In a message dated 8/1/2006 10:09:19 A.M. Central Daylight Time,
[EMAIL PROTECTED] writes:
>Hum, are you sure that "device numbers are canonical across systems"?
>Granted, it is wise to make it so. But I know how to generate a
>different device number for a given physical device in two different
>LPARs. It is especially easy to change the two high-order numbers.
No, I am not sure. That is why I couched my comment with my caveat ("At
least that was the response I got when I asked IBM-MAIN if they were many
months ago."). But there must be something canonical across systems that is
unique to the device for a cross-system ENQ involving devices to work.
Certainly
it cannot be the address of a UCB.
I asked an IBM IOS internals person a year ago, and he said it is possible
to have different device numbers on different LPARs that reach the same
physical device. I asked an IBM control unit person and he said it is not
possible, but I think he was talking about the one-byte channel connection
address
within the control unit and did not understand terminology describing things
outside of the control unit, such as processor architecture. Then I asked
IBM-MAIN, and everyone who responded said the numbers are canonical and must
be
the same. Please email me offline so I can find out how to generate different
device numbers for the same device for an experiment.
Bill Fairchild
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
