[concurrency-interest] The atomicity of StampedLock lock conversion

Doug Lea dl at cs.oswego.edu
Fri Oct 10 09:43:23 EDT 2014

On 10/10/2014 06:12 AM, Chris Vest wrote:
> One might, without further research, come to be ostensibly reasonable conclusion
> that a read-write lock that offers lock conversion, would be able to convert a
> write-lock into a read-lock atomically.


> Reading the documentation for the StampedLock#tryConvertToRead, we’re told that
> if given a write-lock stamp, the method will first release that write-lock, and
> /then/ obtain a read-lock, thus leaving a window for someone else to grab the
> lock ahead of us.

The word "then" doesn't appear in specs. Perhaps we should clarify
by adding "atomically" in:

      * If the lock state matches the given stamp, performs one of
      * If the lock state matches the given stamp, atomically performs one of

      * the following actions. If the stamp represents holding a write
      * lock, releases it and obtains a read lock.  Or, if a read lock,
      * returns it. Or, if an optimistic read, acquires a read lock and
      * returns a read stamp only if immediately available. This method
      * returns zero in all other cases.

And similarly for the others.


More information about the Concurrency-interest mailing list