Yes, the scsi drive is recognized and the appropriate drivers are loaded (I
   did a mkinitrd --preload scsi_mod --preload sd_mod initrd.scsi <kernel
   image>  and  verified  that the modules were included in the initrd by
   mounting it). I also verified that the root filesystem is indeed ext3 (btw,
   how does the system know if the filesystem is of a particular type before it
   gets to /etc/fstab? I haven't figured that out yet) and it was clean. 

     (!) [Ben] Presumably by reading the partition table and doing a little
     string comparison, just as "file" does.

   (?) From what I was able to google, one person said that the journal might
   have been trashed (fixed with fsck) and another said that there was no
   corresponding device in the initrd, which is true. But do I need to create
   a device for the /dev/sda in the initrd and if so, how? 

     (!) [Ben] I've found that a lot of headaches can be avoided by simply
     deleting the journal when there's a related problem. There can be much
     darkness of the spirit and wailing and rending of clothes otherwise.

     (!) [Faber] Yes, it's a bash script called linuxrc. It looks pretty
     straightforward:

#!/bin/bash

echo "Loading scsi_mod.o module"
insmod /lib/scsi_mod.o
echo "Loading sd_mod.o module"
insmod /lib/sd_mod.o
echo "Loading ncr53c8xx.o module"
insmod /lib/ncr53c8xx.o
echo "Loading jbd.o module"
insmod /lib/jbd.o
echo "Loading ext3.o module"
insmod /lib/ext3.o
echo Mounting /proc filesystem
mount -t proc /proc /proc
echo Creating block devices
mkdevices /dev
echo Creating root device
mkrootdev /dev/root
echo 0x0100 > /proc/sys/kernel/real-root-dev
echo Mounting root filesystem
mount -o defaults --ro -t ext3 /dev/root /sysroot
pivot_root /sysroot /sysroot/initrd
umount /initrd/proc

     So straight forward that they're not even using 'modprobe'! -- Thomas Adam

     (!) [Ben] Saaay... that looks interesting. What's that "mkdevices" bit,
     and how does it know which devices to "mk"? That's assuming that the lack
     of an appropriate device is the problem. I'd also add a "-v" to all those
     "insmod" invocations, just for grins - and maybe kill those "echo" lines.

   (?) Any other suggestion for fixing this are welcome. As I've said, I've
   fixed three other [92]Red Hat/SCSI problems within the past few weeks ( 1.
   boot  off of an IDE and mount the scsi, 2. upgrade to Red Hat 9, or 3.
   preload modules in initrd) but short of reinstalling the OS, I can't figure
   out what to try next. 

     (!) [Ben] "pivotroot", IIRC, is only necessary if you want to do an
     "initrd"-type boot, where you fire up a RAMdisk, boot off that, then mount
     your partition on '/' as ext3, and away you go. Mind you, this is all from
     memory - it's been a long time since I did this myself, and the only
     reason for doing it was that I wanted to try a [93]Debian-precompiled
     kernel. These days, I just make an ext3 partition from the start. However,
     you're talking SCSI, so it looks like you're stuck with doing it that way
     - unless you stick a small IDE HD in there just for booting.

     What I'd run into previously is, "mkinitrd" actually uses "/etc/fstab"
     when building the image. I ended up having to tweak "/etc/fstab" to make
     it fit the machine I was building it for - as opposed to the one on which
     it was being built - and then untweak it after running "mkinitrd".

   (?) I've tried that as well. :-( 

     (!) [Ben] Argh. Well... let's peel it back as much as we can. It's been a
     mort of years since I've dealt with SCSI HDs, so I don't know how helpful
     this will be, though.

     First off, I'd go ahead and pop in a Knoppix CD, boot with it, and see if
     I can detect/mount/read the SCSI - and I would watch carefully to see just
     how different the SCSI-related messages are during the process (i.e., do
     you  need  to try a different module?) Then, I'd make sure that the
     "/etc/lilo.conf" on the SCSI is trying to do The Right Thing (you are
     using LILO, right?) - no IDE-specific stuff in there, "root", "boot",
     etc. are set to the correct values (not an "hdX" in the place...) and
     re-run "lilo -v" just to see what the output is and make sure that it's
     properly set. I'd walk through the "initrd" setup to make sure that
     there's  nothing  odd in there, and particularly check that ROOT in
     "/etc/mkinitrd/mkinitrd.conf" is explicitly set (I just remembered having
     a problem with that default "probe" setting at one point; my current one
     says    'ROOT="/dev/hda1    ext3"'.)   Also,   take   a   look   at
     "/etc/mkinitrd/modules".

     Nothing else comes to mind at the moment.

   (?) The partition table holds filesystem information? I didn't see anything
   in the list of types in fdisk. 

     (!) [Ben] Sorry, I should have expanded that. The kernel would read the