[jcifs] Exceptions in AndXServerMessageBlock Causing Timeout Errors

[Michael B Allen]

> I am running version 0.7.1 of JCIFS.  It has been working fine since
> January, but lately I've been getting null pointer exceptions in the
> AndXServerMessageBlock class that apparently causes the socket reading
> thread to die and a timeout exception to be thrown.  This appears to
> happen
> at random, but about once every week.  None of this should matter, but for
> background:  It only happens when transferring large files ~200mb+ to an
> EMC
> Celerra.  This error hasn't appeared when transferring to any other
> servers,
> however, the file sizes are relatively small ~10mb max.
>
> Anyone have any ideas on cause/troubleshooting/etc?

It looks like you're high-level description is correct Charles. Somehow
the InputStream 'in' is being closed and set to null in tryClose at the
same time another operation comes in. Since 'in' is null the response to
the second operation causes the null pointer exception and eventually the
request times out.

> Is this a known issue that has been fixed by a later release of the
> library?

We have not had reports of this particular error but a serious deadlock
was discovered in SmbTransport in May. It was fixed in a way that
sigificantly changed the overall locking of the transport layer. The
previous locking scheme was flawed. Updating to 0.7.11 would probably
solve this problem.

> I don't want to upgrade/regression test unless this has been fixed.

I really think you should update Charles. There is another very serious
flaw in 0.7.1 regarding writing. See the News section on our homepage. I
don't think you will have any surprises from 0.7.11. The changes to the
smb package have all been fixes. No new features were added. The 0.7.11
release is very solid at this point.

Mike