Adventures In G5 Land

[nanopico]

I decided to play around with the G5 this morning.
The dual 2.0ghz G5 I have is the target of todays adventure.
I think this has been tried before and there was an error about cascading interrupts and it just crashed hard.
Well I looked closer.  It is the same issue as the xserve.
This time being the data device in the K2 IO chip.  The issue comes down to not all the interrupt and register information being in the device tree.  This is the same issue right now with booting the xserve with the built in ATA controller.

So I deleted the devices.
And it went a little further.
And then it started complaining about being unable to register the VSP device. Whatever that is.
So I deleted that too.
Then it complained about not being able to initialize the ethernet port.
So guess what I deleted that too.
Then it complained about the firewire port.
So I deleted that as well.
Then it got all the way to the end of the trampoline code and right as it passes off to the ROM it crashes.

I have an error code I haven't looked into yet, but none of the debug options yet give me any further details.

So there is a good chance a G5 may get to at least the Happy Mac boot screen.

Of course if we get to a desktop a lot of hardware will not be useful.

[MacTron]

Of course if we get to a desktop a lot of hardware will not be useful.

... but will be worth to be seen. 

[ELN]

Then it got all the way to the end of the trampoline code and right as it passes off to the ROM it crashes.

This is fantastic! If (as I have assumed but not actually checked) the Trampoline passes control straight to ROM offset 0x310000, then my Nanokernel hacking will be of some use. The disassembly project is on the back burner for the time being, but after my exams finish on Tuesday I will be able to get going again. It shouldn't be very long before I can make small patches to the kernel and build them into a tbxi. Mind pasting the debug output somewhere?

[Mat]

Guys you are incredible!

The entire "now bootable usupported machines-storys" are absolutely great for me. I am sad that I cannot contribute anything as I cannot code.
A huge point for me is the confirmation - now 12 years later - that it all was about marketing from Apple. If I remember how much I discussed, nearly quarreled with friends who belived that it is impossible, and that Apple had to do the cut, ...

And now there is even some possibility that G5s will boot 9 one day!

Marvellous!

[MacOS Plus]

  Mat, you share my thoughts exactly.  All over the internet there were insults and taunts hurled at anyone who had the 'gall' to suggest that trying this even just for fun was possible or worthwhile.  The actual fact of the matter is so elegantly simple - booting my Xserve required only a few characters changed in the Mac OS ROM file and two commands in open firmware to disable one device.  That's a mere three lines of code for a complete boot and a nearly fully functional machine!  Not only that, the exact same disabled device was already working in OS 9 on a PCI card from Sonnet without any additional drivers.  Apple intentionally changed something in the firmware in this chip and apparently nothing else.  Our work on this may eventually correct that issue also.

  In the end, I'm revelling in the opportunity to rub it in the faces of all those who were so rude about it for years.  I now have an Xserve booting OS 9 from an SSD and probably get to be the first in the world to say that!  Obviously nanopico and iMic get most if the credit, and nanopico did boot the CD first, but it's still a big deal.  It can only get better for our community as the work continues.

[Mat]

MacOS Plus, I can absolutely understand your enthusiasm! I even feelt a little bit like you with the new GhostScript (when there was only 1 download from the newly compiled version).

What I do not understand is why nobody tried it earlier?
When I think back to 2003/4 there was a real tumult about the new "not 9 bootable machines". Apple even had to release the latest MDD for 9 again. Imagine if that what nanopico and iMic did/do now, would have happened 6 month after Apple presented the FW800 MDDs!
Don´t get me wrong, the work both are doing is incredible, but also in the past there were good coders around, and as we know now it is possible with 3 lines of code (what an irony!). Why in the hell did nobody back then investigate in this direction!?

What a pity!

[MacOS Plus]

  I'm hard-pressed to understand it either and can only try to guess at it.  There is a good chance that Apple would have sent their notorious legal team after anyone publishing solutions at the time.  It's very unfortunate it didn't happen at the time because information and in-service machines were readily available.  Very few people are in a position these days to take on such a project and be able to source out all the needed documentation.

  I guess the last ten years can be called OS 9's "lost decade".  I never imagined the 'golden years' would come so long after the fact.  It was almost too late to save many of these machines from the scrap heap.  We're so lucky that our little group of dedicated enthusiasts and 'brains' have come together at the right moment to finally see this dream through.  It's high time we struck back at planned obsolescence and literally rewrite history!  (Sent from my five-year-old phone. )

[Mat]

It's high time we struck back at planned obsolescence and literally rewrite history!

There is nothing to add!
You cannot imagine how hapy I am here, reading from people around the world thinking similar to how I do the last decade (remember I never stopped using Mac OS 9 only, never switched to any other OS).

[MacOS Plus]

  If anything, it's become so easy for everyone to connect socially (Did I just say that about computer geeks?!? ) via the internet from all corners of the globe that it allows a critical mass of like-minded people to assemble around a common interest.  I've found it now entirely routine that an internet forum has users from many, many countries.  It's really quite fantastic because this completely changed and accelerated my ability to gain knowledge since prior to the internet.  (Ugh, dating myself here!)

[DieHard]

  If anything, it's become so easy for everyone to connect socially (Did I just say that about computer geeks?!? ) via the internet from all corners of the globe that it allows a critical mass of like-minded people to assemble around a common interest.  I've found it now entirely routine that an internet forum has users from many, many countries.  It's really quite fantastic because this completely changed and accelerated my ability to gain knowledge since prior to the internet.  (Ugh, dating myself here!)

That is the reason for success.  back in the day, we could not share ideas like this.

I remember 2003/4 and I had friends in the apple camp. Also, my company was an authorized Apple repair center. I remember asking about the MDD 1.42 and I was told that the logic board was so radically differenent that it could never boot to OS 9. 

They even extended to lie and I remember a district service manager explaining that Apple Hardwired changes that would prevent the OS 9 ROM file from loading or working at all, I was told that hardware addresses were physically changed and there was no hope for a patch.

Obviously, this was all mis information, we were so focussed on the hardware setups/repairs that we believed everything we were told.  It is clear that at least in the G4 line, it would have been very easy to continue with OS 9 compatibility and marketing (not hardware changes) were to blame

[nanopico]

They even extended to lie and I remember a district service manager explaining that Apple Hardwired changes that would prevent the OS 9 ROM file from loading or working at all, I was told that hardware addresses were physically changed and there was no hope for a patch.

Obviously, this was all mis information, we were so focussed on the hardware setups/repairs that we believed everything we were told.  It is clear that at least in the G4 line, it would have been very easy to continue with OS 9 compatibility and marketing (not hardware changes) were to blame

That's actually pretty funny.  It doesn't matter if the addresses changed. It's all in the open firmware parameters. OS 9 doesn't give to shakes of a rats tail what address are used as long as they are listed in the device tree.  All apple did was change the compatible property.  And the instance of the xserve leave out some of the addresses in firmware leaving the new os x drivers to figure it out.  Which I find amusing as that defeated the whole point of open firmware.

[MacTron]

Most of the info was really out there. But at its time The Os X propaganda was really effective. And the very few users that resisted with Mac Os 9, were calmed with a reedition of the the Mac Os 9 enabled MDD.
The very few that digged for some information, were aware about the existence of the Custer ROM, and a how it reveals that the Mac Os 9 support was a question of the ROM, drivers, and "something more" ...
The  OF flash needed to enable the 7448, 7447 CPUs, shows that something in to this direction should be made.

We don't know exactly what to do until iMic, nanopico and others begin its great work, and  find  the details of the process.

It's not the same to spend today 50-100 bucks in a unsupported iMac, eMac , MacMini,  etc.. and make some tests than spend 1000-2000 bucks at its time to make some test...

[nanopico]

As has been stated before, MacTron was it you?, the U3 is going to be a problem.
Well I'm not sure what the hang up is at this point, but the current error on the G5 is a memory related error (not anything specific) followed by no interrupt controller found.
Up to that point it is able to initialize all hardware and setup the interrupts.
Because it's a memory error and interrupt it is probably U3 not being know well enough.

The ROM has parcels for various driver for early boot.  Some of them related to the Uni-North chip in the other PPC machines.  It might be a case of creating a driver and packaging it as a parcel in the ROM.  Well at least to get further.

I think it's possible to get it working, but it's going to take time.

[nanopico]

It may have been a while on this one, but here I go.

My brain is always coming up with crazy idea's (isn't that how all these threats of updates started?)

Based on what I have been able to do so far, I have a new method of attack on this.
This is low priority, but it popped in my head so I'm going to share it.

The xserve has problems traversing parts of the device tree because of how it interprets interrupts.
This is also one of the problems with the G5.  If I remove a lot of the devices from the tree the trampoline is as happy as a bug in a rug.
It flakes out on what I am guessing is actually the cpu version check. That whole 7447, 7448, 7447a stuff on the G4's

Task is then to locate that check and the lookup table used to load what ever needs to be loaded and put the G5 cpu id in and map the same G4 data to it.  Probably won't work but it's a start.

The other actual major issue is the disk controllers.  You have the SATA controller and the ATA controller.  OS 9 Doesn't know enough about either one of those (thus the reason third party cards emulate scsi).
Well for testing this is easy.  Stick a OS 9 supported SATA or IDE PCI card in and boot off it.

 I know I'm crazy, but one can dream right?

[mrhappy]

 I know I'm crazy, but one can dream right?

Keep dreaming nanopico!!!

[IIO]

  Mat, you share my thoughts exactly.  All over the internet there were insults and taunts hurled at anyone who had the 'gall' to suggest that trying this even just for fun was possible or worthwhile

there are two kinds of people. and the ones who just do it.