(Answer) (Category) Linux on PowerPC FAQ-O-Matic : (Category) Hardware Compatibility : (Category) G3/G4 Upgraded PowerMacs :
G3 Upgrades on NON AV 6100
I found the issues for G3 upgraded NON AV 6100s slightly different than those previously posted here for AV equipped 6100s.
Info below is really a summary of stuff garnered from postings to comp.os.linux.powerpc.
The following works for me on a NON AV 6100/60 with MacOS 8.6 installed on the HFS boot partition with a Newer Techmology card (http://www.newertech.com) - their part no 7MXPA6100.
It DID NOT work with the equivalent card from Sonnet Technology (http://www.sonnettech.com) - their part no C61-G3-512-250.
Basic procedure:
1. change file type of MkLinux Booter from 'scri' to 'INIT' using ResEdit so that it boots after the G3 enabling extension.
2. Ensure, by name of Extension or use of Conflict Catcher 8, that the G3 enabling extension (MAXpwr for Newer, Crescendo for Sonnet) runs before the MkLinux Booter.
3. Install Generic #7 kernel parts (from http://www.mklinux.org/development/kernels.html ), that is, put the Mach Kernel in the MacOS file system Extensions folder and the vmlinux in /mach_servers in the MkLinux file system. The MkLinux Booter from the DR3 CDs was used
4. Added 'mach_options= bus_speed_hz=30000000' line to lilo.conf
With the Newer G3 upgrade, the BogoMips reported is about 8x over that for the motherboard PPC601. Running KDE and most everything else is a lot brisker.
I do not know enough to determine why this procedure did not work for the Sonnet card. The problem appears to be with the Mach Kernel, not the Booter or the vmlinux. Symptomatically, it does the same as it did with the Newer card before I switched from the stock DR3 kernel to Generic #7; the Booter splash window displays the mach_options, boot partition and 'System booting...' and then the machine stares at its navel. My SWAG is that whatever was done in Generic #7 to stop the micro kernel believing it was dealing with an AV equipped machine and get it to use DRAM video may not work for the Sonnet card.
r.barrett@ftel.co.uk
I am using a PPC 6100/60 (not AV) w/ a Sonnet Crescendo (methinks the same one you tried). Got the navel staring, definetly in the kernel, didn't even replace the vmlinux file but had the same results. Called Sonnet, the tech sup guy assured me their senior software engineer was a 'big MKLinux proponent' but MacOS 9 probs are taking precedent and they would hope to have something soon... I am not much of a coder, but am willing to try, any ideas on starting points/references I should read? What might need tweeking? (I am not a Mac guy, grew up on PCs). Gonna give generic 8 mach kernel and maybe the others a random try, will update if the navel staring stops, anyone else had luck with Sonnet cards?

barsamian@entranceramp.com
I have a non-AV 6100, which boots using MkLinux booter and a kernel from YellowDog ftp that was called "NOT_MACH_KERNEL"...something.
It works great! Now I'm trying to build a cross-debugging kernel to see if I can adjust the startup code to avoid the problem that I think is happening with my Sonnet G3 upgrade card (which works like a champ in OS 8.5, 8.6, and 9.x):
When linux boots with the G3 card enabled by the Crescendo extension (newest download - September 2003) using MkLinux booter, it freezes. Bootx, which never worked properly, at least gets to fill the display with colored hash before it locks up.
Examining the code, I think what is happening is the kernel sees the 750 processor (G3), and automatically assumes that it's a new-world ROM, and then tried to manipulate the G3 proessor using the PCI/ROM interface present on newer Macs, but absent on this PowerPC 6100. Somehow, the NewerTek card doesn't let the kernel realize there's another processor present.
Of course, the NewerTek accelerators are reported to work, so this may *NOT* be what's happening.
Maybe it provides Soft-ROM code that mimics the NewWorld ROMS? Or hides the processor from the kernel-probe code?
To find out, I need a little advice - how do I make my compiled kernel boot with MkLinux booter (the only working booter I can use)? Do I have to uncompress it, or is there a specific way to make a "NOT_MACH_KERNEL" that is really a vmlinux file compiled on YellowDog?
As soon as I can boot my new kernel, I can try to serial-debug the boot-up and see if my theory is correct.
Otherwise, I'll have to bug Sonnet into showing me the necessary steps to enable the Crescendo G3 (400/1M) card so I can write a module that will load and enable the Crescendo card *AFTER* booting into linux...
-G!
neuf@power.net
Ok - step one complete: I can cross compile a bootable/stable kernel on my x86 box for the 6100, and I've debugged the serial connection setup to the point that I'm sure it's functional. Unfortunately, I can't seem to get the kgdb (kernel debugger) to compile in the kernel. Working on that. In the meantime, I've cross-compiled gdb 5.3 to run on the x86 box, with the 601/750 PowerPC instruction set as the target, and I'm ready try to figure out why it won't boot the kernel after I've used MacOs and the Crescendo INIT/scri to enable the Crescendo G3/400 accelerator card. If anyone out there knows how to get the kgdb stuff in ppc kernel 2.4.x to compile, please let me know - even if it's just to point out that it's not meant for the ppc platform...
neuf@power.net
[Append to This Answer]
Previous: (Answer) Sonnet has updated their mkLinux compatability listing
Next: (Answer) Will the Performa 6360 or PowerMac 6400 work when upgraded with the NewerTech G3 Card?
This document is: http://www.jonh.net/cgi-bin/lppcfom/fom?file=832
[Search] [Appearance]
This is a Faq-O-Matic 2.718d.
Hosted by anduin.org and SourceForge Logo