We have a couple of legacy applications here, written in C++, that have
always run on windows and *nix. In order to support the *nix versions, we always
have to have a Linux install hanging around (we also have a solaris, but that's
already set up).
So we got a Dell Precision 380 workstation with a SATA drive that had Windows
XP installed on it. Since the Dell didn't come with a floppy drive, I couldn't
very well just create a boot diskette. That would be useless. So based on the
fact that there is a spec called "El Torito", I looked up an article online about Linux Bootdisks. It
looked pretty good, but there was one glaring omission: the author mentions that
you need to load any initial ramdisk via LILO, but doesn't explain how to setup
LILO correctly for a floppy image. I decided to go back to it if necessary and
went to redhat.com to see if there was any insight there.
I ended up finding a section of documentation that outlines how to create an Installation Boot CD-ROM,
but I had already installed it! I later found that if at the bootloader screen
of the install CD I typed the command:
It would in fact load the installation off of the second (IDE) harddrive
(thanks, btw, linux for making me use a modified kernel to support SATA drives.
That's top-notch, guys.)
So I figured that the secret to the sauce was probably in the isolinux.cfg
file that configures all the commands visible at the boot loader screen. First I
tried something like this:
append root=/dev/hdb2 initrd=initrd.img
but that didn't work, it still brings up the install screen. So I figured why
not try the simplest possible solution using the config file, and this is what I
came up with:
default vmlinuz root=/dev/hdb2
Guess what? I now have a Redhat Linux 9 install that will boot from a CD and
I didn't have to mess with a bootmanager or touch the MBR in any way. I don't
have to wait those pesky 10 seconds for either grub or the windows NT bootloader
to select which install to start - I just pop in a CD if I want linux, and open
it if I want WinXP.