31 Jul
2010
31 Jul
'10
6:49 p.m.
As you might remember from a previous post, I am trying to virtualize
/ emulate an old Data General UNIX system to a VM env'.
This is my progress so far (after a lot of reading, searching and
understanding who is who and what does it all mean) :
1. I've DD'ed an exact copy of the raw system disk as an image on my Linux box.
2. Loaded it up with most virtualization products out there - VMware,
VirtualBox, Xen/KVM/Qemu, Bochs and such.
They all work, ie, bootloader works and getting a DG/UX menu to
choose my proper kernel/system disk.
3. This is where they fail, as they cannot find the system's original
Adaptec PCI SCSI Controller (which is an AIC-7880 chip).
Most VM products only support LSI/BusLogic SCSI, and IDE of course.
As I see it, I have a few things I can do to make it work:
1. write an Adaptec driver for a VM product.
2. write an LSI/BusLogic driver for DG/UX or implement virtio drivers into it.
3. get IDE support working.
1st 2 options, as I understand from software eng' and other
professionals, are very difficult.
The 3rd option is interesting :
On one hand, I have an config file with all of DG/UX's supported
hardware, IDE is simply not there (would love for someone else to have
a look if possible. Just say so and I'll send you the file).
_and_ docs say "Data General has not tested or qualified the use of
IDE devices in AViiON AV 2100 systems".
_BUT_, on the other hand, BIOS seems to have plenty of IDE config
options and also docs say:
"IDE Interface - Disabled/Auto" , "OnBoard IDE Controller -
enabled".
" The AV 2100 board set consists of a system board with six PCI
expansion slots, three ISA expansion slots, and several embedded
controller devices (PCI video, SCSI, Network, and IDE).
and "..A SCSI-III AIC-7880 controller for connecting up to seven 8-bit
narrow SCSI devices. You can also connect a mixture of fifteen 8-bit
narrow and 16-bit wide SCSI devices to the controller—maximum of seven
8-bit narrow SCSI devices. PCI-enhanced Integrated Drive Electronics
(IDE) hard disk interface that supports two hard disk drives"
Also " Since the lower bay contains a SCSI CD-ROM drive, only one
other SCSI device can be installed. You can install an IDE device in
the remaining bay. "
And much more. The question is, if I have an exact image of the
primary disk and I did try to load it via an IDE drive (in Qemu/
VirtualBox etc), didn't work out. Like last time, bootloader booted up
and then hang after trying to load the kernel.
Is there something I can try and do as a different config in my image file?
Anything else you can think of?
Please let me know if you need any file sent over.
Any help would be greatly appreciated.
Thanks,
Adam
2 Aug
2 Aug
6:11 a.m.
New subject: [TUHS & QEMU] Making progress with old DG/UX virtualization. Need advice.
Hi,
I've read all your posts in the QEMU mailing list and the TUHS one and I'm
answering to both lists in a hope my mail enlights you and any other curious.
First of all, old UNIX systems (and I put my hand on the fire for DG/UX also), use a
monolithic linked at setup/later time kernel.
That is, even if you get a driver (IDE, virtio, whatsoever), the configuration files, the
kernel, the ramdisk, everything that lets your system boot, MUST HAVE BEEN BOOT from the
AIC controller, the driver is hardcoded, no way to change it.
If you have extensive knowledge of what files a driver setup modifies on DG/UX
specifically (knowledge from other UNIX, forget it, they are as different as Porsche and
Ferrari motors), you can always get a new kernel with the drivers you need to make it boot
and manually put them in your image.
In the case, you meet this requirements, and, you do it, you can then achieve to other
problems. The DG/UX workstations are x86 machines, but nothing swears they are PC
compatible machines, and they can have a different memory map for some critical device, or
include critical devices never found in a PC (like an Intel Macintosh does for example).
Just booting from a BIOS doesn't make the machines be the same (PowerPC Macintosh,
IBM POWER workstations, Genesi Pegasos, are machines that boot OpenFirmware with heavily
different configurations, devices and memory maps).
Also, you are assuming IDE is available in DG/UX just because the controller is present in
the hardware. That hardware was also used for Windows NT. IDE support can be JUST FOR
Windows, and the DG/UX manufacturer just decided to not include an IDE driver in the
kernel (happened in AIX for PCs until last version of all, only SCSI was supported, being
a hugely strange controller in PC worlds).
In the case you opt for making a driver (adding IDE, virtio, or other SCSI support) for
the DG/UX need to say you need, low level knowledge of the hardware, low level knowledge
of the operating system, a working machine (for sure, with the hardware available), a
debug machine (almost sure also), C and maybe assembler knowledge. In a scale of 10, this
puts the difficulty in 8 for most of programmers, and surely if you were one you stacked
with the first option everyone gave you (see next sentence).
The easiest way, and the one that people answered you already in QEMU's mailing list
(in a scale of 10 the difficulty is 6 or even 5), is creating an emulated device
(that's the correct term, not "driver") for an emulator, like QEMU, Bochs,
VirtualBox (forget this option for VMWare, VirtualPC or Parallels) that adds the AIC SCSI
controller you exactly need.
Why is this easiest? You don't need any DG/UX working system, you don't need to
know how DG/UX works, you don't need to compile a kernel, copy it to your image.
You just take the Adaptec's documentation, and start coding, making a SCSI emulated
controller, and testing it with systems you can always reinstall, debug, and check, until
they fully work (Windows, Linux, BSD, take your choice).
And then, you just polish it until your DG/UX boots, or finds the memory map as a mess it
doesnt like.
Finally, please stop begging on all the internet, spend that time coding the driver or
getting the money to pay a programmer that will do.
Sincerely yours,
Natalia Portillo
Claunia.com CEO
QEMU's Official OS Support List maintainer
7:48 a.m.
New subject: [TUHS & QEMU] Making progress with old DG/UX virtualization. Need advice.
Thanks Natalia,
I'll start by answering the insultive part of your answer, as my ego
will not let me go on if I don't:
I am not "begging on all the internet", I am simply seeking solutions,
help and advice, and making sure to update whoever is interested in
the progress I am doing.
Also, I wish to thank you for your insight and well detailed answer.
Finally I got an explanation as to _why_ solution A will not be as
good as solution B. That is what I call a winning argument, and I
thank you for that.
I already have people searching for Adaptec docs and programmers for
the creation of the driver, err, emulated device.
Take care,
Adam
On Mon, Aug 2, 2010 at 9:11 AM, Natalia Portillo <claunia(a)claunia.com> wrote:
Hi,
I've read all your posts in the QEMU mailing list and the TUHS one and I'm
answering to both lists in a hope my mail enlights you and any other curious.
First of all, old UNIX systems (and I put my hand on the fire for DG/UX also), use a
monolithic linked at setup/later time kernel.
That is, even if you get a driver (IDE, virtio, whatsoever), the configuration files, the
kernel, the ramdisk, everything that lets your system boot, MUST HAVE BEEN BOOT from the
AIC controller, the driver is hardcoded, no way to change it.
If you have extensive knowledge of what files a driver setup modifies on DG/UX
specifically (knowledge from other UNIX, forget it, they are as different as Porsche and
Ferrari motors), you can always get a new kernel with the drivers you need to make it boot
and manually put them in your image.
In the case, you meet this requirements, and, you do it, you can then achieve to other
problems. The DG/UX workstations are x86 machines, but nothing swears they are PC
compatible machines, and they can have a different memory map for some critical device, or
include critical devices never found in a PC (like an Intel Macintosh does for example).
Just booting from a BIOS doesn't make the machines be the same (PowerPC Macintosh,
IBM POWER workstations, Genesi Pegasos, are machines that boot OpenFirmware with heavily
different configurations, devices and memory maps).
Also, you are assuming IDE is available in DG/UX just because the controller is present
in the hardware. That hardware was also used for Windows NT. IDE support can be JUST FOR
Windows, and the DG/UX manufacturer just decided to not include an IDE driver in the
kernel (happened in AIX for PCs until last version of all, only SCSI was supported, being
a hugely strange controller in PC worlds).
In the case you opt for making a driver (adding IDE, virtio, or other SCSI support) for
the DG/UX need to say you need, low level knowledge of the hardware, low level knowledge
of the operating system, a working machine (for sure, with the hardware available), a
debug machine (almost sure also), C and maybe assembler knowledge. In a scale of 10, this
puts the difficulty in 8 for most of programmers, and surely if you were one you stacked
with the first option everyone gave you (see next sentence).
The easiest way, and the one that people answered you already in QEMU's mailing list
(in a scale of 10 the difficulty is 6 or even 5), is creating an emulated device
(that's the correct term, not "driver") for an emulator, like QEMU, Bochs,
VirtualBox (forget this option for VMWare, VirtualPC or Parallels) that adds the AIC SCSI
controller you exactly need.
Why is this easiest? You don't need any DG/UX working system, you don't need to
know how DG/UX works, you don't need to compile a kernel, copy it to your image.
You just take the Adaptec's documentation, and start coding, making a SCSI emulated
controller, and testing it with systems you can always reinstall, debug, and check, until
they fully work (Windows, Linux, BSD, take your choice).
And then, you just polish it until your DG/UX boots, or finds the memory map as a mess it
doesnt like.
Finally, please stop begging on all the internet, spend that time coding the driver or
getting the money to pay a programmer that will do.
Sincerely yours,
Natalia Portillo
Claunia.com CEO
QEMU's Official OS Support List maintainer
4 p.m.
New subject: [TUHS & QEMU] Making progress with old DG/UX virtualization. Need advice.
Hi,
El 02/08/2010, a las 08:48, DG UX escribió:
Thanks Natalia,
I'll start by answering the insultive part of your answer, as my ego
will not let me go on if I don't:
I am not "begging on all the internet", I am simply seeking solutions,
help and advice, and making sure to update whoever is interested in
the progress I am doing.
Also, I wish to thank you for your insight and well detailed answer.
Finally I got an explanation as to _why_ solution A will not be as
good as solution B. That is what I call a winning argument, and I
thank you for that.
That's why I sent you the message not because egos
I already have people searching for Adaptec docs and
programmers for
the creation of the driver, err, emulated device.
Great, I wish you my best and offer my repository of operating systems to test the
emulated device on as much systems as possible when it is mature enough.
Take care,
Adam
Natalia Portillo
Claunia.com
On Mon, Aug 2, 2010 at 9:11 AM, Natalia Portillo
<claunia(a)claunia.com> wrote:
> Hi,
>
> I've read all your posts in the QEMU mailing list and the TUHS one and I'm
answering to both lists in a hope my mail enlights you and any other curious.
>
> First of all, old UNIX systems (and I put my hand on the fire for DG/UX also), use a
monolithic linked at setup/later time kernel.
> That is, even if you get a driver (IDE, virtio, whatsoever), the configuration files,
the kernel, the ramdisk, everything that lets your system boot, MUST HAVE BEEN BOOT from
the AIC controller, the driver is hardcoded, no way to change it.
>
> If you have extensive knowledge of what files a driver setup modifies on DG/UX
specifically (knowledge from other UNIX, forget it, they are as different as Porsche and
Ferrari motors), you can always get a new kernel with the drivers you need to make it boot
and manually put them in your image.
>
> In the case, you meet this requirements, and, you do it, you can then achieve to
other problems. The DG/UX workstations are x86 machines, but nothing swears they are PC
compatible machines, and they can have a different memory map for some critical device, or
include critical devices never found in a PC (like an Intel Macintosh does for example).
Just booting from a BIOS doesn't make the machines be the same (PowerPC Macintosh,
IBM POWER workstations, Genesi Pegasos, are machines that boot OpenFirmware with heavily
different configurations, devices and memory maps).
>
> Also, you are assuming IDE is available in DG/UX just because the controller is
present in the hardware. That hardware was also used for Windows NT. IDE support can be
JUST FOR Windows, and the DG/UX manufacturer just decided to not include an IDE driver in
the kernel (happened in AIX for PCs until last version of all, only SCSI was supported,
being a hugely strange controller in PC worlds).
>
> In the case you opt for making a driver (adding IDE, virtio, or other SCSI support)
for the DG/UX need to say you need, low level knowledge of the hardware, low level
knowledge of the operating system, a working machine (for sure, with the hardware
available), a debug machine (almost sure also), C and maybe assembler knowledge. In a
scale of 10, this puts the difficulty in 8 for most of programmers, and surely if you were
one you stacked with the first option everyone gave you (see next sentence).
>
> The easiest way, and the one that people answered you already in QEMU's mailing
list (in a scale of 10 the difficulty is 6 or even 5), is creating an emulated device
(that's the correct term, not "driver") for an emulator, like QEMU, Bochs,
VirtualBox (forget this option for VMWare, VirtualPC or Parallels) that adds the AIC SCSI
controller you exactly need.
>
> Why is this easiest? You don't need any DG/UX working system, you don't
need to know how DG/UX works, you don't need to compile a kernel, copy it to your
image.
>
> You just take the Adaptec's documentation, and start coding, making a SCSI
emulated controller, and testing it with systems you can always reinstall, debug, and
check, until they fully work (Windows, Linux, BSD, take your choice).
>
> And then, you just polish it until your DG/UX boots, or finds the memory map as a
mess it doesnt like.
>
> Finally, please stop begging on all the internet, spend that time coding the driver
or getting the money to pay a programmer that will do.
>
> Sincerely yours,
> Natalia Portillo
> Claunia.com CEO
> QEMU's Official OS Support List maintainer
5014
days inactive
5016
days old
3 comments
2 participants
participants (2)
-
DG UX -
Natalia Portillo