[Info-vax] CentOS has been effectively killed for production use

[David Wade]

On 11/12/2020 00:38, Dennis Boone wrote:
> Dave, I think you've missed some things:
> 
>   > Well lets look at the definitive "Principles of Operations". This is the
>   > bible for what the hardware does. (you will need an IBM ID to read this
>   > but they are free)
> 
>   > it goes into great detail on how to test the processor type but no where
>   > does it mention any difference in the instructions implemented.
> 
> First, giving away the details of the implementation would not be the
> best customer control move.  I'm sure just saying something is missing
> isn't the whole story.
> 
> Second, the features "depended on" (possibly actually used, maybe just
> tested for) by z/OS don't have to be a documented part of the
> architecture at all.
> 
>   > All PUs on a z14 server are physically identical. When the system is
>   > initialized, one integrated firmware processor (IFP) is allocated from
>   > the pool of PUs that is available for the entire system. The other PUs
>   > can be characterized to specific functions (CP, IFL, ICF, zIIP, or SAP).
> 
>   > The function that is assigned to a PU is set by the Licensed Internal
>   > Code (LIC). The LIC is  loaded when the system is initialized at
>   > power-on reset (POR) and the PUs are characterized. Only characterized
>   > PUs include a designated function. Non-characterized PUs are considered
>   > spares. Order at least one CP, IFL, or ICF on a z14 server.
> 
> Yes, the silicon is supposedly identical, and I see no reason to doubt
> that.  Various things I've read over the years indicate that the silicon
> is fairly strongly related to the Power architecture.
> 
> The above doesn't support your contention.  It says that the LIC
> ("microcode") is different for different types of engines.  That means
> that different engines have different behaviors in execution.  Remember
> that 370 family machines have effectively been emulators since the early
> 70s.
> 
> I've heard the VM will actually run on an IFL.  z/OS is the crown
> jewels. :)
> 
> De
> 
I believe that zOS will run on an IFL. It doesn't because it tests the 
processor types and feature bits. I also think folks have had it run on 
an IFL.

I know from the Hercules emulator project that zOS is very picky about 
what it will run on. So newer versions won't run on old hardware. They 
check the CPU type and feature bits.

Hercules allows you to pretend to have features you don't have. Often if 
you enable these zOS runs just fine.

So IBM has no need to disable instructions and keep it secret. It can 
lock zOS using the feature bits.

Disabling "one or two" instructions would also be fairly pointless. Its 
relativity simple to fix. When a "z" box (or any of its predecessors) 
executes an undefined instruction a particular behaviours occurs and a 
fault handler is loaded. The fault handler can then simulate the missing 
instructions.

This was used back in the day to simulate floating point or commercial 
instructions that were options. It still works to this day.

You also say zVM will run. It does, as does any assembler that you write 
for zVM. So if there is a missing instruction it must be invalid in a 
zVM virtual machine running on an iFL.


Dave