[CLUE-Tech] LILO Boot Problem

[Jed S. Baer]

On Wed, 12 Feb 2003 09:15:59 -0700
Randy Arabie <randy at arabie.org> wrote:

> On Wednesday, 12 February 2003 at  8:34:16 -0700, Jed S. Baer
> <thag at frii.com> wrote:
> > On Wed, 12 Feb 2003 06:59:43 -0700
> > Randy Arabie <randy at arabie.org> wrote:
> > 
> > > I think my options are:
> > > 
> > >   --Repartition the disk, creating smaller partitions that the bios
> > >   can handle. Maybe a /, /boot, and /var (in addition to my swap).
> > 
> > AFAIK, the bios doesn't know squat about partitions. It knows only
> > about how to get to blocks on the disk, via head/cylinder/sector
... 
> My thought was by repartitioning I could ensure that my /boot, and all
> the necessary files lilo needs to bootstrap would be within the address
> range of my apparently limited BIOS.

While I can't cite specific references, that has the "ring of truth" to
me. I've always created /boot as a seperate partition. While I've never
dived in deeply to the whole geomety thing, I had one interesting
configuration issue a while back. Even though I created it first (in Disk
Druid, during install), there was apparently something strange about my
/boot partition on my current box, which I'm assuming is the 1024 boundary
issue. However, at the time (using Gentus Linux -- built by ABit from
RH6.x), LILO handled it just fine. When I later installed a newer RH
(Gentus wasn't being maintained), LILO choked. What I wound up doing was
booting from floppy, and installing the Gentus distro LILO. So apparently,
whatever patch ABit made didn't get into the "newer" LILO. However, RH7.2
caused me no trouble. FWIW, I'm at lilo-21.4.4-14 now. But the question
which has been in my mind was how to "force" a partition to "first one on
the disk" using Disk Druid, since my conclusion was that it didn't
necessarily follow order of specification.

Maybe you need a newer LILO? (Or GRUB)

> > My memory might be wrong here, but doesn't the P133 predate LBA? Or,
... 
> I think P133 may predate LBA, but I don't know.  I'm putting this box
> together from "scrap".  The case, mobo, and processor all came together,
> the rest (HD included) are spare parts I had.  I used the BIOS' HD
> auto-detect, and it has NORMAL, LARGE, and LBA options.  The only
> geometry that matches the label on my HD is that reported with the
> NORMAL option.

It's been so long since I looked at the BIOS screen on my P133 (haven't
turned it on in over a year), that I don't remember whether that's normal
or not. I do remember fussing around with LARGE vs. LBA, but it might not
have been on that machine. Have your tried booting the box with those
options, or just NORMAL. The whole point of LBA is that the drive remaps
the geometry, in effect reporting something the BIOS can deal with. It
lies. Typically, in the past anyway, it would over-report the number of
heads, and underreport something else, because the BIOSes of those days
couldn't deal with the number of cylinders (I think) used by the larger
drives. The drive itself would re-map the RAW IO request coming from the
BIOS to correspond to the actual head/cylinder mapping. So, using LBA
mode, the geometry won't look right. But if the reported size is correct,
it's probably OK.

jed
-- 
I wouldn't even think about bribing a rottweiler with a steak that
didn't weigh more than I do. -- Jason Earl