[infinispan-dev] Semaphore vs Lock

[Dan Berindei]

On Tue, Mar 13, 2012 at 2:53 PM, Galder Zamarreño <galder at redhat.com> wrote:
>
> On Mar 8, 2012, at 12:42 PM, Dan Berindei wrote:
>
>> On Wed, Mar 7, 2012 at 2:39 PM, Sanne Grinovero <sanne at infinispan.org> wrote:
>>> On 7 March 2012 12:05, Galder Zamarreño <galder.zamarreno at redhat.com> wrote:
>>>> Hi,
>>>>
>>>> I was reading up about Java's Semaphores (http://docs.oracle.com/javase/6/docs/api/java/util/concurrent/Semaphore.html) and a couple of ideas came to my mind:
>>>>
>>>> 1. Wouldn't it make sense to use binary semaphores instead of locks in Infinispan? We're already having to override ReentrantLock in order to have locks owned by Transactions rather than threads. Initially I thought it might make easier for deadlock detection, but not so sure right now cos we're already changing things to avoid thread ownership of locks.
>>>>
>>
>> We don't support most of the Lock operations, so I think it would be
>> fair to remove 'implements Lock' from the OwnableReentrantLock
>> declaration. But we can't remove the reentrant part, as we acquire the
>> lock when we put a value in L1 in DistributionInterceptor - after we
>> have already acquired the lock once in LockInterceptor (that's before
>> we even consider a pessimistic transaction doing multiple puts on the
>> same key).
>
> True, but reentrant is only needed for non-transactional scenarios. For pessimistric transactions, we have ownable locks where they're owned by transactions.
>

I meant reentrant in a more general sense - OwnableReentrantLock is
reentrant because it allows the same transaction to lock the same key
twice, even if it happens on different threads.

In the pessimistic case you could even have a transaction doing lock +
read + write + unlock + commit and I believe the key should stay
locked until the commit (because the update command implicitly
acquired the lock a second time).

I'm not sure about optimistic transactions - we might get away without
reentrant locks there. But I think there's no point adding a
non-reentrant lock for optimistic transactions if we still need
reentrant locks in the other cases.

Cheers
Dan