[Coco] More new ideas for booting Nitros9

Allen Huffman alsplace at pobox.com
Wed Feb 4 09:55:08 EST 2015


> On Feb 4, 2015, at 7:06 AM, Brett Gordon <beretta42 at gmail.com> wrote:
> 
> Phase Four:  Passing a "commandline" to the kernel from the booter.
> Dynamically tell the kernel what use as /DD, etc....

My suggestion would be to look in to what was done for OSK-9/68000 and beyond. Those kernels use the init module for the memory list, default console device, default disk device (if present), etc. The kernel could consult that module for configuration.

Combining things in to one "kernel" seems natural. Back when I was an OS-9 developer, I never really understood os9p1 and os9p2 and what they were. I learned most of my architecture once I worked for Microware and was teaching classes on OS-9. Looking at OS-9/6809 now, it's quite different than the OS-9 I learned there. And I expect there are some good reasons for it.

Also -- what does OS-9/6809 look like for a GIMIX? Did it also have the split kernel? Was the kernel customized for each system to do init stuff? On later OS-9's, the kernel was independent of anything but processor -- all init stuff was done by the boot code, and the kernel just had a memory map to work in, and relied on init to proceed.

The OS-9 68K 2.4 docs are on www.colorcomputerarchive.com, I noticed, and the Technical Reference Manual would be a good place to look for ideas.

Between 2.4 (which the MM/1, TC70, etc. ran) and the 3.0 release, a few things changed. At some point, ioman was merged in to the kernel, then later, it was split out again so (just a guess) Microware could point out how small it's kernel was (even though it seems pretty rare for systems to run without any I/O -- why use such a powerful kernel just for process scheduling but on open/close/read/write?).

		-- A



More information about the Coco mailing list