[jcifs] Re: jCIFS deadlock issue - when can we expect a fix?

Michael B Allen mba2000 at ioplex.com
Mon Feb 27 16:15:35 GMT 2006


Yup. I'll add it to The List.

Thanks,
Mike

On Mon, 27 Feb 2006 12:14:25 +0100
Ronny Schuetz <usenet-01 at groombridge34.de> wrote:

> Hi Mike,
> 
> > No. I don't think so. All 75 threads should use the same socket to
> > multiplex requests (unless the server periodically closes sockets). If
> > you set ssnLimit to 25 then you should see ~3 sockets.
> 
> Thanks for the explanation. Thats helpful to know.
> 
> After some additional stability tests with 200 threads accessing a
> single windows share and another stability test of our application it
> looks like the deadlock issue is fixed now. In addition to the places
> already communicated, I had to synchronize SmbTree#treeConnect() in the
> same way:
> 
> void treeConnect(ServerMessageBlock andx,
>                  ServerMessageBlock andxResponse ) throws SmbException
> {
>  synchronized(session.transport()._oSetupDisconnectMutex)
>  {
>   // ... add original message body here ...
>  }
> }
> 
> Maybe you can add this change (or a better one) to the next jCIFS
> version. The performance does not seem to be impacted heavily.
> 
> Best regards,
> Ronny
> 


More information about the jcifs mailing list