It seems to be failing in the flash-based initrd code, never getting to the builder code. Now trying to remember how to look at the internal initrd image. It has been far too long!

EDIT: /proc/flash_b0000 contains the boot-time initrd image, which mainly consists of a pre-compiled linuxrc binary. That's the part that is failing for some reason.

I think this is the code that contains "pump", and otherwise simply exists to mount the "real" root filesystem from /dev/hda5. It probably also looks after displaying the "dead empeg" logo when "No Hard Disk Found" happens.

Dunno how "pump" works (that "bootloader protocol" thing I asked about earlier), but the rest of it should be easy to recreate if need be.

Meanwhile I'll poke around some more and see if I can reverse engineer why it is failing here.

EDIT:
The initrd code seems to be having trouble either mounting or reading from /dev/hda4 (the real root filesystem). Weird. It's the hijack kernel that does all of the low-level accesses, so there's not much that could be wrong here.

Just for giggles, I copied an existing hda4 image onto the drive and booted from it without issue. Then took the drive back to my Linux PC and copied the bigdisk builder contents into hda4, and swapped the drive back into the empeg afterward. It boots up, and successfully runs the disk builder.

So.. just got to figure out exactly which step causes the trouble. Next up: manually copy the hda4 image from bigdisk onto the drive (as opposed to the already working file-by-file copy of the same thing).