os_unix.c

sqlite 嵌入式数据库的源码

** 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.
  **
  ** To obtain a SHARED lock, a read-lock is obtained on the 'pending
  ** byte'.  If this is successful, a random byte from the 'shared byte
  ** range' is read-locked and the lock on the 'pending byte' released.
  **
  ** A process may only obtain a RESERVED lock after it has a SHARED lock.
  ** A RESERVED lock is implemented by grabbing a write-lock on the
  ** 'reserved byte'. 
  **
  ** A process may only obtain a PENDING lock after it has obtained a
  ** SHARED lock. A PENDING lock is implemented by obtaining a write-lock
  ** on the 'pending byte'. This ensures that no new SHARED locks can be
  ** obtained, but existing SHARED locks are allowed to persist. A process
  ** does not have to obtain a RESERVED lock on the way to a PENDING lock.
  ** This property is used by the algorithm for rolling back a journal file
  ** after a crash.
  **
  ** An EXCLUSIVE lock, obtained after a PENDING lock is held, is
  ** implemented by obtaining a write-lock on the entire 'shared byte
  ** range'. Since all other locks require a read-lock on one of the bytes
  ** within this range, this ensures that no other locks are held on the
  ** database. 
  **
  ** The reason a single byte cannot be used instead of the 'shared byte
  ** range' is that some versions of windows do not support read-locks. By
  ** locking a random byte from a range, concurrent SHARED locks may exist
  ** even if the locking primitive used is always a write-lock.
  */
  int rc = SQLITE_OK;
  struct lockInfo *pLock = id->pLock;
  struct flock lock;
  int s;

  assert( id->isOpen );
  TRACE7("LOCK    %d %s was %s(%s,%d) pid=%d\n", id->h, locktypeName(locktype), 
      locktypeName(id->locktype), locktypeName(pLock->locktype), pLock->cnt
      ,getpid() );

  /* If there is already a lock of this type or more restrictive on the
  ** OsFile, do nothing. Don't use the end_lock: exit path, as
  ** sqlite3OsEnterMutex() hasn't been called yet.
  */
  if( id->locktype>=locktype ){
    TRACE3("LOCK    %d %s ok (already held)\n", id->h, locktypeName(locktype));
    return SQLITE_OK;
  }

  /* Make sure the locking sequence is correct
  */
  assert( id->locktype!=NO_LOCK || locktype==SHARED_LOCK );
  assert( locktype!=PENDING_LOCK );
  assert( locktype!=RESERVED_LOCK || id->locktype==SHARED_LOCK );

  /* This mutex is needed because id->pLock is shared across threads
  */
  sqlite3OsEnterMutex();

  /* If some thread using this PID has a lock via a different OsFile*
  ** handle that precludes the requested lock, return BUSY.
  */
  if( (id->locktype!=pLock->locktype && 
          (pLock->locktype>=PENDING_LOCK || locktype>SHARED_LOCK))
  ){
    rc = SQLITE_BUSY;
    goto end_lock;
  }

  /* If a SHARED lock is requested, and some thread using this PID already
  ** has a SHARED or RESERVED lock, then increment reference counts and
  ** return SQLITE_OK.
  */
  if( locktype==SHARED_LOCK && 
      (pLock->locktype==SHARED_LOCK || pLock->locktype==RESERVED_LOCK) ){
    assert( locktype==SHARED_LOCK );
    assert( id->locktype==0 );
    assert( pLock->cnt>0 );
    id->locktype = SHARED_LOCK;
    pLock->cnt++;
    id->pOpen->nLock++;
    goto end_lock;
  }

  lock.l_len = 1L;
  lock.l_whence = SEEK_SET;

  /* A PENDING lock is needed before acquiring a SHARED lock and before
  ** acquiring an EXCLUSIVE lock.  For the SHARED lock, the PENDING will
  ** be released.
  */
  if( locktype==SHARED_LOCK 
      || (locktype==EXCLUSIVE_LOCK && id->locktype<PENDING_LOCK)
  ){
    lock.l_type = (locktype==SHARED_LOCK?F_RDLCK:F_WRLCK);
    lock.l_start = PENDING_BYTE;
    s = fcntl(id->h, F_SETLK, &lock);
    if( s ){
      rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY;
      goto end_lock;
    }
  }


  /* If control gets to this point, then actually go ahead and make
  ** operating system calls for the specified lock.
  */
  if( locktype==SHARED_LOCK ){
    assert( pLock->cnt==0 );
    assert( pLock->locktype==0 );

    /* Now get the read-lock */
    lock.l_start = SHARED_FIRST;
    lock.l_len = SHARED_SIZE;
    s = fcntl(id->h, F_SETLK, &lock);

    /* Drop the temporary PENDING lock */
    lock.l_start = PENDING_BYTE;
    lock.l_len = 1L;
    lock.l_type = F_UNLCK;
    fcntl(id->h, F_SETLK, &lock);
    if( s ){
      rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY;
    }else{
      id->locktype = SHARED_LOCK;
      id->pOpen->nLock++;
      pLock->cnt = 1;
    }
  }else if( locktype==EXCLUSIVE_LOCK && pLock->cnt>1 ){
    /* We are trying for an exclusive lock but another thread in this
    ** same process is still holding a shared lock. */
    rc = SQLITE_BUSY;
  }else{
    /* The request was for a RESERVED or EXCLUSIVE lock.  It is
    ** assumed that there is a SHARED or greater lock on the file
    ** already.
    */
    assert( 0!=id->locktype );
    lock.l_type = F_WRLCK;
    switch( locktype ){
      case RESERVED_LOCK:
        lock.l_start = RESERVED_BYTE;
        break;
      case EXCLUSIVE_LOCK:
        lock.l_start = SHARED_FIRST;
        lock.l_len = SHARED_SIZE;
        break;
      default:
        assert(0);
    }
    s = fcntl(id->h, F_SETLK, &lock);
    if( s ){
      rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY;
    }
  }
  
  if( rc==SQLITE_OK ){
    id->locktype = locktype;
    pLock->locktype = locktype;
  }else if( locktype==EXCLUSIVE_LOCK ){
    id->locktype = PENDING_LOCK;
    pLock->locktype = PENDING_LOCK;
  }

end_lock:
  sqlite3OsLeaveMutex();
  TRACE4("LOCK    %d %s %s\n", id->h, locktypeName(locktype), 
      rc==SQLITE_OK ? "ok" : "failed");
  return rc;
}

/*
** Lower the locking level on file descriptor id to locktype.  locktype
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
**
** It is not possible for this routine to fail if the second argument
** is NO_LOCK.  If the second argument is SHARED_LOCK, this routine
** might return SQLITE_IOERR instead of SQLITE_OK.
*/
int sqlite3OsUnlock(OsFile *id, int locktype){
  struct lockInfo *pLock;
  struct flock lock;
  int rc = SQLITE_OK;

  assert( id->isOpen );
  TRACE7("UNLOCK  %d %d was %d(%d,%d) pid=%d\n", id->h, locktype, id->locktype, 
      id->pLock->locktype, id->pLock->cnt, getpid());

  assert( locktype<=SHARED_LOCK );
  if( id->locktype<=locktype ){
    return SQLITE_OK;
  }
  sqlite3OsEnterMutex();
  pLock = id->pLock;
  assert( pLock->cnt!=0 );
  if( id->locktype>SHARED_LOCK ){
    assert( pLock->locktype==id->locktype );
    if( locktype==SHARED_LOCK ){
      lock.l_type = F_RDLCK;
      lock.l_whence = SEEK_SET;
      lock.l_start = SHARED_FIRST;
      lock.l_len = SHARED_SIZE;
      if( fcntl(id->h, F_SETLK, &lock)!=0 ){
        /* This should never happen */
        rc = SQLITE_IOERR;
      }
    }
    lock.l_type = F_UNLCK;
    lock.l_whence = SEEK_SET;
    lock.l_start = PENDING_BYTE;
    lock.l_len = 2L;  assert( PENDING_BYTE+1==RESERVED_BYTE );
    fcntl(id->h, F_SETLK, &lock);
    pLock->locktype = SHARED_LOCK;
  }
  if( locktype==NO_LOCK ){
    struct openCnt *pOpen;

    /* Decrement the shared lock counter.  Release the lock using an
    ** OS call only when all threads in this same process have released
    ** the lock.
    */
    pLock->cnt--;
    if( pLock->cnt==0 ){
      lock.l_type = F_UNLCK;
      lock.l_whence = SEEK_SET;
      lock.l_start = lock.l_len = 0L;
      fcntl(id->h, F_SETLK, &lock);
      pLock->locktype = NO_LOCK;
    }

    /* Decrement the count of locks against this same file.  When the
    ** count reaches zero, close any other file descriptors whose close
    ** was deferred because of outstanding locks.
    */
    pOpen = id->pOpen;
    pOpen->nLock--;
    assert( pOpen->nLock>=0 );
    if( pOpen->nLock==0 && pOpen->nPending>0 ){
      int i;
      for(i=0; i<pOpen->nPending; i++){
        close(pOpen->aPending[i]);
      }
      sqliteFree(pOpen->aPending);
      pOpen->nPending = 0;
      pOpen->aPending = 0;
    }
  }
  sqlite3OsLeaveMutex();
  id->locktype = locktype;
  return rc;
}

/*
** Close a file.
*/
int sqlite3OsClose(OsFile *id){
  if( !id->isOpen ) return SQLITE_OK;
  sqlite3OsUnlock(id, NO_LOCK);
  if( id->dirfd>=0 ) close(id->dirfd);
  id->dirfd = -1;
  sqlite3OsEnterMutex();
  if( id->pOpen->nLock ){
    /* If there are outstanding locks, do not actually close the file just
    ** yet because that would clear those locks.  Instead, add the file
    ** descriptor to pOpen->aPending.  It will be automatically closed when
    ** the last lock is cleared.
    */
    int *aNew;
    struct openCnt *pOpen = id->pOpen;
    pOpen->nPending++;
    aNew = sqliteRealloc( pOpen->aPending, pOpen->nPending*sizeof(int) );
    if( aNew==0 ){
      /* If a malloc fails, just leak the file descriptor */
    }else{
      pOpen->aPending = aNew;
      pOpen->aPending[pOpen->nPending-1] = id->h;
    }
  }else{
    /* There are no outstanding locks so we can close the file immediately */
    close(id->h);
  }
  releaseLockInfo(id->pLock);
  releaseOpenCnt(id->pOpen);
  sqlite3OsLeaveMutex();
  id->isOpen = 0;
  TRACE2("CLOSE   %-3d\n", id->h);
  OpenCounter(-1);
  return SQLITE_OK;
}

/*
** Turn a relative pathname into a full pathname.  Return a pointer
** to the full pathname stored in space obtained from sqliteMalloc().
** The calling function is responsible for freeing this space once it
** is no longer needed.
*/
char *sqlite3OsFullPathname(const char *zRelative){
  char *zFull = 0;
  if( zRelative[0]=='/' ){
    sqlite3SetString(&zFull, zRelative, (char*)0);
  }else{
    char zBuf[5000];
    zBuf[0] = 0;
    sqlite3SetString(&zFull, getcwd(zBuf, sizeof(zBuf)), "/", zRelative,
                    (char*)0);
  }
  return zFull;
}


#endif /* SQLITE_OMIT_DISKIO */
/***************************************************************************
** Everything above deals with file I/O.  Everything that follows deals
** with other miscellanous aspects of the operating system interface
****************************************************************************/


/*
** Get information to seed the random number generator.  The seed
** is written into the buffer zBuf[256].  The calling function must
** supply a sufficiently large buffer.
*/
int sqlite3OsRandomSeed(char *zBuf){
  /* We have to initialize zBuf to prevent valgrind from reporting
  ** errors.  The reports issued by valgrind are incorrect - we would
  ** prefer that the randomness be increased by making use of the
  ** uninitialized space in zBuf - but valgrind errors tend to worry
  ** some users.  Rather than argue, it seems easier just to initialize
  ** the whole array and silence valgrind, even if that means less randomness
  ** in the random seed.
  **
  ** When testing, initializing zBuf[] to zero is all we do.  That means
  ** that we always use the same random number sequence.* This makes the
  ** tests repeatable.
  */
  memset(zBuf, 0, 256);
#if !defined(SQLITE_TEST)
  {
    int pid, fd;
    fd = open("/dev/urandom", O_RDONLY);
    if( fd<0 ){
      time((time_t*)zBuf);
      pid = getpid();
      memcpy(&zBuf[sizeof(time_t)], &pid, sizeof(pid));
    }else{
      read(fd, zBuf, 256);
      close(fd);
    }
  }
#endif
  return SQLITE_OK;
}

/*
** Sleep for a little while.  Return the amount of time slept.
*/
int sqlite3OsSleep(int ms){
#if defined(HAVE_USLEEP) && HAVE_USLEEP
  usleep(ms*1000);
  return ms;
#else
  sleep((ms+999)/1000);
  return 1000*((ms+999)/1000);
#endif
}

/*
** Static variables used for thread synchronization
*/
static int inMutex = 0;
#ifdef SQLITE_UNIX_THREADS
static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
#endif

/*
** The following pair of routine implement mutual exclusion for
** multi-threaded processes.  Only a single thread is allowed to
** executed code that is surrounded by EnterMutex() and LeaveMutex().
**
** SQLite uses only a single Mutex.  There is not much critical
** code and what little there is executes quickly and without blocking.
*/
void sqlite3OsEnterMutex(){
#ifdef SQLITE_UNIX_THREADS
  pthread_mutex_lock(&mutex);
#endif
  assert( !inMutex );
  inMutex = 1;
}
void sqlite3OsLeaveMutex(){
  assert( inMutex );
  inMutex = 0;
#ifdef SQLITE_UNIX_THREADS
  pthread_mutex_unlock(&mutex);
#endif
}

/*
** The following variable, if set to a non-zero value, becomes the result
** returned from sqlite3OsCurrentTime().  This is used for testing.
*/
#ifdef SQLITE_TEST
int sqlite3_current_time = 0;
#endif

/*
** Find the current time (in Universal Coordinated Time).  Write the
** current time and date as a Julian Day number into *prNow and
** return 0.  Return 1 if the time and date cannot be found.
*/
int sqlite3OsCurrentTime(double *prNow){
  time_t t;
  time(&t);
  *prNow = t/86400.0 + 2440587.5;
#ifdef SQLITE_TEST
  if( sqlite3_current_time ){
    *prNow = sqlite3_current_time/86400.0 + 2440587.5;
  }
#endif
  return 0;
}

#endif /* OS_UNIX */