Re: IBM S/370 and binutils

Re: IBM S/370 and binutils

[David and Jannette Uczen]

If EM_I370 is different from Amdahl's ABI, wouldn't it be better to change
i370 to conform to Amdahl's?

--David

-----Original Message-----
From: H.J. Lu <hjl@lucon.org>
To: Ian Lance Taylor <ian@zembu.com>
Cc: linas@linas.org <linas@linas.org>; uczen@mint.net <uczen@mint.net>;
binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>
Date: Friday, October 29, 1999 1:05 PM
Subject: Re: IBM S/370 and binutils


>>
>> Talk to registry@sco.com.  They maintain the lists of official ELF
>> machine numbers.  They'll give you a number if you ask, although
>> sometimes you have to ask a few times.  It may help to pull in
>> somebody from IBM; they seem to take big companies more seriously than
>> free software people.
>
>I am looking at the draft of the next SVR4 gABI. 9 is reserved. In the
>previous email from Joel at SCO, it will be used for EM_S370.
>
>>
>> First, though, is there some reason that it is inappropriate to use
>> EM_S370?  I don't know anything about the 370, so that may be a stupid
>> question.
>
>Linas, does Linux/i370 use the same ABI as defined by Amdahl's EM_S370?
>If we do, we can use it. Otherwise, we have to apply for a new EM_I370.
>
>
>--
>H.J. Lu (hjl@gnu.org)
>
>

Re: IBM S/370 and binutils

[linas]

Hi David,

It's been rumoured that David and Jannette Uczen said:
> 
> If EM_I370 is different from Amdahl's ABI, wouldn't it be better to change
> i370 to conform to Amdahl's?

When I wrote this thing, I was not aware that Amdahl had defined an
ELF ABI.  I would like to ask you to help me get specs for the thing.
Certainly, I can't make any changes without the specs.

If Amdahl's ABI is old, inefficient, or doesn't support dynamic
linking/shared libs, then this would be an occasion to abandon old
problems.   This does not mean that you or I or someone else could not
also create the parts needed to adhere to Amdahl's ABI and add them to
gcc & gas. It just means that there is room for two ABI's if that 
seems appropriate.

If I get a copy of the spec's, I will pour through them and try
to match what they've done with what I have in place.  If it seems 
close, I'll volunteer to make at least the initial changes to bring 
them in line.  In order of effort, from least to greatest:

-- stack layout
-- static linkage and register conventions
-- dynamic linkage mechanisms
-- new assembler directives
-- system call compatibility

Let me know.

--linas

Re: IBM S/370 and binutils

[David and Jannette Uczen]

I am willing to help in any way I can. I will contact Chris Swartout again
and see if I can get an ABI.

--David

-----Original Message-----
From: linas@linas.org <linas@linas.org>
To: David and Jannette Uczen <uczen@mint.net>
Cc: linas@linas.org <linas@linas.org>; ian@zembu.com <ian@zembu.com>;
binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>;
cas30@oes.amdahl.com <cas30@oes.amdahl.com>; LINUX-VM@VM.MARIST.EDU
<LINUX-VM@VM.MARIST.EDU>
Date: Saturday, October 30, 1999 2:45 PM
Subject: Re: IBM S/370 and binutils


>Hi David,
>
>It's been rumoured that David and Jannette Uczen said:
>>
>> You should be able to get an ABI from Amdahl Corporation, through UTS
>> Technical Support. I have contacted a Chris A. Swartout
>> (cas30@oes.amdahl.com), who verified that the value of EM_S370 (9) is
used
>> for IBM System/370 Processors. This value still exists in the binutils
>> source (/binutils/include/elf/common.h).
>> I have had the value officially registered with SCO (registry@sco.com). A
>> new ABI is going to be published soon.
>> I myself am trying to get ABI information, but haven't heard back yet.
>
>You apologized for a long overview in your last note, so let me do
>the same here.
>
>Be aware that there are multiple ABI specs.  The one kept by SCO
>outlines the structure of the ELF file format that is common to *all*
>ELF-based architectures.  You can download this spec as a PDF file
>from the SCO website; my webpage gives the pointer to it.  The binutils
>contain a back-end that knows how to manipulate this geenric file
>format.
>
>Seperately, there are processor-specific ABI's that define things like
>the stack layout, the static linkage and register conventions, the
>dynamic linkage mechanisms, and occasionally even the system call
>interfaces.  I beleive SCO maintains the Intel ABI, Sun maintains the
>PowerPC ABI, and SGI the MIPS ABI.  It is this later document, that
>hopefully Amdahl maintains for the IBM processors (isn't it a curious
>accident of fate that processor ABI's seem to be maintained by
>competitors?).
>
>In the gnu scheme of things, stack layout, register conventions are
>handled by the compiler gcc.  Dynamic linking is handled mostly by
>the binutils ld loader, with assist from gcc.  The system call
>interfaces are a shared responsibility between the C Library glibc
>and the kernel (Linux or UTS in Amdahl's case).  As you can see,
>ABI conformance has prevasive effects, and full, 100% conformance
>can be a rather overwhelming task, especially when subtlties with
>glibc and the OS kernel enter the picture.
>
>--linas
>
>

Re: IBM S/370 and binutils

[linas]

Hi David,

It's been rumoured that David and Jannette Uczen said:
> 
> You should be able to get an ABI from Amdahl Corporation, through UTS
> Technical Support. I have contacted a Chris A. Swartout
> (cas30@oes.amdahl.com), who verified that the value of EM_S370 (9) is used
> for IBM System/370 Processors. This value still exists in the binutils
> source (/binutils/include/elf/common.h).
> I have had the value officially registered with SCO (registry@sco.com). A
> new ABI is going to be published soon.
> I myself am trying to get ABI information, but haven't heard back yet.

You apologized for a long overview in your last note, so let me do
the same here.

Be aware that there are multiple ABI specs.  The one kept by SCO 
outlines the structure of the ELF file format that is common to *all*
ELF-based architectures.  You can download this spec as a PDF file
from the SCO website; my webpage gives the pointer to it.  The binutils
contain a back-end that knows how to manipulate this geenric file
format.

Seperately, there are processor-specific ABI's that define things like
the stack layout, the static linkage and register conventions, the
dynamic linkage mechanisms, and occasionally even the system call
interfaces.  I beleive SCO maintains the Intel ABI, Sun maintains the
PowerPC ABI, and SGI the MIPS ABI.  It is this later document, that 
hopefully Amdahl maintains for the IBM processors (isn't it a curious 
accident of fate that processor ABI's seem to be maintained by
competitors?).

In the gnu scheme of things, stack layout, register conventions are
handled by the compiler gcc.  Dynamic linking is handled mostly by
the binutils ld loader, with assist from gcc.  The system call
interfaces are a shared responsibility between the C Library glibc 
and the kernel (Linux or UTS in Amdahl's case).  As you can see,
ABI conformance has prevasive effects, and full, 100% conformance
can be a rather overwhelming task, especially when subtlties with 
glibc and the OS kernel enter the picture.

--linas

Re: IBM S/370 and binutils

[David and Jannette Uczen]

Just a quick correction in my last reply. When I was talking about the
mainframe it uses up to 32GB of memory, not 32MB.

--David

-----Original Message-----
From: Ian Lance Taylor <ian@zembu.com>
To: linas@linas.org <linas@linas.org>
Cc: uczen@mint.net <uczen@mint.net>; binutils@sourceware.cygnus.com
<binutils@sourceware.cygnus.com>; linas@linas.org <linas@linas.org>
Date: Friday, October 29, 1999 12:36 PM
Subject: Re: IBM S/370 and binutils


>   Date: Fri, 29 Oct 1999 11:50:28 -0500 (CDT)
>   From: linas@linas.org
>
>   > -----Original Message-----
>   > From: Ian Lance Taylor <ian@zembu.com>
>   > To: uczen@mint.net <uczen@mint.net>
>   > Cc: binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>
>   > Date: Wednesday, October 27, 1999 7:40 PM
>   > Subject: Re: IBM S/370 and binutils
>   >
>   >
>   > >   From: "David and Jannette Uczen" <uczen@mint.net>
>   > >   Date: Wed, 27 Oct 1999 19:22:49 -0400
>   > >
>   > >   I am forwarding this message to inform the binutils developers
that I
>   > >   believe that any further use of the ELF machine type EM_I370
(0xF00F)
>   > should
>   > >   be discontinued.
>
>   I picked this value in the absence of any other suitable value.
>   There is no formal ELF ABI that would specify such a value,
>   although I have been working on & off with folks inside of IBM
>   (the TJ Watson research center) to create such an ABI.  Clearly,
>   among the first steps would be to designate a machine type.
>
>   Please let me know who I should contact to get this (and also,
>   a type for the binary format) reserved in the most suitably
>   official manner possible.
>
>Talk to registry@sco.com.  They maintain the lists of official ELF
>machine numbers.  They'll give you a number if you ask, although
>sometimes you have to ask a few times.  It may help to pull in
>somebody from IBM; they seem to take big companies more seriously than
>free software people.
>
>First, though, is there some reason that it is inappropriate to use
>EM_S370?  I don't know anything about the 370, so that may be a stupid
>question.
>
>Ian
>
>

Re: IBM S/370 and binutils

[David and Jannette Uczen]

You should be able to get an ABI from Amdahl Corporation, through UTS
Technical Support. I have contacted a Chris A. Swartout
(cas30@oes.amdahl.com), who verified that the value of EM_S370 (9) is used
for IBM System/370 Processors. This value still exists in the binutils
source (/binutils/include/elf/common.h).
I have had the value officially registered with SCO (registry@sco.com). A
new ABI is going to be published soon.
I myself am trying to get ABI information, but haven't heard back yet.

--David

-----Original Message-----
From: linas@linas.org <linas@linas.org>
To: David and Jannette Uczen <uczen@mint.net>
Cc: ian@zembu.com <ian@zembu.com>; binutils@sourceware.cygnus.com
<binutils@sourceware.cygnus.com>; linas@linas.org <linas@linas.org>
Date: Friday, October 29, 1999 12:22 PM
Subject: Re: IBM S/370 and binutils


>> -----Original Message-----
>> From: Ian Lance Taylor <ian@zembu.com>
>> To: uczen@mint.net <uczen@mint.net>
>> Cc: binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>
>> Date: Wednesday, October 27, 1999 7:40 PM
>> Subject: Re: IBM S/370 and binutils
>>
>>
>> >   From: "David and Jannette Uczen" <uczen@mint.net>
>> >   Date: Wed, 27 Oct 1999 19:22:49 -0400
>> >
>> >   I am forwarding this message to inform the binutils developers that I
>> >   believe that any further use of the ELF machine type EM_I370 (0xF00F)
>> should
>> >   be discontinued.
>
>Ian,
>
>I picked this value in the absence of any other suitable value.
>There is no formal ELF ABI that would specify such a value,
>although I have been working on & off with folks inside of IBM
>(the TJ Watson research center) to create such an ABI.  Clearly,
>among the first steps would be to designate a machine type.
>
>Please let me know who I should contact to get this (and also,
>a type for the binary format) reserved in the most suitably
>official manner possible.
>
>--linas
>
>

Re: IBM S/370 and binutils

[David and Jannette Uczen]

I have already registered EM_S370 (9) with SCO. It should appear in the new
ABI.

There should be no problem using EM_S370. In all the header files I have
seen, EM_S370 has either had the correct value (9) or has been reserved.

The IBM System/370 Processor family has a long history, dating back to the
early 1960's. Each new processor is fully upwards compatible with all
previous chips. Although they were primarily water-cooled through the late
1980's, the latest generations, the System/390 processors, are air-cooled
CMOS microprocessors. There are even boards (P/390) that fit and run in PC's
at about 9 MIPS. The current "mainframe" versions (actually they are about
the size of a coke machine now) come with 1 to 12 processors, run at up to
1000+ MIPS, and have up to 32MB of memory.

The S/370 is a 16 register, 32-bit processor, however, IBM is expected to
release a 64-bit version in mid to late 2000. The current version now
support IEEE floating point. The S/370 is a big-endian machine.

Amdahl (I forget his first name), one of the developers of the S/370
processor, left IBM and started Amdahl Corporation and began manufacturing
S/370-compatible processors, much like AMD makes x86-compatible processors.

In the late 1980's, Amdahl developed UTS, a SVR4-compliant operating system
that runs on S/370 and compatible processors. They assigned EM_S370 (9) as
an ELF machine type, although SCO has no record of this. As I said earlier,
this value is now registered.

With this in mind, I think it would be inappropriate NOT to use EM_S370.
Besides only having one machine value for a particular processor (which I
believe is a goal of ELF), it would also facilitate porting GNU projects to
the UTS operating system.

Sorry for the long-winded blurb on S/370, but I thought you might want a
little background.

--David



-----Original Message-----
From: Ian Lance Taylor <ian@zembu.com>
To: linas@linas.org <linas@linas.org>
Cc: uczen@mint.net <uczen@mint.net>; binutils@sourceware.cygnus.com
<binutils@sourceware.cygnus.com>; linas@linas.org <linas@linas.org>
Date: Friday, October 29, 1999 12:36 PM
Subject: Re: IBM S/370 and binutils


>   Date: Fri, 29 Oct 1999 11:50:28 -0500 (CDT)
>   From: linas@linas.org
>
>   > -----Original Message-----
>   > From: Ian Lance Taylor <ian@zembu.com>
>   > To: uczen@mint.net <uczen@mint.net>
>   > Cc: binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>
>   > Date: Wednesday, October 27, 1999 7:40 PM
>   > Subject: Re: IBM S/370 and binutils
>   >
>   >
>   > >   From: "David and Jannette Uczen" <uczen@mint.net>
>   > >   Date: Wed, 27 Oct 1999 19:22:49 -0400
>   > >
>   > >   I am forwarding this message to inform the binutils developers
that I
>   > >   believe that any further use of the ELF machine type EM_I370
(0xF00F)
>   > should
>   > >   be discontinued.
>
>   I picked this value in the absence of any other suitable value.
>   There is no formal ELF ABI that would specify such a value,
>   although I have been working on & off with folks inside of IBM
>   (the TJ Watson research center) to create such an ABI.  Clearly,
>   among the first steps would be to designate a machine type.
>
>   Please let me know who I should contact to get this (and also,
>   a type for the binary format) reserved in the most suitably
>   official manner possible.
>
>Talk to registry@sco.com.  They maintain the lists of official ELF
>machine numbers.  They'll give you a number if you ask, although
>sometimes you have to ask a few times.  It may help to pull in
>somebody from IBM; they seem to take big companies more seriously than
>free software people.
>
>First, though, is there some reason that it is inappropriate to use
>EM_S370?  I don't know anything about the 370, so that may be a stupid
>question.
>
>Ian
>
>

Re: IBM S/370 and binutils

[linas]

It's been rumoured that Ian Lance Taylor said:
>    Ian has your i370 patch. I believe he has been swamped and he has
>    a long backlog for binutils patches. In the meantime, I will keep
>    maintaining my binutils for Linux.
> 
> You're right, I do have the patch.  At first I thought I didn't, but I
> wasn't checking my older archive (I'm so far behind that I have two
> archives now).  Sorry for the confusion.  I'll try to bump it up in
> priority.

Ugh.  I hate being swamped.  I know what it feels like. Thanks.

--linas

Re: IBM S/370 and binutils

[H.J. Lu]

> 
> Talk to registry@sco.com.  They maintain the lists of official ELF
> machine numbers.  They'll give you a number if you ask, although
> sometimes you have to ask a few times.  It may help to pull in
> somebody from IBM; they seem to take big companies more seriously than
> free software people.

I am looking at the draft of the next SVR4 gABI. 9 is reserved. In the
previous email from Joel at SCO, it will be used for EM_S370.

> 
> First, though, is there some reason that it is inappropriate to use
> EM_S370?  I don't know anything about the 370, so that may be a stupid
> question.

Linas, does Linux/i370 use the same ABI as defined by Amdahl's EM_S370?
If we do, we can use it. Otherwise, we have to apply for a new EM_I370.


-- 
H.J. Lu (hjl@gnu.org)

Re: IBM S/370 and binutils

[H.J. Lu]

> 
> Talk to registry@sco.com.  They maintain the lists of official ELF
> machine numbers.  They'll give you a number if you ask, although
> sometimes you have to ask a few times.  It may help to pull in
> somebody from IBM; they seem to take big companies more seriously than
> free software people.

I have been working with the person who maintains SVR4 ABI. I will
see what I can do.

> 
> First, though, is there some reason that it is inappropriate to use
> EM_S370?  I don't know anything about the 370, so that may be a stupid
> question.
> 



-- 
H.J. Lu (hjl@gnu.org)

Re: IBM S/370 and binutils

[Ian Lance Taylor]

   Date: Fri, 29 Oct 1999 09:28:52 -0700 (PDT)
   From: hjl@lucon.org (H.J. Lu)

   > I was not aware that Ian was the maintainer for the FSF copy of
   > binutils, and/or that patches sent to hjlu didn't 'automagically' 
   > end up in Ian's source tree.  I signed all the copyright assignments 
   > some 6-9 months ago, and hope to pursue with Ian the formalities of
   > getting the changes merged in.

   Ian has your i370 patch. I believe he has been swamped and he has
   a long backlog for binutils patches. In the meantime, I will keep
   maintaining my binutils for Linux.

You're right, I do have the patch.  At first I thought I didn't, but I
wasn't checking my older archive (I'm so far behind that I have two
archives now).  Sorry for the confusion.  I'll try to bump it up in
priority.

   Ian, Linas has some new changes to i370. Linas or I can send you an
   updated patch after you check in the current i370 patch you got.

OK.

Ian

Re: IBM S/370 and binutils

[Ian Lance Taylor]

   Date: Fri, 29 Oct 1999 11:50:28 -0500 (CDT)
   From: linas@linas.org

   > -----Original Message-----
   > From: Ian Lance Taylor <ian@zembu.com>
   > To: uczen@mint.net <uczen@mint.net>
   > Cc: binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>
   > Date: Wednesday, October 27, 1999 7:40 PM
   > Subject: Re: IBM S/370 and binutils
   > 
   > 
   > >   From: "David and Jannette Uczen" <uczen@mint.net>
   > >   Date: Wed, 27 Oct 1999 19:22:49 -0400
   > >
   > >   I am forwarding this message to inform the binutils developers that I
   > >   believe that any further use of the ELF machine type EM_I370 (0xF00F)
   > should
   > >   be discontinued.

   I picked this value in the absence of any other suitable value.
   There is no formal ELF ABI that would specify such a value,
   although I have been working on & off with folks inside of IBM 
   (the TJ Watson research center) to create such an ABI.  Clearly,
   among the first steps would be to designate a machine type.  

   Please let me know who I should contact to get this (and also, 
   a type for the binary format) reserved in the most suitably 
   official manner possible.

Talk to registry@sco.com.  They maintain the lists of official ELF
machine numbers.  They'll give you a number if you ask, although
sometimes you have to ask a few times.  It may help to pull in
somebody from IBM; they seem to take big companies more seriously than
free software people.

First, though, is there some reason that it is inappropriate to use
EM_S370?  I don't know anything about the 370, so that may be a stupid
question.

Ian

Re: IBM S/370 and binutils

[H.J. Lu]

> 
> It's been rumoured that Alan Modra said:
> > 
> > On Wed, 27 Oct 1999, David and Jannette Uczen wrote:
> > 
> > > I built binutils-2.9.5.0.14 with target=i370-elf, assembled an empty .s file
> > > to get an ELF binary. When I used readelf, it said the machine type was
> > > 0xf00f (unkown).
> 
> I was not aware that Ian was the maintainer for the FSF copy of
> binutils, and/or that patches sent to hjlu didn't 'automagically' 
> end up in Ian's source tree.  I signed all the copyright assignments 
> some 6-9 months ago, and hope to pursue with Ian the formalities of
> getting the changes merged in.
> 

Ian has your i370 patch. I believe he has been swamped and he has
a long backlog for binutils patches. In the meantime, I will keep
maintaining my binutils for Linux.

Ian, Linas has some new changes to i370. Linas or I can send you an
updated patch after you check in the current i370 patch you got.

Thanks.

-- 
H.J. Lu (hjl@gnu.org)

Re: IBM S/370 and binutils

[linas]

> -----Original Message-----
> From: Ian Lance Taylor <ian@zembu.com>
> To: uczen@mint.net <uczen@mint.net>
> Cc: binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>
> Date: Wednesday, October 27, 1999 7:40 PM
> Subject: Re: IBM S/370 and binutils
> 
> 
> >   From: "David and Jannette Uczen" <uczen@mint.net>
> >   Date: Wed, 27 Oct 1999 19:22:49 -0400
> >
> >   I am forwarding this message to inform the binutils developers that I
> >   believe that any further use of the ELF machine type EM_I370 (0xF00F)
> should
> >   be discontinued.

Ian,

I picked this value in the absence of any other suitable value.
There is no formal ELF ABI that would specify such a value,
although I have been working on & off with folks inside of IBM 
(the TJ Watson research center) to create such an ABI.  Clearly,
among the first steps would be to designate a machine type.  

Please let me know who I should contact to get this (and also, 
a type for the binary format) reserved in the most suitably 
official manner possible.

--linas

Re: IBM S/370 and binutils

[linas]

...
> > I built binutils-2.9.5.0.14 with target=i370-elf, assembled an empty .s file

BTW, a better place to discuss i370 related things is on the
linux-vm mailing list.  This is where most of the developers hang out.

LINUX-VM@VM.MARIST.EDU  


--linas

Re: IBM S/370 and binutils

[linas]

It's been rumoured that Alan Modra said:
> 
> On Wed, 27 Oct 1999, David and Jannette Uczen wrote:
> 
> > I built binutils-2.9.5.0.14 with target=i370-elf, assembled an empty .s file
> > to get an ELF binary. When I used readelf, it said the machine type was
> > 0xf00f (unkown).

I was not aware that Ian was the maintainer for the FSF copy of
binutils, and/or that patches sent to hjlu didn't 'automagically' 
end up in Ian's source tree.  I signed all the copyright assignments 
some 6-9 months ago, and hope to pursue with Ian the formalities of
getting the changes merged in.

--linas

Re: IBM S/370 and binutils

[Alan Modra]

On Wed, 27 Oct 1999, David and Jannette Uczen wrote:

> I built binutils-2.9.5.0.14 with target=i370-elf, assembled an empty .s file
> to get an ELF binary. When I used readelf, it said the machine type was
> 0xf00f (unkown).
> 
> The source I got is at ftp://ftp.valinux.com/pub/support/hjl/binutils , to
> which I was directed by Linas Vepstas ( http://www.linas.org ) who is working
> on a port of Linux to S/370.
> 
> Is this not official? Thanks.

This is H.J. Lu's official release of binutils including the official
Linas Vepstas' IBM370 patches, which don't happen to yet be in the
official GNU binutils sources.  :-)

I can understand the confusion from someone exposed to free software
mainly through linux, as most (all?) linux distributions use H.J. Lu's
binutils.  Thus H.J.'s binutils are 'official' to most linux users.  H.J.
has supported linux with very rapid response to bugs and new feature
requests, while the Cygnus/FSF binutils releases have lagged somewhat.
(CVS development sources are available to everyone though).  Part of the
reason for the lag is that the Cygnus/FSF binutils releases need to 
support practically every architecture on the planet, while H.J. focuses
on architectures that run linux.  Also the FSF is rather strict about such
things as copyright assignment.

Ian, BTW, is the official grand-poobah binutils maintainer.

> -----Original Message-----
> From: Ian Lance Taylor <ian@zembu.com>
> >
> >As far as I know, the GNU binutils do not use that value.  The
> >official sources may be found at
> >    http://sourceware.cygnus.com/binutils
> >
> >There may be patches out there which use it.  If I get those patches,
> >I will bear your comments in mind.

Re: IBM S/370 and binutils

[David and Jannette Uczen]

I built binutils-2.9.5.0.14 with target=i370-elf, assembled an empty .s file
to get an ELF binary. When I used readelf, it said the machine type was
0xf00f (unkown).

The source I got is at ftp://ftp.valinux.com/pub/support/hjl/binutils , to
which I was directed by Linas Vepstas ( http://www.linas.org ) who is working
on a port of Linux to S/370.

Is this not official? Thanks.
-----Original Message-----
From: Ian Lance Taylor <ian@zembu.com>
To: uczen@mint.net <uczen@mint.net>
Cc: binutils@sourceware.cygnus.com <binutils@sourceware.cygnus.com>
Date: Wednesday, October 27, 1999 7:40 PM
Subject: Re: IBM S/370 and binutils


>   From: "David and Jannette Uczen" <uczen@mint.net>
>   Date: Wed, 27 Oct 1999 19:22:49 -0400
>
>   I am forwarding this message to inform the binutils developers that I
>   believe that any further use of the ELF machine type EM_I370 (0xF00F)
should
>   be discontinued.
>
>As far as I know, the GNU binutils do not use that value.  The
>official sources may be found at
>    http://sourceware.cygnus.com/binutils
>
>There may be patches out there which use it.  If I get those patches,
>I will bear your comments in mind.
>
>Ian
>
>

Re: IBM S/370 and binutils

[Ian Lance Taylor]

   From: "David and Jannette Uczen" <uczen@mint.net>
   Date: Wed, 27 Oct 1999 19:22:49 -0400

   I am forwarding this message to inform the binutils developers that I
   believe that any further use of the ELF machine type EM_I370 (0xF00F) should
   be discontinued.

As far as I know, the GNU binutils do not use that value.  The
official sources may be found at
    http://sourceware.cygnus.com/binutils

There may be patches out there which use it.  If I get those patches,
I will bear your comments in mind.

Ian

IBM S/370 and binutils

[David and Jannette Uczen]

I am forwarding this message to inform the binutils developers that I
believe that any further use of the ELF machine type EM_I370 (0xF00F) should
be discontinued.

Somehow, Amdahl Corporation did not register their machine type (IBM
System/370 Processor) with the SVR4 registry, which at the time was AT&T.
The have a commercially available SVR4-compliant operating system called
UTS. In accordance with the SVID, they use ELF as their object/binary format
with a machine type of EM_S370 (9). As you can see by the attached messages,
the value is now an officially registered value.

As far as Amdahl's implementation of relocation, program loading and dynamic
linking (all defined as processor-specific in the ABI), I am requesting
information from Amdahl.

Subsequently, work has been done with binutils and GCC to utilize EM_I370,
the exact same processor. Since I don't think it's wise to have to values
for the same processor, I suggest that any further development of binutils
or GCC use EM_S370 instead.

I think part of the confusion involves Amdahl. Amdahl makes S/370 compatible
processors, just as AMD make x86 compatible processors. IBM software runs on
Amdahl and visa-versa. Actually, Hitachi also makes compatible machines.

As far as autoconf target values, i.e. i370-*, I don't know if those should
be converted to s370-* or not.

ELF support is there for S/370 in hopes, I assume, of running Linux.
However, by using EM_S370, the GNU projects should easily be ported to UTS
as well.

I am considering designing an ESD/XSD backend to support the OS/390, VM and
VSE communities. Currently, only commerical compilers are available. With
the machine defintions already in place, I am hoping to extend Dave Pitt's
work. He has developed some patches that produce HLASM-compatible assembler
(IBM's Assembler), but it still needs work. If a ESD/XSD backend existed for
BFD, then GAS could also be used. This should allow targets like
s370-amdahl-uts4, s370-ibm-mvs, s370-ibm-os390, s370-ibm-oe, s370-ibm-vm,
s370-ibm-vse, s370-ibm-linux, s370-elf, s370-esd, s370-xsd, etc..


Please offer comments to my proposal.

Thanks,
David Uczen
Mainframe Technical Support
Hannaford Bros. Co.

P.S. A new version of the SVR4 ABI will be published soon, which should
contain EM_S370 as an official value.


-----Original Message-----
From: David Uczen <duczen@hannaford.com>
To: uczen@mint.net <uczen@mint.net>
Date: Wednesday, October 27, 1999 6:40 PM
Subject: Re: Request for new ELF machine type


>
>
>
>---------------------- Forwarded by David Uczen/Tech Support/Corp/HBC on
>10/27/99 06:38 PM ---------------------------
>
>
>"Silverstein J." <jds@sco.COM> on 10/27/99 08:30:05 AM
>
>To:   David Uczen/Tech Support/Corp/HBC@HBC
>cc:   registry@sco.COM
>Subject:  Re: Request for new ELF machine type
>
>
>
>On Tue, Oct 26, 1999 at 08:01:36PM -0400, David Uczen wrote:
>>
>>
>> Joel,
>>
>> Here is the reply I got back from Amdahl UTS Technical Support. I hope
this
>> confirms the value of 9 for EM_S370, or I can have Chris Swartout contact
you.
>>
>> Thanks,
>> David Uczen
>> Mainframe Technical Support
>> Hannaford Bros. Co.
>
>David,
>
>Okay, we have assigned EM_S370 with value 9 to mean
>IBM System/370 Processor.
>--
>Joel Silverstein
>SCO Development Systems
>jds@sco.com (908) 790-2371
>
>All opinions expressed are my own.
>
>>
>>
>> ---------------------- Forwarded by David Uczen/Tech Support/Corp/HBC on
>> 10/26/99 07:56 PM ---------------------------
>>
>>
>> cas30@oes.amdahl.com (Chris A. Swartout) on 10/26/99 03:25:12 PM
>>
>> To:   David Uczen/Tech Support/Corp/HBC@HBC
>> cc:   utsproducts@oes.amdahl.com
>> Subject:  Re: UTS Use of ELF
>>
>>
>>
>>   David and Joel,
>>
>>   The elf header structure and legal values of the e_machine element
>>   (as of the assignment of the Amdahl value) are included below:
>>
>> typedef struct {
>>         unsigned char   e_ident[EI_NIDENT];     /* ident bytes */
>>         Elf32_Half      e_type;                 /* file type */
>>         Elf32_Half      e_machine;              /* target machine */
>>         Elf32_Word      e_version;              /* file version */
>>         Elf32_Addr      e_entry;                /* start address */
>>         Elf32_Off       e_phoff;                /* phdr file offset */
>>         Elf32_Off       e_shoff;                /* shdr file offset */
>>         Elf32_Word      e_flags;                /* file flags */
>>         Elf32_Half      e_ehsize;               /* sizeof ehdr */
>>         Elf32_Half      e_phentsize;            /* sizeof phdr */
>>         Elf32_Half      e_phnum;                /* number phdrs */
>>         Elf32_Half      e_shentsize;            /* sizeof shdr */
>>         Elf32_Half      e_shnum;                /* number shdrs */
>>         Elf32_Half      e_shstrndx;             /* shdr string index */
>> } Elf32_Ehdr;
>>
>> #define EM_NONE         0               /* e_machine */
>> #define EM_M32          1               /* AT&T WE 32100 */
>> #define EM_SPARC        2               /* Sun SPARC */
>> #define EM_386          3               /* Intel 80386 */
>> #define EM_68K          4               /* Motorola 68000 */
>> #define EM_88K          5               /* Motorola 88000 */
>> #define EM_486          6               /* Intel 80486 */
>> #define EM_860          7               /* Intel i860 */
>> #define EM_MIPS         8               /* Mips R2000 */
>> #define EM_S370         9               /* Amdahl       */
>> #define EM_NUM          10
>>
>>   This is definately still used by Amdahl, and our SVR4 product (UTS4)
>>   uses elf compilers and binaries.
>>
>>   I hope this answers your questions.
>>
>>   Chris.
>>
>> David Uczen recently spake thusly:
>> >Subject: UTS Use of ELF
>> >Date: Tue, 26 Oct 1999 12:22:11 -0400
>> >From: "David Uczen" <duczen@hannaford.com>
>> >To: utsproducts@oes.amdahl.com
>> >Message-ID: <85256816.005A3AF7.00@mta.hannaford.com>
>> >Mime-Version: 1.0
>> >
>> >
>> >
>> >Hello,
>> >
>> >I may be mistaken, but I have heard that your UTS operating system is
>> >SVR4-compliant. If this is true, you should be using ELF binaries. If
you are
>> >using ELF binaries, I would like to know what the value of your
e_machine is
>in
>> >the ELF header.
>> >
>> >I found somewhere a value of EM_S370 (9) that was supposedly used by
Amdahl.
>> >However, SCO, the legal owner of the SVR4 code and offical registry of
>> e_machine
>> >values has no record of this.
>> >
>> >I am attempting to (re)establish EM_S370 as an official value. Could you
tell
>> me
>> >if you do or ever have used EM_S370? And if you have used it, is it 9
and is
>it
>> >still valid?
>> >
>> >Thanks for your help,
>> >David Uczen
>> >Hannaford Bros. Co.
>> >(207) 885-3421
>> >
>> >FYI: See attached message(s):
>> >
>> >---------------------- Forwarded by David Uczen/Tech Support/Corp/HBC on
>> >10/26/99 12:18 PM ---------------------------
>> >
>> >
>> >"Silverstein J." <jds@sco.COM> on 10/26/99 10:18:56 AM
>> >
>> >To:   David Uczen/Tech Support/Corp/HBC@HBC
>> >cc:
>> >Subject:  Re: Request for new ELF machine type
>> >
>> >
>> >
>> >Sorry it has taken so long to reply.  We have been digging
>> >through old mail, trying to see if there is any record
>> >of a value for Amdahl.  We haven't been able to find
>> >positive confirmation of that.  We will be glad to assign a
>> >new number for EM_S370.  If you would like to use the value 9,
>> >however, we will need for you to get someone from Amdahl in touch
>> >with us to verify that they are using 9 to mean S370.
>> >--
>> >Joel Silverstein
>> >SCO Development Systems
>> >jds@sco.com (908) 790-2371
>> >
>> >All opinions expressed are my own.
>> >
>> >On Mon, Oct 18, 1999 at 11:24:24AM -0400, David Uczen wrote:
>> >>
>> >>
>> >> With that in mind, I would like to request that EM_S370 be assigned
since
>it
>> >may
>> >> (or may not) be used by Amdahl. (NOTE: I have found some documentation
that
>> >> EM_S370 used to have a value of 9, but I don't know if Amdahl still
uses
>it.)
>> >>
>> >> Thank you very much,
>> >>
>> >> David Uczen
>> >> Mainframe Technical Support
>> >> Hannaford Bros. Co.
>> >>
>> >>
>> >>
>> >>
>> >>
>> >> "Silverstein J." <jds@sco.COM> on 10/18/99 09:07:24 AM
>> >>
>> >> To:   David Uczen/Tech Support/Corp/HBC@HBC
>> >> cc:   registry@sco.COM
>> >> Subject:  Re: Request for new ELF machine type
>> >>
>> >>
>> >>
>> >> On Fri, Oct 15, 1999 at 02:41:32PM -0400, David Uczen wrote:
>> >> >
>> >> >
>> >> > I would like to request an new machine type for the ELF standard:
>> >> >
>> >> > The processor is the IBM System/370 Processor (Bi-polar and CMOS).
>> >> >
>> >> > Proposed constant: EM_S370
>> >> >
>> >> > Proposed description: IBM System/370 Architecture
>> >> >
>> >> > Requesting organization:  My self, for two reasons: 1) I am
interested in
>> >> > developing tools that use ELF on the S/370 platform and would like
to use
>> an
>> >> > official value; and 2) The GNU binutils project has selected an
arbitrary
>> >> value
>> >> > (EM_I370 and 0xF00F), while a commercial SVR4-compliant system,
Amdahl
>> >> > Corporation's UTS, already exists.
>> >> >
>> >> >
>> >> > If it is necessary to distinquish between different processors
although
>> they
>> >> are
>> >> > upwardly compatible, then I would also like to propose:
>> >> >
>> >> > The processor is the IBM System/390 CMOS Microprocessor.
>> >> >
>> >> > Proposed constant: EM_S390
>> >> >
>> >> > Proposed description: IBM System/390 Architecture
>> >> >
>> >> > Requesting organization:  My self, in addition to the above reasons,
the
>> >S/390
>> >> > is NOT always downwardly compatible, i.e. IEEE floating point
>instructions
>> >> will
>> >> > not work for S/370.
>> >> >
>> >> >
>> >> > Thank you,
>> >> >
>> >> > David Uczen
>> >> > Mainframe Technical Support
>> >> > Hannaford Bros. Co.
>> >> >
>> >>
>> >> We can assign 2 values.  I would recommend, however, that you
>> >> use a single value and then use a bit in the e_flags field
>> >> of the ELF header to indicate that S390 floating-point support is
>> >> required for programs that use those instructions.  I think
>> >> it is probably simpler in the tools.  If the processors are
>> >> otherwise compatible, it keeps things more consistent.  Let
>> >> us know how you want to go.
>> >> --
>> >> Joel Silverstein
>> >> SCO Development Systems
>> >> jds@sco.com (908) 790-2371
>> >>
>> >> All opinions expressed are my own.
>> >>
>> >>
>> >>
>> >
>> >
>> >
>> >
>> >
>>
>>
>> --
>>   Chris A. Swartout    | Amdahl Corporation            |   N6WCP
>>   cas30@oes.amdahl.com | 1250 East Arques Ave  M/S 278 |
>>  (408) 746-6590        | Sunnyvale,  CA, 94088-3470    |
>>
>> Disclaimer: All statements are my own and do NOT represent Amdahl in any
way.
>> -------------------------------------------------------------------------
----
>>
>>
>
>
>
>
>