Sometimes it is necessary to lock a file at the disk level. Since there is no
native Java way of locking a file, we have to implement our own.
Unfortunately, it appears as though we can only assume that rename is atomic.
We base the locking on this then: rename the lock file and update it with the
lock owners. How it works: for every file we want to lock we create an _lock
file. This file contains information about who is locking the file, and in
what mode. (Single writer, multiple readers.) To guarantee atomicity of
update, we move (rename) the actual file each time we want to lock it and
update the lock file. When this is done, we move (rename) it back. Almost
like a two-phase commit protocol! Currently we don't support re-entrant
- Mark Little (firstname.lastname@example.org)
- $Id: FileLock.java 2342 2006-03-30 13:06:17Z $
- JTS 1.0.
public static final int F_RDLCK = 0;
public static final int F_WRLCK = 1;
+ ", " + timeout + ", " + retry + " )");
public boolean lock (int lmode)
return lock(lmode, false);
public synchronized boolean lock (int lmode, boolean create)
public synchronized boolean unlock ()
int number = 0, mode = 0;
for (int i = 0; i < _retry; i++)