MQSeries.net :: View topic

RogerLacroix · Posted: Mon Mar 13, 2006 6:14 am Post subject:

Attention IBM Hurlsey,

Setup: Windows XP Pro SP2 & WMQ v6.0.1.0 (Fresher Pack 01)

Over the last couple of weeks, I have been desperately trying to get the client-side security exit of MQAUSX to work with MQ Explorer v6 (Eclipse based).

No matter what I tried, I could not get the Java client-side security exit to work with MQ Explorer v6. I tried every combination using Client Channel Table to get it to work but with no luck.

What I don't understand is for IBM's WBIMB Eclipse Tool Kit it is a very simple setup. But for an IBM supplied MQ tool, it is extremely complicated. (Yes, irony.)

Anyway, while reading the MQ docs for a while, I found a new Java class called: MQExternalSecurityExit. This got me thinking why would anyone want to call a native DLL or shared object from a Java program when you could just as easily call a Java written security exit. Then it hit me, maybe IBM added MQExternalSecurityExit to have backwards compatibility between MQ Explorer v5.* (MMC) and MQ Explorer v6 (Eclipse).

So I put in the name of the Windows DLL for the MQAUSX client-side security exit in the Client Channel Table that MQ Explorer v6 was using and bingo, it worked (sort of). Being surprised that it worked is an understatement. It is fine that IBM has attempted to add backwards compatibility but what about using a simple Java written security exit like WBIMB Eclipse Tool Kit?

Now to my 'sort of worked' comment: Here's how the MQAUSX client-side security exits works under programs like MO71 (MQMon), MQ Explorer v5 or other programs using Client Channel Table:

- When the client-side security exit is invoked for MQXR_SEC_MSG it receives a buffer with DataLength of 250 and AgentBufferLength of 3968.
- The client-side exit responds by updating the AgentBuffer with 300 bytes. (There is code to allocate a new buffer if the AgentBuffer is too small.)

Here's what happens under MQ Explorer v6.
- When the client-side security exit is invoked for MQXR_SEC_MSG it receives a buffer with DataLength of 250 and AgentBufferLength of 250.
- The client-side exit responds by allocating a new buffer of 300 bytes, fills the new buffer with the data and it sets ExitResponse2 with MQXR2_USE_EXIT_BUFFER.
- The server-side receives the new AgentBuffer, processes it but when it responds, the client crashes. I don't mean my client code because it is not invoked again but rather the MQ client library.

At first I thought it was my code. So, I set the conditional test to also always allocate a new buffer and set ExitResponse2 with MQXR2_USE_EXIT_BUFFER. This works perfectly with MO71, MQ Explorer v5, etc... but always fails with MQ Explorer v6.

Does anyone at IBM have any idea what the problem is? (Other than a bug in MQExternalSecurityExit). I supposed I going to have to open a PMR?

Of course in a really weird twist, if I setup the server-side and client-side to have the first buffer to be 300 bytes (or larger) then everything works perfectly (because the client-side can then reuse the AgentBuffer).

But this solution is totally impractical because that would mean my customers would have to upgrade all client-side and server-side at the same time. (Plus sending 50 bytes of filler is a really stupid idea anyway!) Several customers have rather large number of installations of MQAUSX and if I propose this, then they would want someone's head. (And I really don't think it should be mine!!)

So, I only see one easy way out. Does anyone at IBM know how to use a Java written client-side security exit with MQ Explorer v6?? (Documented or undocumented? - Supported or unsupported? At this point, I really don't care.) Please someone at IBM give me a tip, trick, etc..

Regards,
Roger Lacroix
Capitalware Inc.
http://www.capitalware.biz/
_________________
Capitalware: Transforming tomorrow into today.
Connected to MQ!
Twitter