Very Computer

I needed the SCSI drive out of my desktop for another machine, so I added
an IDE drive, made similar-sized partitions, and copied the filesystems
over with cpio.  I also adjusted /etc/fstab etc. to change sda to hda,
and re-ran lilo to reinstall LILO.  "lilo -v" doesn't show any errors.

I can boot the machine with a custom boot disk using "rescue
root=/dev/hda1" and it works fine.  But when I boot from the hard drive,
LILO only gets to "LI" and then stops.

The LILO docs say that this means either that I've moved things in /boot
without rerunning lilo, or a geometry mismatch.  I know the former isn't
the case (I've rerun lilo lots of times).

When I boot with Tom's bootroot, the number of heads on hda shows up
as 0, and the number of cylinders is wrong.  This makes me wonder about
the geometry problem.

Here are the things I've tried so far:

- Copying the MS-DOS (standard) MBR using "dd bs=446" onto the disk, and
  switching lilo.conf to have boot=/dev/hda1 (which is marked active)
- Adding "linear" to lilo.conf -- is this needed since the BIOS is set
  to use LBA?
- Adding disk geometry to lilo.conf -- using an "hda=" argument with the
  Tom's disk does make the geometry right, but it doesn't fix LILO

Can anyone suggest what I should try next?  It doesn't seem like
reinstalling this machine should be necessary.  Also any pointers to
places I can go to read more, troubleshooting tools, etc... most of this
seems pretty opaque, and I don't know any way to inspect what's on the
disk to figure out where it's falling over.

Thanks in advance for any help!

----------machine details-----------------------------------------------
Intel Seattle 440BX motherboard (Phoenix BIOS)
        (IDE drives set to "Auto" and "LBA" in the BIOS)
RedHat 6.1 (2.2.12 and 2.2.14 kernels)
LILO 0.21
hda: ST38410A, 8223MB w/512kB Cache, CHS=16708/16/63
hdc: ST39140A, 8693MB w/448kB Cache, CHS=17662/16/63
Adaptec 2940UW with SCSI CD-ROM and tape

----------disk details--------------------------------------------------
Disk /dev/hda: 16 heads, 63 sectors, 16708 cylinders
Units = cylinders of 1008 * 512 bytes

        Device Boot    Start       End    Blocks   Id  System
        /dev/hda1   *         1       261    131430   83  Linux
        /dev/hda2           262     16706   8288280    5  Extended
        /dev/hda5           262      4423   2097534   83  Linux
        /dev/hda6          4424      6504   1048710   83  Linux
        /dev/hda7          6505      7545    524550   82  Linux swap
        /dev/hda8          7546     16706   4617030   83  Linux

Filesystem           1k-blocks      Used Available Use% Mounted on
/dev/hda1               127269     50113     70585  42% /
/dev/hda8              4544532   3779992    533692  88% /export
/dev/hda5              2064144   1192256    767012  61% /usr
/dev/hda6              1032056    134680    844944  14% /var

----------/etc/lilo.conf------------------------------------------------
boot=/dev/hda
#boot=/dev/hda1
root=/dev/hda1
map=/boot/map
install=/boot/boot.b
prompt
timeout=50
default=linux
read-only
#linear

#disk=/dev/hda
#    bios=0x80
#    sectors=63
#    heads=16
#    cylinders=16708

image=/boot/vmlinuz-2.2.14-12
    label=linux
    initrd=/boot/initrd-2.2.14-12.img
#    append="hda=16708,16,63"  # Redundant?
    password=foobar
    restricted

image=/boot/vmlinuz-2.2.12-20
    label=linux.old
    initrd=/boot/initrd-2.2.12-20.img
#    append="hda=16708,16,63"  # Redundant?
    password=foobar

------------------------------------------------------------------------

--

W: +1 651 638-6136

      Could it be that your old kernel (+mods) only had support for SCSI
and not IDE?  Rescue disks often have IDE support, and so may be able to
read your drive when your own kernel cannot.  

diwan.vcf < 1K Download

Quote:>I needed the SCSI drive out of my desktop for another machine, so I added
>an IDE drive, made similar-sized partitions, and copied the filesystems
>over with cpio.  I also adjusted /etc/fstab etc. to change sda to hda,
>and re-ran lilo to reinstall LILO.  "lilo -v" doesn't show any errors.

                                            Snip

 Brent
         If you have just copied from your Scsi Disk to your IDE Disk the
kernel will have the wrong root dev i.e /dev/sda1 instead of /dev/hda1.
The fact that you can boot using a rescue disk with root=/dev/hda1
seems to prove this. What you  need to do is boot using your rescue
disk and then rdev kernelname of the kernel on your hard disk, this
will show what root device the kernel is looking for. Then you have to
rdev kernelname /dev/hda1 assuming hda1 is your root device. Now
do rdev -R kernelname 1 to mount root device in readonly mode. If
you then dd if=kernelname of=/dev/fd0 you will create a boot disk of
your kernel, if you can boot off this floppy your kernel is o/k. Any more
problems with lilo must then be due to lilo.conf. See man rdev.

                                    Best of Luck
                                      Phil Jones

No, this is the "stock" kernel from RedHat; I didn't build it myself.
And I did have one working IDE drive when my primary was a SCSI.

--

W: +1 651 638-6136

Ah, "rdev"!  That's a useful tool.  Thanks for pointing it out.
Well, "rdev kernelname" showed /dev/sdd4 but there was no such device
on my system (the root was /dev/sda1).  I think this is because it's the
"stock" RedHat kernel (I didn't compile it); that's probably some box at
RedHat.

So I changed it to /dev/hda1 and (just to be sure) rebuilt the initrd
and reran lilo.  Sadly, I get the same "LI".  I also tried again with
and without geometry, with and without "linear"; nothing works.

Quote:> Now
> do rdev -R kernelname 1 to mount root device in readonly mode. If
> you then dd if=kernelname of=/dev/fd0 you will create a boot disk of
> your kernel, if you can boot off this floppy your kernel is o/k.

Another helpful tip; I didn't know you could do that.  Well, the
readonly flag was already set.  I made the boot disk, and it works
fine.  So:

Quote:> Any more problems with lilo must then be due to lilo.conf.

I just wish I knew what they were.  :-)  Thanks, Phil, for an informative
reply.

I did notice something in the log I hadn't seen before:

"hd0: C/H/S=64448/0/63 from BIOS ignored"

and then it goes on to print "hda:" with the correct geometry.  So I'm
really suspecting that somehow the odd geometry reported from the BIOS is
the culprit here... even though I've specified it in the lilo.conf file.

I also tried John Coffman's disk geometry checker; it reported some
very strange C/H/S numbers and "0 sectors = 0K", but the EDD numbers
were correct.  Maybe I should try this new lba32 mode.

--

W: +1 651 638-6136