On Sun, 6 Aug 2000, gabriel rosenkoetter wrote:

> On Sun, Aug 06, 2000 at 03:25:53PM -0400, John Klos wrote:
> > Open Firmware will not boot the kernel from a UFS disk.
> 
> That should be amended to "some versions of Open Firmware", and the
> problem is more accurately that, when using an EIDE drive with those
> versions of OF, we can't both boot from and then use a file system on
> the same (EIDE) disk. (The issue being that OF leaves it in a state
> we're not familiar with, and we don't know how to recover.)

The problem more specifically is that OF initializes the IDE
controller in 0646U2 mode which the pciide driver does not
(yet) understand sufficiently, even after Manuel's addition of U2
support for pciide. It appears that OF does other strange things with
the interrupts of that chip.  My work-around has been to move the IDE
disk to the second connector on the motherboard that normally connects
to CD-ROM drive. The boot drive is now ide0/disk@.... instead of
ide1/disk@ and you don't have a CD-ROM anymore (unless you
re-cable) but you now can boot and run from the same IDE drive.

 > 
> It'd be pretty hard to boot from "a UFS disk" anyway, since we use BSD
> FFS (and don't support UFS at all, last I checked; some System
> V-derived OSes use it... SunOS for sure).
> 
> Despite extraneous firmware bugginess, all five of my OF 1.0.5 and all
> eight of OF 2.x machines boot just fine, both from SCSI and IDE hard
> drives. (Well, to be fair, there are five 6500s not actually booted
> yet, but they're identical to the one that's up.)

The work-around I described works for G3/G4's with OF 3.x and a
CMD646U2 ide controller.

-- 
    - aew