I'm install base Debain system on an old P3 box, with SATA/RAID controller on it.
The box does not support SATA at all! Using rescuecd I have boot the newly installed system "rescuehd root=/dev/sdb1" and make a boot floppy, using "mkboot -r /dev/sdb1". It is working, boot the right kernel image - that is good.
The floppy now day is quite unreliable. In addition I would like to make a monolityc kernel, all neded modules compiled into the kernel - kernel image could not to be placed on a simple floppy (1.44 MBytes);( - that is wrong.

How to make a boot CD, same simple as boot floppy (it is contain a lilo.cfg a map and an image to be load)?
(I do not need live CD)

Can you just use the Ultimate Boot CD or supergrub?

Method 2 of Task J3 in the last link of my signature has the exact solution you after.

No, I think not, by two reasons:
1. I need the real installed kernel, on using of live CD (such as rescuecd or ultimate boot CD) it uses his own kernel, what I'm customize for my box.

2. I need a NON interactive boot process, I do not want to run every time when power failure, to boot my box using keyboard and display.

Good work! I need to learn grub - I'm always use good, old lilo in my practice. First of al what is the real difference between them - seem to be the grub have a special file "stage2_eltorito" and allow to make a bootable iso image, which could be burn with just any CD/DVD burning app.

grub, for first time, does not work for me.
Disk configuration:
- IDE: /dev/hdb - it is simply a CDROM
- SCSI: /dev/sda - one SCSI disk
- SATA: /dev/sdb - bot/root/home device
/dev/sdb1 - root - grub installed here(?)
/dev/sdb5 - /home
/dev/sdb6 - swap

I'm reinstall Debian 4.0 base configuration, with grub on root partition. Using rescuecd (as later) I'm boot new system's root (with rescue cd's kernel 2.6.22) and make grub boot floppies as described in grub manual. Reboot, with floppy ...
root (hd1,0)
kernel vmlinuz
initrd /initrd.img
boot
after get5 all block devices including IDE, SCSI and SATA it is stuck on "Waiting for root filesystem ..."
after timeout "Check root bootarg cat /proc/cmdline
or missing modules cat /proc/modules ls /dev
ALERT! does not exist. Dropping to a shell!"
What to do?

I thought this thread is to make a bootable CD?


You make a bootable floppy while in a root terminal of Debian (use "su then followed by root password)

Method 1

Method 2

Your Grub commands may work if it is like this
You need to have Grub built into Debian. Let me know if it doesn't

> I thought this thread is to make a bootable CD?
Yes! but with a base Debian installation it is not convenient (at least).
My target is to rebuilt/renew my home server, it must be reliable, as possible fast and convenient (file server, printer server, FAX server, e-mail client, database server ...)
First of all I have realized that the best choice (quality/finance) is move to SATA. It is also proven by some primitive benchmarks (such as hdparm) that the best perform is SATA as hotplug for backup (plug in an USB 2.0 external hdd is slow not only the USB but hole SCSI and IDE subsystem - about 30 %). Unfortunatelly my box does not support it more then an addon SATA/RAID card - at first it could not boot through usuall BIOS settings. I can boot from floppy-CD - IDE HDD - SCSI addon (last I do not try).
I want to build box, using SATA hdd as the main "bulk" storage including the system partition. Using a SATA mobil rack I can (manually) periodically save hole system and files as quick as possible, and if the main hdd goes wrong, I can simply replace with a more or less fresh backup - I've got two similar SATA hdd for this purpuse.

Now about booting!
I never use before the grub - lilo work good for me (I have managed many boxes with dual boot with windows xp and linux). More lilo have a good working utility script "mkboot", but it could make only floppy boot disk I need the same feature for CD, reasons:
1. Floppy does not reliable!
2. Boot floppy have only 1.44 Mbytes storage capacity, what mean I can not use monolitic or larger kernel! - boot should contain at least kernel, map, and a bottstrap software.

root (hd1,0)
kernel /boot/vmlinuz root=/dev/sdb1 ro
# it is strange, /boot/vmlinuz is wrong, in Debian there is a link to the /boot directory
# /vmlinuz -> /boot/kernel-2.6.18-4-686 - which is the real bzImage
# root=/dev/sdb1 ro - what is this? Is these are kernel boot arguments? In GRUB
# manual 0.97 "kernel2 command does not realize another "root" command!
initrd /boot/initrd.img
# same as for vmlinuz - there is alink to the /boot directory to file initrd-2.6.18-4-686.img

I will try it - this means reainstall Debian using netinst with bootloader grub;(

I only use the vmlinuz and initrd files supplied by you.

The common practice is the vmlinuz should be symbolic=linked to the latest kernel. So is initrd file. They could be placed in the root directory "/".

In General you can boot to any latest pair of the vmlinux and initrd in the /boot directory which should have everything.

Let me tell you what you have missed out by using Lilo

(1) Grub can boot any system "manually". The commands are identical to those in menu.lst. For example when you see a Grub screen in Debian, instead of selecting a system you can press the "c" key to get a Grub prompt to boot any PC system.

(2) There is no installed PC a Grub prompt cannot boot. Using your example you can type the follwing lines to boot Debian
or simply
(3) You can write a menu.lst to boot operating systems even before you install them. Lilo will cut your head off if you do the same with lilo.conf.

(4) Grub can be installed in a floppy, CD, USB device, internal hard disk or external hard disk.

(5) Grub can be installed without an operating system attached or even booted by Windows' MBR.

(6) Lilo need between 2 to 3 times work than Grub to multi-boot because you need to update Lilo every time you alter lilo.conf. For Grub never!

(7) Lilo booting screen is static and can only accept 27 images to boot. Grub uses a scrolling screen so that you can throw 200 systems at it.

(8) You can play tunes with Grub by having a set of configuration menus in the same Distro to boot a very complex set of operating systems. Lilo has only one lilo.conf. See this example.

Thanks for your help saikee!
You give me a good comparison why grub is better then lilo!
After many hours of hardware testing, many different configurations I decided to use a small old IDE hdd - about 10 G should be anough for me (my old system partition was about 2 G). SCSI subsystem has not only SCSI drives (which could be up to 15 pieces) but porting/hosting SATA and USB devices! Every time when I'm attaching something to this system I would have to manage change my boot drive, fstab and hole system goes instable. I think better to use a slow and small, bet reliable IDE harddisk to hold the system, and mounting using fstab everything - as last I need to rewrite fstab and do not risks boot
I have seen some tool to manage permanent SCSI -> devicename connections - scsidev, I would better work with these.

PS: on last try to install debian on SATA I have choose to install grub into "master boot device". It wasa vonderful, system booting from SATA w/o problems. After I was inspoected the system and I realized that the master boot device was a SCSI harddisk which is suppose to be use for database server (hidden from users) - nice, but is not so much usefull, to have master boot record on a disk only for data But grub works excellent - i would use it!

The Master Boot Record is the first sector of your bootable hard disk you nominated in the Bios.

You may have a war on your hand thinking of using a Pata disk will get you out of the current hole. Linux with kernel 2.6.20 and newer now standardised all hard disk names as sda, adb, sdc, sdd, sde, sdf...etc for every Pata, SCSI, Sata or USB hard disk. Only older kernels recognise and use hda, hdc, hdd and hdd.

As long as you keeps an eye on which one is the boot disk the names sda, sdb...etc can be made permanent becasue they are generated according to the order the hard disks are detected in the mobo.

> Only older kernels recognise and use hda, hdc, hdd and hdd.
Are you sure? The rescue CD uses linux-2.6.20.11 and there are hda, hdb, hdc and hdd?
The problem is that for reliable boot process need a fix hdd, on that case it would have the fix device name. For this purpose the best choice (I think) is IDE, for my old box.

But you are not booting the rescue CD permanently, aren't you.

Some distros still try to maintain the Pata disk names for the interim period but the trend is to move over because new kernel use the libATA driver to detect all hard disks. It is not only the device names that matter because a Pata can have 63 partitions but the unified sda, sdb etc cannot have more than 15 partitions. Some installers actually recognised your Pata disk device names and use them but the installed Linux switches to the new notation.

You don't have a problem if you do not install or use a newer Linux in your box.

> The best computing teacher in the world is the dumb box in front of me. It knows everything
> and doesn't lie.

I'm affraid, you right!
I do not working with many distros, I1m use only Debian, if I want to use all of it is power and benefits (security) you need to move
Also I'm not sure that is it impossible! - my "old" configuration is working on linux-2.6.18.1 for many years!