[Qemu-devel] Open Hack'Ware 0.4 release candidate

[Qemu-devel] Open Hack'Ware 0.4 release candidate

[J. Mayer]

I just released OHW-0.4-rc
If no more regression is found, this will become the current version
used in Qemu-PPC.
Please try it and report unknown problems or successes, if any.

You'll find the binary image here:
<http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4-rc_bin.tar.bz2>
the sources with subversion informations here:
<http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4-rc_svn.tar.bz2>
and the updated status here:
<http://perso.magic.fr/l_indien/OpenHackWare/0.4/README>

-- 
J. Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] Open Hack'Ware 0.4 release candidate

[Natalia Portillo]

What you say about the BeOS:

The BeOS kernel isn't in the ISO part of the CD-ROM, but in the Big  
Endian BeFS part of it.
The BeBox firmware understand this system and loads the PEFF kernel as  
I commented you a time ago.
The BeOS Loader for MacOS seems to load the kernel from the BeFS also  
and then let control to it removing MacOS from memory.

For getting BeOS compatibility there are two ways:
First of all, emulate a BeBox. This shouldn't be a very hard way, as  
the BeBox is simply a PReP machine with a lot of chips (SCSI, UARTs,  
etc) and with an specific BIOS (I think, not similar to OpenFirmware at  
any point).
Second of all, see how the BeOS Loader loads the kernel in a MacOS. The  
loader don't like at all the software ROM, and probably ignores totally  
OpenFirmware and relies to ToolBox to get loaded and then do hardware  
probes (it does directly SCSI commands to load the system on a real  
Macintosh.)

About Windows NT did you get the kernel to load?
I have also 3.5 and 3.51 so I can help you also in this.

About Pegasos and MorphOS did you pass ago the UART check we had  
problems before?

You know I offer my help for anything, and I have know more spare time  
than before.

Regards,
Natalia Portillo

El 20/03/2005, a las 13:39, J. Mayer escribió:

> I just released OHW-0.4-rc
> If no more regression is found, this will become the current version
> used in Qemu-PPC.
> Please try it and report unknown problems or successes, if any.
>
> You'll find the binary image here:
> <http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4- 
> rc_bin.tar.bz2>
> the sources with subversion informations here:
> <http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4- 
> rc_svn.tar.bz2>
> and the updated status here:
> <http://perso.magic.fr/l_indien/OpenHackWare/0.4/README>
>
> -- 
> J. Mayer <l_indien@magic.fr>
> Never organized
>
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>

Re: [Qemu-devel] Open Hack'Ware 0.4 release candidate

[J. Mayer]

On Mon, 2005-03-21 at 02:56 +0000, Natalia Portillo wrote:
> What you say about the BeOS:
> 
> The BeOS kernel isn't in the ISO part of the CD-ROM, but in the Big  
> Endian BeFS part of it.

I did not say the kernel was in the El-Torito image.
I just said the CDROMs I got contain ISO filesystems.

> The BeBox firmware understand this system and loads the PEFF kernel as  
> I commented you a time ago.
> The BeOS Loader for MacOS seems to load the kernel from the BeFS also  
> and then let control to it removing MacOS from memory.

I though I remembered booting directly BeOS CDROM on a Mac, but I may be
wrong on this point...

> For getting BeOS compatibility there are two ways:
> First of all, emulate a BeBox. This shouldn't be a very hard way, as  
> the BeBox is simply a PReP machine with a lot of chips (SCSI, UARTs,  
> etc) and with an specific BIOS (I think, not similar to OpenFirmware at  
> any point).
> Second of all, see how the BeOS Loader loads the kernel in a MacOS. The  
> loader don't like at all the software ROM, and probably ignores totally  
> OpenFirmware and relies to ToolBox to get loaded and then do hardware  
> probes (it does directly SCSI commands to load the system on a real  
> Macintosh.)

For now, I'm not in a hurry to see BeOS booting...
I just made a try and saw that it failed. Moreover, I think it would
only boot on OldWorld Macs, that are not supported in qemu.

> About Windows NT did you get the kernel to load?
> I have also 3.5 and 3.51 so I can help you also in this.
> 
> About Pegasos and MorphOS did you pass ago the UART check we had  
> problems before?

I did not tried hard to make MorphOs and NT boot in qemu.
My first goals are to have all Linux kernel booting, then all BSD
(including MacOS X).
Feel free to make tries and add a new hardware target for Pegasos or
make NT boot.

> You know I offer my help for anything, and I have know more spare time  
> than before.

If you'd like to make NT and/or MorphOs boot in qemu, I can give you
support and add what's needed in OHW. It would be great if it works, but
it's not my priorities, there's still a lot of work to do before
qemu-PPC / OHW can reasonably be considered as "production-ready"
software for the end user.


> El 20/03/2005, a las 13:39, J. Mayer escribió:
> 
> > I just released OHW-0.4-rc
> > If no more regression is found, this will become the current version
> > used in Qemu-PPC.
> > Please try it and report unknown problems or successes, if any.
> >
> > You'll find the binary image here:
> > <http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4- 
> > rc_bin.tar.bz2>
> > the sources with subversion informations here:
> > <http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4- 
> > rc_svn.tar.bz2>
> > and the updated status here:
> > <http://perso.magic.fr/l_indien/OpenHackWare/0.4/README>

-- 
J. Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] Open Hack'Ware 0.4 release candidate

[Natalia Portillo]

Hi!

>> What you say about the BeOS:
>>
>> The BeOS kernel isn't in the ISO part of the CD-ROM, but in the Big
>> Endian BeFS part of it.
>
> I did not say the kernel was in the El-Torito image.
> I just said the CDROMs I got contain ISO filesystems.
It should have three tracks, one in HFS/ISO with El-Torito, one in 
little endian BeFS (BeOS/x86) and one in big endian BeFS (BeOS/PPC),
if not, the CD is incorrect.

>> The BeBox firmware understand this system and loads the PEFF kernel as
>> I commented you a time ago.
>> The BeOS Loader for MacOS seems to load the kernel from the BeFS also
>> and then let control to it removing MacOS from memory.
>
> I though I remembered booting directly BeOS CDROM on a Mac, but I may 
> be
> wrong on this point...
I don't think you got that, as the zbeos in the ISO/HFS partition seems 
to be a x86 executable (I should see this in more detail), and OF 
doesn't recognize BeFS.

>> For getting BeOS compatibility there are two ways:
>> First of all, emulate a BeBox. This shouldn't be a very hard way, as
>> the BeBox is simply a PReP machine with a lot of chips (SCSI, UARTs,
>> etc) and with an specific BIOS (I think, not similar to OpenFirmware 
>> at
>> any point).
>> Second of all, see how the BeOS Loader loads the kernel in a MacOS. 
>> The
>> loader don't like at all the software ROM, and probably ignores 
>> totally
>> OpenFirmware and relies to ToolBox to get loaded and then do hardware
>> probes (it does directly SCSI commands to load the system on a real
>> Macintosh.)
>
> For now, I'm not in a hurry to see BeOS booting...
> I just made a try and saw that it failed. Moreover, I think it would
> only boot on OldWorld Macs, that are not supported in qemu.
Yes, it will only boot on OldWorld, but also on BeBox (that are PReP 
machines).
There is one person interested on emulating a BeBox.
Do you think will be easy to add BeFS support to OHW using OpenBeOS or 
Linux's BeFS code?


>> About Windows NT did you get the kernel to load?
>> I have also 3.5 and 3.51 so I can help you also in this.
>>
>> About Pegasos and MorphOS did you pass ago the UART check we had
>> problems before?
>
> I did not tried hard to make MorphOs and NT boot in qemu.
> My first goals are to have all Linux kernel booting, then all BSD
> (including MacOS X).
> Feel free to make tries and add a new hardware target for Pegasos or
> make NT boot.
>
>> You know I offer my help for anything, and I have know more spare time
>> than before.
>
> If you'd like to make NT and/or MorphOs boot in qemu, I can give you
> support and add what's needed in OHW. It would be great if it works, 
> but
> it's not my priorities, there's still a lot of work to do before
> qemu-PPC / OHW can reasonably be considered as "production-ready"
> software for the end user.
Please do it, and if you can send me a binary qemu from the lastest CVS 
will be great.
I think NT should boot with only the support for the executable format, 
at least to a point that shows anything interesting on screen (surely a 
big blue screen of death xD).
MorphOS SHOULD work if we mix the qemu's PC emulation with OHW and 
PowerPC cpu, as Pegasos is a standard PC motherboard with an OF clone 
and a PowerPC G3/G4 chip.

If anyone in the list is interested on investing for Pegasos emulation 
I have a Pegasos board that only needs Registered PC-133 SDRAM and an 
ATX power supply to boot, and when booted, we can extract the fully 
hardware map and OF tree.

Regards

>
>> El 20/03/2005, a las 13:39, J. Mayer escribió:
>>
>>> I just released OHW-0.4-rc
>>> If no more regression is found, this will become the current version
>>> used in Qemu-PPC.
>>> Please try it and report unknown problems or successes, if any.
>>>
>>> You'll find the binary image here:
>>> <http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4-
>>> rc_bin.tar.bz2>
>>> the sources with subversion informations here:
>>> <http://perso.magic.fr/l_indien/OpenHackWare/0.4/OpenHackWare-0.4-
>>> rc_svn.tar.bz2>
>>> and the updated status here:
>>> <http://perso.magic.fr/l_indien/OpenHackWare/0.4/README>
>
> -- 
> J. Mayer <l_indien@magic.fr>
> Never organized
>

Re: [Qemu-devel] Open Hack'Ware 0.4 release candidate

[J. Mayer]

On Mon, 2005-03-21 at 23:43 +0000, Natalia Portillo wrote:
> Hi!

Hi,

> >> What you say about the BeOS:
> >>
> >> The BeOS kernel isn't in the ISO part of the CD-ROM, but in the Big
> >> Endian BeFS part of it.
> >
> > I did not say the kernel was in the El-Torito image.
> > I just said the CDROMs I got contain ISO filesystems.
> It should have three tracks, one in HFS/ISO with El-Torito, one in 
> little endian BeFS (BeOS/x86) and one in big endian BeFS (BeOS/PPC),
> if not, the CD is incorrect.

Seems that it depends on which CDROMs. I have some BeOS ones that really
don't look this way.

> >> The BeBox firmware understand this system and loads the PEFF kernel as
> >> I commented you a time ago.
> >> The BeOS Loader for MacOS seems to load the kernel from the BeFS also
> >> and then let control to it removing MacOS from memory.
> >
> > I though I remembered booting directly BeOS CDROM on a Mac, but I may 
> > be
> > wrong on this point...
> I don't think you got that, as the zbeos in the ISO/HFS partition seems 
> to be a x86 executable (I should see this in more detail), and OF 
> doesn't recognize BeFS.

Right. The El-torito image I checked is for x86.

> >> For getting BeOS compatibility there are two ways:
> >> First of all, emulate a BeBox. This shouldn't be a very hard way, as
> >> the BeBox is simply a PReP machine with a lot of chips (SCSI, UARTs,
> >> etc) and with an specific BIOS (I think, not similar to OpenFirmware 
> >> at
> >> any point).
> >> Second of all, see how the BeOS Loader loads the kernel in a MacOS. 
> >> The
> >> loader don't like at all the software ROM, and probably ignores 
> >> totally
> >> OpenFirmware and relies to ToolBox to get loaded and then do hardware
> >> probes (it does directly SCSI commands to load the system on a real
> >> Macintosh.)
> >
> > For now, I'm not in a hurry to see BeOS booting...
> > I just made a try and saw that it failed. Moreover, I think it would
> > only boot on OldWorld Macs, that are not supported in qemu.
> Yes, it will only boot on OldWorld, but also on BeBox (that are PReP 
> machines).
> There is one person interested on emulating a BeBox.
> Do you think will be easy to add BeFS support to OHW using OpenBeOS or 
> Linux's BeFS code?

It seems BeFS is not well documented. There was a BeFS project for Linux
but it seems that it has been stopped (no activity for a few years) and
was never integrated into standard kernels.
If BeBox is just a PREP machine, the first step to emulate it would be
to try to boot the BeOS kernel using Qemu PREP target with -kernel
argument.
Then, we may don't need any BeFS emulation to start it.

> >> About Windows NT did you get the kernel to load?
> >> I have also 3.5 and 3.51 so I can help you also in this.
> >>
> >> About Pegasos and MorphOS did you pass ago the UART check we had
> >> problems before?
[...]
> I think NT should boot with only the support for the executable format, 
> at least to a point that shows anything interesting on screen (surely a 
> big blue screen of death xD).

Well, I'd have to merge the code I got for NT XCOFF loading, then and
retry...

> MorphOS SHOULD work if we mix the qemu's PC emulation with OHW and 
> PowerPC cpu, as Pegasos is a standard PC motherboard with an OF clone 
> and a PowerPC G3/G4 chip.

Then, it's exactly what a PREP board is...
So, a few modifications from PREP target (mostly base addresses, I
guess) should be OK.

[...]

-- 
J. Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] Open Hack'Ware 0.4 release candidate

[Natalia Portillo]

Hi!


>>>> What you say about the BeOS:
>>>>
>>>> The BeOS kernel isn't in the ISO part of the CD-ROM, but in the Big
>>>> Endian BeFS part of it.
>>>
>>> I did not say the kernel was in the El-Torito image.
>>> I just said the CDROMs I got contain ISO filesystems.
>> It should have three tracks, one in HFS/ISO with El-Torito, one in
>> little endian BeFS (BeOS/x86) and one in big endian BeFS (BeOS/PPC),
>> if not, the CD is incorrect.
>
> Seems that it depends on which CDROMs. I have some BeOS ones that 
> really
> don't look this way.
Ok but surely they are x86 only

> Right. The El-torito image I checked is for x86.
>
>>>> For getting BeOS compatibility there are two ways:
>>>> First of all, emulate a BeBox. This shouldn't be a very hard way, as
>>>> the BeBox is simply a PReP machine with a lot of chips (SCSI, UARTs,
>>>> etc) and with an specific BIOS (I think, not similar to OpenFirmware
>>>> at
>>>> any point).
>>>> Second of all, see how the BeOS Loader loads the kernel in a MacOS.
>>>> The
>>>> loader don't like at all the software ROM, and probably ignores
>>>> totally
>>>> OpenFirmware and relies to ToolBox to get loaded and then do 
>>>> hardware
>>>> probes (it does directly SCSI commands to load the system on a real
>>>> Macintosh.)
>>>
>>> For now, I'm not in a hurry to see BeOS booting...
>>> I just made a try and saw that it failed. Moreover, I think it would
>>> only boot on OldWorld Macs, that are not supported in qemu.
>> Yes, it will only boot on OldWorld, but also on BeBox (that are PReP
>> machines).
>> There is one person interested on emulating a BeBox.
>> Do you think will be easy to add BeFS support to OHW using OpenBeOS or
>> Linux's BeFS code?
>
> It seems BeFS is not well documented. There was a BeFS project for 
> Linux
> but it seems that it has been stopped (no activity for a few years) and
> was never integrated into standard kernels.
> If BeBox is just a PREP machine, the first step to emulate it would be
> to try to boot the BeOS kernel using Qemu PREP target with -kernel
> argument.
> Then, we may don't need any BeFS emulation to start it.
Uh, as far as I know the BeFS filesystem for Linux is in the kernel ^o)
You can download a fully explanation of the BeFS from Dominic 
Giampaolo's webpage (the designer of BeFS).
Also the OpenBeOS project has a working filesystem driver (at least 
readonly)

Without that in OHW or a BeOS/PPC compatible machine I'm not able to 
extract the kernel (the drivers I said are for little endian BeFS and 
not for big endian one)

>> I think NT should boot with only the support for the executable 
>> format,
>> at least to a point that shows anything interesting on screen (surely 
>> a
>> big blue screen of death xD).
>
> Well, I'd have to merge the code I got for NT XCOFF loading, then and
> retry...
Say me

>> MorphOS SHOULD work if we mix the qemu's PC emulation with OHW and
>> PowerPC cpu, as Pegasos is a standard PC motherboard with an OF clone
>> and a PowerPC G3/G4 chip.
>
> Then, it's exactly what a PREP board is...
> So, a few modifications from PREP target (mostly base addresses, I
> guess) should be OK.
The qemu prep emulation currently emulates the serial controller?
Remember that we debugged the mos kernel and find that it was trying to 
access a non implemented command in the rs232 controller

Re: [Qemu-devel] Open Hack'Ware 0.4 release candidate

[J. Mayer]

On Tue, 2005-03-22 at 23:27 +0000, Natalia Portillo wrote:
[...]
> > It seems BeFS is not well documented. There was a BeFS project for 
> > Linux
> > but it seems that it has been stopped (no activity for a few years) and
> > was never integrated into standard kernels.
> > If BeBox is just a PREP machine, the first step to emulate it would be
> > to try to boot the BeOS kernel using Qemu PREP target with -kernel
> > argument.
> > Then, we may don't need any BeFS emulation to start it.
> Uh, as far as I know the BeFS filesystem for Linux is in the kernel ^o)

Damned, you're right !

> You can download a fully explanation of the BeFS from Dominic 
> Giampaolo's webpage (the designer of BeFS).

Yes, but this documentation says "this book does not contain exhaustive
bit-level details of BFS". But it may be sufficient to have a read-only
implementation...

> Also the OpenBeOS project has a working filesystem driver (at least 
> readonly)
> 
> Without that in OHW or a BeOS/PPC compatible machine I'm not able to 
> extract the kernel (the drivers I said are for little endian BeFS and 
> not for big endian one)

I guess there's some other way to extract the kernel, but I must make
some tests in order to be sure of this...

> >> I think NT should boot with only the support for the executable 
> >> format,
> >> at least to a point that shows anything interesting on screen (surely 
> >> a
> >> big blue screen of death xD).
> >
> > Well, I'd have to merge the code I got for NT XCOFF loading, then and
> > retry...
> Say me
> 
> >> MorphOS SHOULD work if we mix the qemu's PC emulation with OHW and
> >> PowerPC cpu, as Pegasos is a standard PC motherboard with an OF clone
> >> and a PowerPC G3/G4 chip.
> >
> > Then, it's exactly what a PREP board is...
> > So, a few modifications from PREP target (mostly base addresses, I
> > guess) should be OK.
> The qemu prep emulation currently emulates the serial controller?
> Remember that we debugged the mos kernel and find that it was trying to 
> access a non implemented command in the rs232 controller

Qemu PREP emulates all the basic devices of a PREP machine, including
serial drivers. But Pegasos IO seem to be completely different, as far
as I remember. Then, a way of guessing where the device are would be:
1/ copy hw/ppc_prep.c into hw/ppc_pegasos.c, modify hw/ppc.c and vl.c to
recognize "pegasos" board type
2/ Display all IO and memory accesses out of known ranges.
3/ Launch the kernel and try to guess what it tries to do when accessing
devices.
I did guess it tries to probe a serial port and enter an infinite loop
when it does not find it.

-- 
J. Mayer <l_indien@magic.fr>
Never organized