On Friday, 03/30/2007 at 10:54 ZE2, Shimon Lebowitz <[EMAIL PROTECTED]> 
wrote:
> >
> > I tested this years ago when we encountered problems with messages 
that
> > seemed to get lost
> >
> But is it documented?
> Things from 'years ago' sometimes change with no
> announcement when they are undocumented internals.
> 
> I am not arguing with you, I myself have no vague idea,
> just wondering if this old 'feature' is guaranteed to still
> be true.

>From Ch. 16 (*MSG) of the CP Programming Services book:
"The Message system service uses the IUCV default value of 255 for the 
number of
outstanding messages allowed on the path unless the user specifies a value 
for the
MSGLIM parameter. If a value is specified on the IUCV CONNECT for MSGLIM 
to
*MSG, that value is used for the number of messages allowed on the path. 
If this
message limit is exceeded, any additional incoming messages are routed 
directly to
the virtual machine console or alternate console, and the virtual machine 
is not
notified about these messages."

But because WAKEUP reads the message in immediately and queues it in CMS, 
even when you're not in WAKEUP at the moment, the CP message limit only 
comes into play when
(a) Your virtual machine is disabled for external interrupts
(b) The rate of message arrival is faster than the virtual machine running 
WAKEUP can read & queue them

I remember some of the original attempts at IBMers writing *MSG handlers. 
They only did the IUCV RECEIVE when you issued their "get next message" 
command.  People who used those programs were often victimized by the 
MSGLIM.

Alan Altmark
z/VM Development
IBM Endicott

Reply via email to