[NTLUG:Discuss] Promise Bootstrap stuff

[Aaron Goldblatt]

I just bought a Promise IDE controller (Ultra100TX2) to put into my Big
Server Box (an AMD K6-2/350 with 256 megs of RAM, w00h00!) in an effort
to do away with one of my machines, and thus have less to administer
(and fewer points of failure).

The short version is that it worked but it's funky.

I now have four hard drives and two CD drives (one RW, one ROM) in this
machine.  The Promise controller offers two IDE channels; the on-board
Via chipset offers the other two.

It looks like this:

ide0 Promise 0
ide1 Promise 1
ide2 Via 0
ide3 Via 1

ide0 hda
ide1 hdc hdd
ide2 hde 
ide3 hdg hdh (cd drives here)

It works okay.  It even boots properly, and it's running.  I'm happy.

But I noticed a quirk.  Orignally, hde came from a bootable Linux box,
so it's still got boot block on it.  hda was this machine's boot drive
before the changes, so it's got a boot block on it.

I can tell from the way it boots (watching the startup scripts) but also
by observing the differences in Lilo that the system is bootstrapping
from hde, which is then booting the kernel on hda.

It works; Lilo is happy except that it gripes about a drive geometry
problem when I rerun lilo.  It successfully executes and the changes I
made to lilo.conf are reflected on the next boot.

The kernel, if it makes any difference, is 2.4.18-pre9.

My question is:  Since it seems to work, and I'm planning on this being
a pretty stable configuration for the time being, should I worry about
getting it to bootstrap from hda, or should I leave it alone?

ag