gmane.comp.emulators.hercules390.general

RE: Compiling Hercules for 64-bit Windows

"Fish" (David B. Trout — 2012-05-25T01:36:46

[...]
[...]
[...]

Forgive me for asking but I'm just curious: how do you debug Hercules code
using this method?

(i.e. set breakpoints, step through code, etc)

I've personally found it easier to build and especially to debug Hercules
using Visual Studio rather than the command line.

<shrug>

Re: LCS MAC adress mismatch under Linux

jsganino — 2012-05-24T23:12:43


Juergen,

   No need to apologize for being "boring".  Your discovery and comments solved for me a week's worth of frustration in trying to get a TCP/IP connection working from a Hercules guest on a Linux (SLES 11 SP2, a 3.0 kernel) host using ethernet bridging.  It had never occurred to me that the host and guest ends of the tap device would have or could have different MAC addresses (even though I had these maddeningly inexplicable messages in my syslog "br0: received packet on tap0 with own address as source address").

   For the benefit of all, the configuration that worked for me in the end required:

1) In the hercules.cnf file, a relatively bland LCS definition:

0C00.2 LCS -n /dev/net/tun xxx.xxx.xxx.xxx

2) After bringing up TCPIP in my Hercules guest, (which launches hercifc and creates the tap0 device), I adjust its configuration:

ip link show tap0

This reports the MAC address.  Then, I just set the MAC address to a different one, by adding one to the last octet.

ip link set tap0 address xx:

Re: Things to think: Hercules-390 running on mainframe

2012-05-24T16:19:22

I do not agree with your stack, the right one is Linux->KVM->Hercules->zOperating. The KVM will
pass the cpu control to zOperating in the same way z/VM does because the only thing that KVM is doing
is to give access to the hardware feature for virtualization in the CP.

My proposition is to have one Hypervisor Open Source, a kind of thing that I found people trying
to do with QEMU-s390 that I think hercules can do much better because it was built from bottom
up to IBM architecture.

I am not saying that z/VM is a bad, I know and I use it. I guess that port the hercules cpu emulator and memory manager maybe not difficult but I have no idea if I am right, only one developer can issue an opinion and my hope is that one of then can write something here.



--- In hercules-390-hHKSG33TihhbjbujkaE4pw< at >public.gmane.org, "DeWayne Thomas" <DeWayne< at >...> wrote:

Re: Things to think: Hercules-390 running on mainframe

DeWayne Thomas — 2012-05-24T15:33:47

Greetings,

 

I have what might be considered a 'contrarian' view.  My response to this
would be "Why?".  z-Series already has a hypervisor with 45+ years of
enhancements and tuning to run guests as efficiently as possible.

 

As an intellectual exercise or for a diversion I could understand pursing
this path, but as far as KVM on z goes, it strikes me as a solution looking
for a problem.  Even more, it brings to mind Edsger Dijkstra's comment on
PL/1, "belongs more to the problem set than to the solution set".

 

However, inside and outside IBM, KVM is approaching 'religion' status, so
logical arguments have increasingly less weight.

 

Linux->KVM->Linux->Hercules->zOperating System...This is just too many
layers to consider seriously.  Currently the overhead CP imposes to run a
guest is generally < 3%.  I can't imagine improving on that by adding more
layers.

 

All IMHO, of course.

 

DeWayne

 

--- In hercules-390-hHKSG33TihhbjbujkaE4pw< at >public.gmane.org, "jorge.ventura< at >..." <jorge.ventura< at >...>
wro

Re: Re: Compiling Hercules for 64-bit Windows

Mike Schwab — 2012-05-24T03:12:38

Yes.  Intel was forced to adopt the AMD64 (x64) instruction set it
became popular and the Itanium (IA or IA64) did not make any headway
in the consumer market (x86).

On Wed, May 23, 2012 at 8:03 PM, Robert Hodge <quatras.design-/E1597aS9LQAvxtiuMwx3w< at >public.gmane.org> wrote:

Re: Things to think: Hercules-390 running on mainframe

Adam Thornton — 2012-05-24T00:49:47

On May 23, 2012, at 11:25 AM, Ray Mullins wrote:


http://www.fsf.net/~adam/NT-on-390-desktop.png

....and then from a few years later, here I'm running z/VM (64-bit) under z/VM on 31-bit hardware:

http://www.fsf.net/~adam/z900-on-S390-Desktop.png

Adam

[Non-text portions of this message have been removed]

Re: Compiling Hercules for 64-bit Windows

Robert Hodge — 2012-05-24T01:03:51


--- In hercules-390-hHKSG33TihhbjbujkaE4pw< at >public.gmane.org, Roger Bowler <rogerbowler< at >...> wrote:

-[snip]- 

 winbuild\bzip2\x64 and winbuild\pcre\x64 under the Hercules directory.

Even if you have an Intel system?

Re: Things to think: Hercules-390 running on mainframe

peter_flass — 2012-05-23T23:23:12


 
My bias is showing, but I think *one* instance of windoze would be two or three too many.

Re: Things to think: Hercules-390 running on mainframe

peter_flass — 2012-05-23T23:20:23

I don't think you can run VM/370 on current systems, even under VM.  I suppose if you had some business case for running VM/370 instead of newer OSs...

Compiling Hercules for 64-bit Windows

Roger Bowler — 2012-05-23T22:02:16

Here is how to compile a 64-bit Hercules for Windows (x64) using only the 
freely-downloadable Microsoft tools on 64-bit Windows 7:

1. Install Visual C++ 2010 Express from the Visual Studio 2010 Express page:
    http://www.microsoft.com/visualstudio/en-us/products/2010-editions/express

2. Install Microsoft Windows SDK 7.1 from:
    http://www.microsoft.com/en-us/download/details.aspx?id=8279
    IMPORTANT: select all components *except* the Visual C++ Compilers

3. Install Visual C++ 2010 SP1 Compiler Update for the Windows SDK 7.1 from:
    http://www.microsoft.com/en-us/download/details.aspx?id=4422

4. Go to the start menu and choose "All Programs"
    - "Microsoft Windows SDK v7.1" - "Microsoft Windows SDK 7.1 Command Prompt"

5. If you require gzip or bzip2 for disk or tape compression, or if you
    require PCRE for the Hercules Automatic Operator facility, you should
    install the AMD64 versions of these programs in winbuild\zlib\x64
    winbuild\bzip2\x64 and winbuild\pcre\x64 under the Hercules

Re: Re: Things to think: Hercules-390 running on mainframe

Mike Schwab — 2012-05-23T19:12:55

Actually, isn't z/VM out of the emulator business?  Too bad they could
not add emulation of new instructions to allow companies running older
hardware to use new instructions.  Probably not some hardware
capabilities, but certainly end user instructions.

On Wed, May 23, 2012 at 10:53 AM, jorge.ventura-ur4TIblo6goN+BqQ9rBEUg< at >public.gmane.org
<jorge.ventura-ur4TIblo6goN+BqQ9rBEUg< at >public.gmane.org> wrote:

Re: Things to think: Hercules-390 running on mainframe

Ray Mullins — 2012-05-23T16:25:27

2012/5/23 Urtzi Larrieta Alvarez <ularrieta-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org>

 <snip>

Besides, something that could be quite revolutionary is creating an intel
It's already been done and appears to have died - see Mantissa
Corporation's z86VM . They announced it a few years ago at a SHARE
conference and it fizzled due to lack of interest and the new zBX
announcements from IBM. It was emulating the actual x86 (maybe x64 as well)
instruction set, and was actually running native rather than under zLinux.
Unfortunately, their blog has not been updated since March 2010. Maybe
mentioning it here will get a response from Mantissa; I know they have
posted to the various Hercules lists in the past.

Of course, there are already Bochs and other Intel instruction set
emulations. I think it was Adam Thornton who did the (in)famous zVM->Linux
on z->Bochs->Windows 2000 demonstration. I see he's made a brief appearance
in this thread, so he can brag about it here.  :)

Cheers,
Ray

Re: Things to think: Hercules-390 running on mainframe

2012-05-23T15:53:45

I was thinking a replacement for z/VM... Some business can make use mainframes refurbished at low cost but OS is still expensive.

A don't understand z/VM in deep but I think that the concept is similar to other hardware emulators, the bus and devices has still to be emulated or at least partly emulated.

Anyway, I agree that the applicability is narrow. I was reading that there is a fork from QEMU-s390 that the goal is to implement the z/Arch instead of PC but I don't think that it can do better then Hercules.

--- In hercules-390-hHKSG33TihhbjbujkaE4pw< at >public.gmane.org, Adam Thornton <adam< at >...> wrote:

Re: Things to think: Hercules-390 running on mainframe

Jeremy Nicoll - zf yahoo — 2012-05-23T15:27:35


Why would you not just run os images in PR/SM LPARs, with a share of the
real hardware?

Re: Things to think: Hercules-390 running on mainframe

Urtzi Larrieta Alvarez — 2012-05-23T15:12:54

Hi:

Well, my opinion in this fact is quite conservative, that is, creating
another software layer, in this case Hercules, to emulate CPU that is taken
directly from the mainframe CPC, and provide support for devices that are
currently online to the host thanks to the Channel subsystem? I think that
this only sacrifices performance, and has no benefits at all.

It could be reasonable if you have several instances of Hercules running in
the same LPAR like VMware in PC does. But for that, exists z/VM without
performance impact and the posibility of creating thousands of zLinux
images.

Besides, something that could be quite revolutionary is creating an intel
x86-x64 emulator under zLinux, in order to have literally several instances
of MS Windows running under zLinux in an LPAR of the mainframe.

Regards,

Urtzi Larrieta


2012/5/23 jorge.ventura-ur4TIblo6goN+BqQ9rBEUg< at >public.gmane.org <jorge.ventura-ur4TIblo6goN+BqQ9rBEUg< at >public.gmane.org>



[Non-text portions of this message have been removed]



----------

Re: Things to think: Hercules-390 running on mainframe

Adam Thornton — 2012-05-23T15:11:50

The thing is, if you have a business need for real iron, the licensing cost of z/VM is not much of an obstacle.

Things to think: Hercules-390 running on mainframe

2012-05-23T14:50:03

I have been thinking about if there are any good reason to run Hercules-390 on a mainframe and I am guessing that may be a good idea.
Linux is becoming an option as one hypervisor in some platforms, x86, s390. This is thanks to KVM (Kernel Virtual Machine) that
is ready for both plataforms and can manage a virtual CPU and memory for the guest using the hardware support available in the
host CPU.

The point is that the virtualization is for CPU and memory only, for that reason another software is needed to complement the full
virtual hardware with bus, hard disk and other parts. These softwares like QEMU, VMWere or VirtualBox follow the model of PC
what means that only OSs that run on PC (Windows, FreeBSD, QNX and Linux) hardware can make use of this virtual systems.

My think is that Hercules-390 is an emulator for z/Architecture and if part of the code that emulates CPU and memory be replaced
by KVM on s390,  we can have instead of an emulator, a mainframe virtual machine under linux that can be used to run

Re: Window close disable disabled

Dave Wade — 2012-05-21T18:33:46

The later snapshots from http://www.smrcc.org.uk/members/g4ugm/snapshots 
all seem to behave the same ....

Re: Re: Does anyone have Z/series compatiblity macros ?

John P. Hartmann — 2012-05-21T13:17:48

HLASM will generate a list of all machine instructions with operand
formats, if asked to do so.  Maybe you have a friend who can do that for
you.  Of course, you need to somehow specify what is your current
instruction set, but as a start you could pass those instructions to your
assembler and process the resulting listing.  And depending on your
architecture level may take additional optional operands.

There are masses of constraints on the various operands that you can
discover only by reading the Principles of Operation.  Nowadays there are
even operands that must be specified (and even be even), but are ignored.


[Non-text portions of this message have been removed]

Re: Re: Does anyone have Z/series compatiblity macros ?

Binyamin Dissen — 2012-05-21T13:07:25

The LARL macro has to be adjusted to generate an error if the target address
is at an odd address.

On Mon, 21 May 2012 12:57:56 -0000 "Robert Hodge" <quatras.design-/E1597aS9LQAvxtiuMwx3w< at >public.gmane.org>
wrote:

:>The Jay Moseley collection doesn't have everything, but it's a great start.  Thanks so much for bringing this to my attention.
:>
:>I was thinking of the general problem of creating these macros.  It seems that they should/could be generated by a program that took specific information like mnemonic, instruction format, and opcodes from some kind of parameter or specification file, and created the macro in whatever format is needed (with the necessary control cards for DOS or MVS for adding the macros to a library).  If I had the time and ambition, I'd do it myself.  Maybe I will find the time someday to try this.
:>
:>Robert
:>
:>--- In hercules-390-hHKSG33TihhbjbujkaE4pw< at >public.gmane.org, "ScottC" <sccosel< at >...> wrote:
:>>
:>> 
:>> 
:>> --- In hercules-390-hHKSG33TihhbjbujkaE4pw< at >public.gmane.org

Re: Does anyone have Z/series compatiblity macros ?

Robert Hodge — 2012-05-21T12:57:56

The Jay Moseley collection doesn't have everything, but it's a great start.  Thanks so much for bringing this to my attention.

I was thinking of the general problem of creating these macros.  It seems that they should/could be generated by a program that took specific information like mnemonic, instruction format, and opcodes from some kind of parameter or specification file, and created the macro in whatever format is needed (with the necessary control cards for DOS or MVS for adding the macros to a library).  If I had the time and ambition, I'd do it myself.  Maybe I will find the time someday to try this.

Robert

--- In hercules-390-hHKSG33TihhbjbujkaE4pw< at >public.gmane.org, "ScottC" <sccosel< at >...> wrote: