[Qemu-devel] AIX emulated on x86 host

[Qemu-devel] AIX emulated on x86 host

[glen.c.bodie]

Sorry - first message had non-plain text by mistake. Trying again ...

I have an old AIX machine (IBM RS/6000 running AIX v5.x) and it's about to 
fall apart. I would like to migrate that machine's functions onto one of 
my VMware hosts, all of which are DELL 2950 servers (x86 architecture). I 
know VMware only runs x86 architecture guests. So I am planning a Windows 
2003 Server guest, running QEMU as really the only thing it is doing and 
inside QEMU I want to run that AIX machine's functions.

Fundamental question - think that will work?

If the answer is YES then I need to worry about creating the PPC disk 
image, loading AIX onto it, loading my AIX applications and migrating the 
data from the old hardware to the new emulated machine. Think I'm nuts? 
Got a better alternative?

Regards - Glen

Re: [Qemu-devel] AIX emulated on x86 host

[Stefan Hajnoczi]

On Wed, Oct 27, 2010 at 9:23 PM,  <glen.c.bodie@esso.ca> wrote:
> Sorry - first message had non-plain text by mistake. Trying again ...
>
> I have an old AIX machine (IBM RS/6000 running AIX v5.x) and it's about to
> fall apart. I would like to migrate that machine's functions onto one of
> my VMware hosts, all of which are DELL 2950 servers (x86 architecture). I
> know VMware only runs x86 architecture guests. So I am planning a Windows
> 2003 Server guest, running QEMU as really the only thing it is doing and
> inside QEMU I want to run that AIX machine's functions.
>
> Fundamental question - think that will work?
>
> If the answer is YES then I need to worry about creating the PPC disk
> image, loading AIX onto it, loading my AIX applications and migrating the
> data from the old hardware to the new emulated machine. Think I'm nuts?
> Got a better alternative?

I'm not aware of people doing this or whether the ppc targets
supported by QEMU can even run AIX hardware/firmware-wise.  Perhaps
someone has a definitive answer?

Stefan

Re: [Qemu-devel] AIX emulated on x86 host

[Natalia Portillo]

The last thing I've heard about that was long time ago when I and Jocelyn tried to make it work with OpenHackWare to no luck.

Anyway, right now, it does not work either, I just don't know if anyone is working on it.

I know that the AIX boot process is quite different (using MBR partition scheme, different memory maps).

Sorry but right now, "Does AIX/RS6000 work on QEMU?" is NO.

Regards,
Natalia Portillo

El 28/10/2010, a las 16:34, Stefan Hajnoczi escribió:

> On Wed, Oct 27, 2010 at 9:23 PM,  <glen.c.bodie@esso.ca> wrote:
>> Sorry - first message had non-plain text by mistake. Trying again ...
>> 
>> I have an old AIX machine (IBM RS/6000 running AIX v5.x) and it's about to
>> fall apart. I would like to migrate that machine's functions onto one of
>> my VMware hosts, all of which are DELL 2950 servers (x86 architecture). I
>> know VMware only runs x86 architecture guests. So I am planning a Windows
>> 2003 Server guest, running QEMU as really the only thing it is doing and
>> inside QEMU I want to run that AIX machine's functions.
>> 
>> Fundamental question - think that will work?
>> 
>> If the answer is YES then I need to worry about creating the PPC disk
>> image, loading AIX onto it, loading my AIX applications and migrating the
>> data from the old hardware to the new emulated machine. Think I'm nuts?
>> Got a better alternative?
> 
> I'm not aware of people doing this or whether the ppc targets
> supported by QEMU can even run AIX hardware/firmware-wise.  Perhaps
> someone has a definitive answer?
> 
> Stefan
> 

Re: [Qemu-devel] AIX emulated on x86 host

[Alexander Graf]

On 28.10.2010, at 08:42, Natalia Portillo wrote:

> The last thing I've heard about that was long time ago when I and Jocelyn tried to make it work with OpenHackWare to no luck.
> 
> Anyway, right now, it does not work either, I just don't know if anyone is working on it.


FWIW, Andreas Färber is working on it. Those efforts are really just at the beginning though, so don't expect to see anything soon.

Also, windows is a second-class platform for Qemu. It works, but not all features are supported and testing is usually done on Linux. So if you're planning to do anything with Qemu, please try to do it on a Linux host.


Alex

Re: [Qemu-devel] AIX emulated on x86 host

[Andreas Färber]

Am 28.10.2010 um 22:14 schrieb Alexander Graf:

> On 28.10.2010, at 08:42, Natalia Portillo wrote:
>
>> The last thing I've heard about that was long time ago when I and  
>> Jocelyn tried to make it work with OpenHackWare to no luck.
>>
>> Anyway, right now, it does not work either, I just don't know if  
>> anyone is working on it.
>
>
> FWIW, Andreas Färber is working on it. Those efforts are really just  
> at the beginning though, so don't expect to see anything soon.

There are three main issues currently:
1) QEMU doesn't emulate a machine that would run current versions of  
AIX in reality.
2) We don't have a true ppc64 OpenBIOS, just some ppc64 support in the  
32-bit version.
3) RTAS is just a stub in OpenBIOS, it doesn't even signal an error  
when used.

What I am working on:
* ppc64 compatibility of OpenBIOS code, i.e. getting things compiled  
at all
* 32-bit RTAS binary with minimal functionality
* a ppc64-softmmu chrp machine and accompanying differentiations in  
OpenBIOS

I occasionally push my WIP to:
http://repo.or.cz/w/qemu/afaerber.git - aix branch
http://repo.or.cz/w/openbios/afaerber.git - ppc branch

How you could help:
* Provide us with device trees from real AIX-capable CHRP machines  
that we can model memory layout and OF tree from. Additionally, any  
info on what devices/chipsets we would need to emulate would be  
helpful. (I'm currently basing things on mac99 emulation in lack  
thereof.)
* Subscribe to the OpenBIOS mailing list [1] and review and test our  
ppc64 patches.
* Any advice on how to get GCC 4.3+ powerpc64-linux cross-compilers  
built [2] would be appreciated. We could then better rule out compiler  
bugs.
* You might try making the ppc-softmmu PReP emulation work with  
OpenBIOS and check if that would be supported by AIX 5. Massimo  
Montecchi inquired about that but didn't post any patches since then;  
you could try contacting him. If PReP were supported by AIX 5, it  
would still require RTAS support in OpenBIOS though.

Regards,
Andreas

[1] http://www.openfirmware.info/mailman/listinfo/openbios
[2] http://lists.openbios.org/pipermail/openbios/2010-October/005343.html

[Qemu-devel] Re: AIX emulated on x86 host

[Andreas Schwab]

Andreas Färber <andreas.faerber@web.de> writes:

> * Any advice on how to get GCC 4.3+ powerpc64-linux cross-compilers built
> [2] would be appreciated. We could then better rule out compiler  bugs.

You don't need any of the target libraries, just the compiler, since you
are purely in freestanding land.  So all you need is "make all-gcc".

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [Qemu-devel] AIX emulated on x86 host

[glen.c.bodie]

[-- Attachment #1: Type: text/plain, Size: 2635 bytes --]

Andreas - it sounds like an interesting development project, but I'm going 
to have to decline. The amount of effort to get my application running on 
AIX emulated under QEMU seems to be greater than the effort we were hoping 
to avoid in the first case by porting the code to Windows. TNSTAAFL, 
thanks.

Regards - Glen

-----------------------------------------------------------------------------
Am 28.10.2010 um 22:14 schrieb Alexander Graf:

> On 28.10.2010, at 08:42, Natalia Portillo wrote:
>
>> The last thing I've heard about that was long time ago when I and 
>> Jocelyn tried to make it work with OpenHackWare to no luck.
>>
>> Anyway, right now, it does not work either, I just don't know if 
>> anyone is working on it.
>
> FWIW, Andreas Färber is working on it. Those efforts are really just 
> at the beginning though, so don't expect to see anything soon.

There are three main issues currently:
1) QEMU doesn't emulate a machine that would run current versions of 
   AIX in reality.
2) We don't have a true ppc64 OpenBIOS, just some ppc64 support in the 
   32-bit version.
3) RTAS is just a stub in OpenBIOS, it doesn't even signal an error 
   when used.

What I am working on:
* ppc64 compatibility of OpenBIOS code, i.e. getting things compiled 
  at all
* 32-bit RTAS binary with minimal functionality
* a ppc64-softmmu chrp machine and accompanying differentiations in 
  OpenBIOS

I occasionally push my WIP to:
http://repo.or.cz/w/qemu/afaerber.git - aix branch
http://repo.or.cz/w/openbios/afaerber.git - ppc branch

How you could help:
* Provide us with device trees from real AIX-capable CHRP machines 
  that we can model memory layout and OF tree from. Additionally, any 
  info on what devices/chipsets we would need to emulate would be 
  helpful. (I'm currently basing things on mac99 emulation in lack 
  thereof.)
* Subscribe to the OpenBIOS mailing list [1] and review and test our 
  ppc64 patches.
* Any advice on how to get GCC 4.3+ powerpc64-linux cross-compilers 
  built [2] would be appreciated. We could then better rule out compiler 
  bugs.
* You might try making the ppc-softmmu PReP emulation work with 
  OpenBIOS and check if that would be supported by AIX 5. Massimo 
  Montecchi inquired about that but didn't post any patches since then; 
  you could try contacting him. If PReP were supported by AIX 5, it 
  would still require RTAS support in OpenBIOS though.

Regards,
Andreas

[1] http://www.openfirmware.info/mailman/listinfo/openbios
[2] http://lists.openbios.org/pipermail/openbios/2010-October/005343.html


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3624 bytes --]

Re: [Qemu-devel] AIX emulated on x86 host

[Jernej Simončič]

On Tuesday, November 2, 2010, 14:53:08, glen.c.bodie@esso.ca wrote:

> Andreas - it sounds like an interesting development project, but I'm going
> to have to decline. The amount of effort to get my application running on
> AIX emulated under QEMU seems to be greater than the effort we were hoping
> to avoid in the first case by porting the code to Windows. TNSTAAFL, 
> thanks.

Since you're porting an application, you may find it easier to use
Windows' Services for Unix Applications, which gives you a pretty
complete POSIX environment on top of the NT kernel.

-- 
< Jernej Simončič ><><><><>< http://eternallybored.org/ >

Judgement comes from experience. Experience comes from poor judgement.
       -- Robert Lee's Truce

[Qemu-devel] Re: AIX emulated on x86 host

[Andreas Färber]

Am 30.10.2010 um 14:49 schrieb Andreas Schwab:

> Andreas Färber <andreas.faerber@web.de> writes:
>
>> * Any advice on how to get GCC 4.3+ powerpc64-linux cross-compilers  
>> built
>> [2] would be appreciated. We could then better rule out compiler   
>> bugs.
>
> You don't need any of the target libraries, just the compiler, since  
> you
> are purely in freestanding land.  So all you need is "make all-gcc".

make all-gcc
make all-target-libgcc
make install-gcc
make install-target-libgcc

did the trick, thanks a lot!

You wouldn't happen to know what to do about an "undefined reference  
to `_GLOBAL_OFFSET_TABLE_'"?

Output format is binary and supposed to be relocatable (tried -fpic  
and -fpie; for an RTAS blob relocated by OpenBIOS).
Compiles with above 4.5.1 powerpc64-linux-gnu-gcc but fails with  
powerpc-elf-gcc.
Is this a symbol I must provide in freestanding land? If yes, how? Or  
am I using the wrong compiler flags?

Andreas

[Qemu-devel] Re: AIX emulated on x86 host

[Andreas Schwab]

Andreas Färber <andreas.faerber@web.de> writes:

> You wouldn't happen to know what to do about an "undefined reference to
> _GLOBAL_OFFSET_TABLE_'"?

That symbol is generated by the linker, so it depends on how you call
it.

> Output format is binary and supposed to be relocatable (tried -fpic and
> -fpie; for an RTAS blob relocated by OpenBIOS).

Does it understand ELF PIC relocations?

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."