[jcifs] Re: Another fix for Transport deadlock issue ?

[Ronny Schuetz]

Hi Tim,

> We tried using the 1.2.9 release of jcifs, but noticed throughput is
> significantly lower due to the new Transport.setupDiscoLock monitor. We
> were only experience the deadlock when a Transport experiences an
> IOException:

The fix basically addresses the symptom only, it does not fix the real
issue. I didn't had the time to do that and to be honest I didn't wanted
to dive that deep into the code. So the fix is working, but it is
definitely not optimal - I'd really like to see the real issue fixed.
Maybe Michael can do that some day, as he has the complete overview.

The fix works for us and the performance is fine as well, but we might
have a different use case than you, we're very seldom accessing a single
server with multiple threads.

It was required to put the additional synchronization blocks to exactly
the places they're now as I could easily reproduce deadlocks otherwise
(the test was accessing a single Windows share with 200 threads; which
is maybe not a real use case). Avoiding the deadlocks in all cases was
and is a bit more important for me than performance. Sometimes we've to
trade performance for stability.

Best regards,
Ronny