Re: Please Complete the Docs

[Felix Miata]

Matthias Granberry wrote:

> On Sun, 2 Sep 2001, Felix Miata wrote:
 
> > All GRUB documentation makes reference to "/boot/chain.os2" without
> > making any reference to how to get or create it. The docs make no
> > suggestion about how to boot OS/2 from (hd0,5) or (hd1,9). Simple fails
> > to boot OS/2 Warp 4.5 from anything other than (hd0,[1-4]). If this
> > capability is incomplete, please so indicate in the docs and TODO list.
> > If it is complete, please indicate how to implement in menu.lst and
> > anything else that may be required.

> I don't know exactly how to find the chain.os2, but I do know that the
> only way that I have managed to boot OS/2 via the grub was to chainload
> the OS/2 boot manager (or whatever it's called) from a separate
> partition.  I imagine that there is a simpler way to go about it, but I
> have yet to find it.
 
> As I recall, a number of linux distros have a chain.os2.  I know that the
> redhat machine that I am using right now in the comp lab has a
> 640-byte file called "os2_d.b" sitting in /boot.  I imagine this is the
> same thing plus 128 bytes of metadata.
 
> You might be able to install os/2 and copy the boot sector by typing dd
> if=/dev/hda of=chain.os2 bs=512 count=1 after installing it.  If it
> doesn't clobber the mbr, then dd if=/dev/hda(whatever partition it boots
> from) of=... should probably work instead.
 
> I hope this clears it up.

Not terribly clear to someone still new at Linux, but certainly
something worth trying - except maybe . . .

There is no problem making OS/2 boot directly via chainload from
/dev/hda[1-4]. This is no different from using grub to boot windoze. One
difference I'm aware of with booting OS/2 from /dev/hda[5-x] is that
those partition boot records are not on their first sectors, but instead
at the first sector of the second logical head, usually 63 (64?) sectors
after beginning of cyl on a modern (large) HD. Seems to me like at the
very least, that X'd sector would be the one to dd, assuming dd of a
particular cylinder is all that is required for grub to load OS/2.

I suspect that that alone might not be sufficient. My guess is that the
lost first track would need to be accounted for with some added
metatdata or modification of that sector in order for OS/2 boot success.
Maybe not. Worth a try anyway.

The reason getting Grub to boot OS/2 without the IBM Boot Manager has to
do with a shortcoming of BM on a >8Gb HD on which windoze must coexist.
IBM refuses to fix the problem caused by the M$ introduction of the 0Fh
extended partition type. IBM BM and storage access drivers pretend the
contents of 0Fh partitions don't exist, and BM is incapable of changing
partition types during boot management, while windoze corrupts partition
contents when windoze accessible partitions exist within a non-0Fh
container beyond cyl 1023. Grub would solve this incompatibility if it
could load OS/2 directly from beyond /dev/hda4, due to its ability to
set the types as appropriate for the selected non-Linux OS. Chainloading
IBM BM looks like a partial solution at best.
-- 
"The highest glory of the American Revolution was this: it connected,
in one indissoluble bond, the principles of civil government with the
principles of Christianity."                President John Quincy Adams

Felix Miata  ***  http://mrmazda.members.atlantic.net/