Files
oldlinux-files/study/boot/rootdiskette/bootloaders.tex
2024-02-19 00:25:23 -05:00

1467 lines
62 KiB
TeX

\documentclass[12pt,a4paper]{article}
\usepackage{times}
\usepackage{html}
\title{Linux Boot Loaders Compared}
\author{L.C. Benschop}
\begin{document}
\maketitle
Copyright \copyright{}2002, L.C. Benschop, Eindhoven, The
Netherlands. Permission is granted to make verbatim copies of this
document.
\tableofcontents
\section{introduction}
If you use Linux on a production system, you will only see it a few
times a year. If you are a hobbyist who compiles many
kernels or who uses many operating systems, you may see it several
times per day. Of course I mean the boot loader.
There are several different ways to boot Linux and every method has
specific advantages and drawbacks. Therefore it is important to know
which boot loading methods exist for Linux and how they compare to
each other.
We discuss only boot loaders that meet the following criteria:
\begin{itemize}
\item They must support Linux. The boot loader must load the Linux
kernel itself, boot loaders that could chain load another loader,
which in turn boots Linux, are not discussed,
\item They must run on the Intel PC platform.
\item They must be freely available complete with source code.
\item No network boot loaders are discussed.
\end{itemize}
We discuss the following boot loaders:
\begin{itemize}
\item Boot Sector for the Linux kernel.
\item LILO
\item GNU GRUB
\item SYSLINUX
\item LOADLIN
\end{itemize}
Other less well-known boot loaders do exist and may support the
criteria. I know of the following:
\begin{itemize}
\item
\htmladdnormallinkfoot{nuni}{http://www.ibiblio.org/pub/Linux/system/boot/loaders/}
is a boot loader that does not use BIOS, but is otherwise rather
limited. It boots {\em only} from IDE disks.
\item \htmladdnormallinkfoot{Gujin}{http://gujin.sourceforge.net} is a
boot loader that understands file systems, just like GRUB.
\item \htmladdnormallinkfoot{{\tt mbr03} and {\tt
e2boot}}{ftp://spruce.he.net/pub/jreiser/} are an MBR boot
selector and a matching Linux boot loader that fits into the first
1k boot block of an ext2 partition, so no file space is used by them.
\end{itemize}
Note: some boot loaders are rapidly evolving, so some features may
have been added after this article was written.
\section{How Boot Loaders Work}
\subsection{What BIOS does for us}
The BIOS is the firmware in the ROM of a PC. When the PC is powered
up, the BIOS the first program that runs. All other programs must be
loaded into RAM first. The BIOS contains the following parts:
\begin{itemize}
\item POST (Power On Self Test). The running counter that counts the
kilobytes of main memory is the most visible part of the POST.
\item The Setup Menu, that lets you set some parameters and lets you
adjust the real time clock. Most modern BIOS versions let you set
the boot order, the devices that BIOS checks for booting. These can
be A (the first floppy disk), C (the first hard disk), CD-ROM and
possibly other disks as well. The first device in the list will be
tried first. Older BIOS-es have only one boot order: A, C. So the
BIOS will try to boot from A first and if there is no diskette in
the drive it tries to boot from C.
\item The boot sector loader. This loads the first 512-byte sector
from the boot disk into RAM and jumps to it. This is where the boot
loaders described in this paper start.
\item The BIOS interrupts. These are simple device drivers that
programs can use to access the screen, the keyboard and disks. Boot
loaders rely on them, most operating systems do not (the Linux
kernel does not use BIOS interrupts once it has been started). MSDOS
does use BIOS interrupts.
\end{itemize}
Apart from the main BIOS there are extension ROMs, which are started
by the main BIOS. Every VGA card has one. Also SCSI host
adapters and Ethernet cards can have an extension ROM. It is even
possible to put an EPROM on an Ethernet card to boot a machine over
the network without any disks.
As far as boot loading facilities are concerned, the PC BIOS is very
primitive compared to that of other computer systems. The only thing
it knows about disks is how to load the first 512-byte sector.
\begin{itemize}
\item The first sector of a diskette can be loaded at address
0000:7C00. The last two bytes of the sector are checked for the
values 0x55 and 0xAA, this as a rough sanity check. If these are OK,
the BIOS jumps to the address 0000:7C00.
\item Booting from a hard disk is very similar to booting from a
diskette. The first sector of a hard disk (often called the Master
Boot Record) is loaded at 0000:7C00 and next the BIOS jumps to
it. The MBR program must move itself to an address that is different
from 0000:7C00 as it is supposed to load a different boot sector
from a partition to address 0000:7C00 and jump to that.
\item Modern BIOS versions can treat a certain file on a CD-ROM as a
diskette image. They pretend to boot from a diskette by loading the
first 512 bytes of the file to 0000:7C00 and jumping to it. Every
attempt to access the same diskette using the BIOS routines, will be
redirected to the image file on CD-ROM. Some other ways to boot a
CD-ROM may also be supported (with an emulated hard disk or with no
disk emulation at all).
\end{itemize}
When the boot sector is loaded, the CPU is in real mode. For those who
are unfamiliar with 80x86 architecture: real mode is very limited
compared to 32-bit protected mode (in which Linux runs). For example:
data outside a 64K segment can only be accessed if you change a
segment register and data outside the first 1MB of address space
(which contains 640kB of main memory) cannot be accessed at all. As
{\tt gcc} does not know about real mode, programs compiled with it can
only be run in real mode with some tricks and with severe memory size
restrictions. This is the reason why most boot loaders (except GRUB)
are written in assembly. All boot sector programs, even that of GRUB,
are written in assembly\footnote{Real mode was not one of the
strongest points of the GNU assembler, not until recent
versions. Therefore two other assemblers have been used for real-mode
programs such as boot loaders. These are {\tt as86} (in use by LILO
and until recently by the real mode code in the Linux kernel) and {\tt
nasm} (in use by SYSLINUX).}.
In theory a boot loader could do its job by directly accessing the
bare metal, but 512 bytes would be a very small space for that.
The boot loader has access to BIOS interrupts, which are subroutines
that can be invoked by the INT instruction (software
interrupts). These work in real mode only. The following routines are
used by most boot loaders.
\begin{itemize}
\item INT 0x10 for screen output.
\item INT 0x16 for keyboard input.
\item INT 0x13 for disk I/O. The parameters to specify sectors on disk
used to have a very limited range. Originally it was only possible
to specify 1024 cylinders on a hard disk, while hard disks can have
more cylinders. This imposed limitations on where it was allowed to
put the boot loader and any files accessed by it. You were forced to
create a small partition near the start of the hard disk and put the
boot loader there. There are three categories of BIOS:
\begin{itemize}
\item BIOS versions earlier than 1995 could only access IDE disks of
around 500MB, as the BIOS sector and head numbers corresponded
directly to the register values on the IDE interface.
\item BIOS versions between 1995 and 1998 can access IDE disks up to
about 8GB. They translate the cylinder, head and sector numbers from
the INT 0x13 call to different values that better utilize the
allowable ranges on the IDE interface.
\item BIOS versions of 1998 or later have a new calling interface
using linear block addresses.
\end{itemize}
In any of the three categories you have BIOS-es that have bugs that
cause them to stop at a lower disk size limit (a category 2 BIOS
that should work up to 8GB, but stops at 2GB). In those cases it
makes sense to upgrade to a new BIOS version.
\item INT 0x15 is a catch-all for many BIOS functions, one of which
is moving data to extended memory (the BIOS is required to switch to
protected mode temporarily to do that). Other functions are for disabling
the A20 gate and for determining the memory size.
\end{itemize}
\subsection{Parts of a boot loader}
A boot loader typically consists of three programs:
\begin{itemize}
\item The boot sector program is directly loaded by the BIOS at boot
time and is only 512 bytes in size.
\item The second stage program is loaded by the boot sector program
and it does everything you expect the boot loader to do.
\item The boot loader installer is not run when the system is booted,
but it is used to install the boot loader and the second stage
program onto the boot disk. These have to be stored in special
locations, so they cannot be copied with {\tt cp}.
\end{itemize}
\subsubsection{boot sector program}
The boot sector program can only be 512 bytes in size and not all 512
bytes are even available in all cases. The last two bytes must be 0x55
and 0xAA for the BIOS. The Master Boot Record on a hard disk contains
the partition table of 64 bytes, so only the first 446 bytes can be used.
If the boot sector program must exist on a DOS partition or on a DOS
diskette, there must be a parameter block at the start of the boot sector.
Because of these size restrictions, boot sector programs are just
about the only remaining examples of programs on the PC platform that
are truly optimized for size and have to be written in assembly for
this reason. Further, a boot sector program cannot do everything you
want a boot loader to do. Usually a boot sector program does one of
the following things (not all three in one program):
\begin{itemize}
\item Load another boot sector. This is typical for a boot sector
program that lives in the master boot record of a hard disk. It can
find the first sector of the selected {\em active} partition and
{\em chain load} that. The MBR program that came traditionally with
MS-DOS has no ability to change the active partition at boot
time. There are other boot sector programs that let you select a
partition by pressing a key, such as the MBR program of LILO.
\item Load a second stage boot loader. It is generally not possible
for a boot sector program to look into the directory for a file with
a specific name and load that into memory, but exceptions exist, at
least for DOS file systems\footnote{For an interesting example see
the file {\tt sector.S} in the {\tt diagnose} subdirectory of the
LILO boot loader. The {\tt SYSLINUX} program does almost the same
thing, but really loads just the first part of the second stage
loader, which in turn loads the rest.}. Most boot sector programs
find the second stage by sector number rather than by name. The
sector numbers have to be put into the boot sector by the boot
loader installer.
\item Load the kernel directly. A kernel is typically much larger than
a second stage boot loader.
The boot sector program in the Linux kernel loads the kernel
directly into memory without the need for a second stage boot
loader. As the kernel is located in contiguous sectors on a
diskette, there is no need to traverse file system data
structures. However, for {\tt bZimage} kernels the boot sector
program cheats; it invokes a subroutine in the {\tt setup} part of
the kernel for loading the rest of the kernel into high memory.
The boot loader {\tt e2boot} fits into the first 1kB block of an
ext2 partition (it is twice as big as a boot sector program), but
with some tricks it finds both the kernel and the RAM disk by name
on an ext2 partition and loads them into memory.
Also the boot sector on a DOS disk does not utilize a second stage
boot loader to load the MS-DOS kernel files {\tt IO.SYS} and {\tt
MSDOS.SYS}. The structure of an MSDOS file system is simple enough to
find a file with a specific name in the root directory and load it
into memory, at least part of it..
\end{itemize}
\subsubsection{second stage of boot loader}
This is the real boot program. It contains the user interface and the
kernel loader. It can be anywhere from 6.5 kilobytes (LILO) to
over 100 kilobytes (GRUB) in size. It contains the following functions:
\begin{itemize}
\item User interface. It is either a simple command line (old versions
of LILO), a menu or both. It allows you to select any number of
operating systems and to specify additional parameters to the
operating system. The available options are specified by a
configuration file. Modern versions of boot loaders can show their
menu in a bitmap picture.
\item Operating system loader. loads the operating system into memory
and runs it. Alternatively we can load another boot loader specific
to another operating system and let it run. This is called chain
loading.
\end{itemize}
{\tt LOADLIN} is not a complete boot loader, but it has only the
second stage (without the user interface). It is run from DOS, and it
can make use of DOS system calls to read files from disk. What makes
its task harder than that of a normal boot loader, it that it must be
able to work its way out of some types of memory managers.
\subsubsection{Boot loader installer}
The third part of the boot loader is only run when the boot loader is
installed on a disk. As opposed to the boot sector program and second
stage, this is a normal Linux program. In the case of LILO the
installer must be rerun each time the configuration is changed or any
file has been updated. It performs the following tasks:
\begin{itemize}
\item Install the boot sector. If the boot sector will be installed in
the MBR of a hard disk or on a DOS file system, not all 512 bytes
may be overwritten, but the {\em partition table} or the {\em DOS
parameter block} must be preserved.
\item Tell the boot sector where the second stage boot loader
is. Usually it writes one or more sector addresses into the boot loader.
\item Tell the second stage boot loader where all relevant information
is (configuration, kernels). This is the case with LILO. LILO
creates a map file that contains all relevant sector addresses and
puts pointers to the map file in the boot sector and/or second stage
boot loader.
\end{itemize}
\subsection{Loading the operating system}
\subsubsection{Loading the Linux kernel}
A Linux kernel consists of the following parts:
\begin{itemize}
\item The boot sector (see {\tt arch/i386/boot/bootsect.S}). This is only
run if Linux is booted directly from a diskette.
\item The setup part with real mode initialization code (transition to
protected mode) (see {\tt arch/i386/boot/setup.S})
\item The rest of the kernel, which in turn consists of the following parts:
\begin{itemize}
\item Protected mode initialization code \\
(see {\tt arch/i386/boot/compressed/head.S}).
\item A decompression program (compiled C code, see \\{\tt
arch/i386/boot/compressed/misc.c} and {\tt lib/inflate.c}).
\item The real kernel, which is compressed. After decompression this
kernel consists of the following parts:
\begin{itemize}
\item Protected mode initialization code (see {\tt
arch/i386/boot/head.S}).
\item Main C routine (see {\tt init/main.c}).
\item All the rest. Relevant for this discussion is the RAM disk driver
(see {\tt drivers/block/rd.c}). This will be further explained in
section \ref{ramsect}
\end{itemize}
\end{itemize}
\end{itemize}
A Linux boot loader should support the following tasks:
\begin{itemize}
\item Loading the Linux kernel into memory.
\begin{itemize}
\item The boot sector (which will not be run) and the setup part of
the kernel are loaded near the top of low memory (usually address
9000:0000).
\item If it is not a {\tt bzImage} type kernel, the rest of the
kernel will be loaded in low memory at 0000:1000).
\item If it is a {\tt bzImage} type kernel, the rest of the kernel
will be loaded in high memory, starting at 0x100000.
\end{itemize}
The boot loader does not know or care that the rest of the kernel
contains a compressed part.
\item Passing a command line to the kernel. This command line can come
from a configuration file or it can be entered interactively by the
user.
\item Loading an initial RAM disk into memory and passing it to the kernel.
The initial RAM disk will be loaded near the top of high memory
above the kernel.
\item Starting the kernel. The boot loader must set some relevant
parameters in the setup part and jump to it. From there the kernel
takes control and the boot loader's part is over. Next the kernel
starts as follows:
\begin{itemize}
\item The setup code saves relevant parameters and the command line
for later.
\item The setup code checks how much memory is available, performs
some hardware initialization and prepares to enter protected mode.
\item In protected code, the rest of the kernel is being
decompressed\footnote{At this point we are not in a position to use
BIOS anymore and the kernel console driver has not been initialized
yet. The decompressor contains a very primitive screen output
routine, so it can show the ``Uncompressing Linux'' message.}.
\item After decompression the {\tt head.S} and the main C routine
will initialize all parts of the kernel. This will show a lot of
messages.
\item Finally the {\tt init} process will be started (or first its {\tt
linuxrc} variant on an initial RAM disk).
\end{itemize}
\end{itemize}
\subsubsection{Chain loading}
Most boot loaders are designed to boot only one operating system. LILO
knows only to load Linux kernels and the DOS boot sector can load
only.DOS. If you want to select between several different operating
systems, it is not likely that you find a boot loader that can load
all of them. But at least every operating system is
supposed to be booted by a 512-byte boot sector that can be loaded by
the BIOS and there lies the key. Any sufficiently advanced boot loader
supports chain loading.
If a boot loader loads a boot sector of another operating system, it
is called chain loading. This boot sector can directly be loaded from
the disk or partition or it can be loaded from a file. For the other
operating system it should not make a difference whether its boot
sector was loaded by the BIOS or by another boot loader. In reality a
boot sector of a partition is normally loaded by a master boot record
program. Also in this case, it should make no difference if this is
loaded by a more advanced boot loader (such as LILO) instead.
The following chain loading scenarios are possible.
\begin{itemize}
\item Linux boot loaders can chain load almost any other operating system.
\item Linux boot loaders can be chain loaded by boot managers of other
operating systems. It is possible to boot Linux from the Windows NT
boot manager or the OS/2 boot manager.
\item Linux boot loaders can chain load Linux boot loaders as
well. This may make sense on computers with several independent
Linux systems installed, where each Linux installation has its own
local LILO and these can be selected by a central instance of LILO
to chain load them. Instances of LILO can even exchange command
lines between them.
\end{itemize}
\subsection{Configuring the boot loader}
Both LILO and GRUB have a configuration file that specifies several
menu options, each representing either a Linux kernel or a different
operating system to boot. For each Linux kernel a command line and an
initial RAM disk can be specified. Apart from syntactic details the
contents of these configuration files look remarkably similar. But
there is an essential difference:
\begin{itemize}
\item LILO reads its configuration file at installation time. Every
time the configuration file, the kernel or any initial RAM disk is
changed, the LILO installer must be rerun. At boot time the
configuration file is not read. The second stage boot program of LILO
does not know how to find files in the file system. It relies on a map
file to find data blocks of the necessary files. This map file was
created by the LILO installer.
\item GRUB and also SYSLINUX read their configuration files at boot
time. You can install the boot loader once and just change configuration
files, kernels and RAM disk images without trouble. The second stage
boot program knows how to find files in the file system.
\end{itemize}
\section{Example Installations}
\subsection{Example root file system and kernel}
In my article \htmladdnormallinkfoot{``Getting Linux into Small
Machines''}{http://www.xs4all.nl/\~{}lennartb/linux.html} I described how
to build a kernel and a root file system to put on a Linux diskette.
The relevant files are:
\begin{itemize}
\item {\tt zImage} the Linux kernel image.
\item {\tt root.img.gz} the compressed root file system image.
\end{itemize}
You can download a boot disk image {\tt myboot.img.gz} from my home
page and extract these files from it, so you can try them out with
various other boot loaders.
\begin{verbatim}
gunzip myboot.img.gz
mount -o loop myboot.img /mnt
cd /home/lennartb/myboot
cp /mnt/boot/zImage .
cp /mnt/boot/root.img.gz .
umount /mnt
\end{verbatim}
In the following sections I will show how to install this kernel and a
RAM disk with this root file system on a diskette using each of the
boot loaders discussed in this article. Whenever possible I will
provide a boot time choice between the RAM disk and an alternative
root file system that could be on a hard disk partition. i have tried
each of the diskettes, just to make sure they really work.
\subsection{Linux Boot Sector}
\label{lbssect}
The oldest way to boot Linux is the kernel's own boot sector. This
method has several limitations:
\begin{itemize}
\item It can load a Linux kernel only from a diskette with no normal
file system on it.
\item It does not support a kernel command line. Instead, several
parameters can be set in the boot sector using the {\tt rdev}
program. At boot time there is absolutely no way to select any
options.
\item It does not support initial RAM disks. It supports a different
type of RAM disk instead.
\end{itemize}
It cannot boot from overformatted diskettes (1680kB on a 3.5''HD
diskette), but it has zero file system overhead.
If the kernel and RAM disk together are 1200kB or less you can {\tt
dd} the {\em exact same disk image} to both a 5.25'' and a 3.5'' diskette
and they will both work. There are few if any boot loaders that can
do {\em that}!
First transfer the kernel and the compressed RAM disk image to the
diskette. As the kernel is less than 450kB in size, we can specify an
offset for the RAM disk of 450kB ({\tt seek} option to {\tt dd}). This
way the kernel and the RAM disk do not overlap on the diskette.
\begin{verbatim}
dd if=zImage of=/dev/fd0
dd if=root.img.gz of=/dev/fd0 bs=1k seek=450
\end{verbatim}
We need one more thing to get this working: we need to set the RAM
disk offset in the kernel. First calculate 16384 plus the offset you
used with {\tt dd}, in our case it is 16384+450=16834. Use this
number in the following {\tt rdev} commands:
\begin{verbatim}
rdev /dev/fd0 /dev/fd0
rdev -r /dev/fd0 16834
\end{verbatim}
It is important also to set the root device to {\tt /dev/fd0},
otherwise it won't work.
We could have put the RAM disk image on a separate diskette. In this
case we should not specify an offset with {\tt dd} and the {\tt rdev}
commands must be as follows:
\begin{verbatim}
rdev /dev/fd0 /dev/fd0
rdev -r /dev/fd0 49152
\end{verbatim}
When this diskette is booted, the RAM disk will be loaded after the
kernel initialization messages are shown, as opposed to the initial
RAM disk that will be shown in other examples, which will be loaded by
the boot loader before the kernel starts initializing. This type of RAM
disk can also be used with other boot loaders.
Though the Linux boot sector cannot load a Linux kernel from an
overformatted diskette there is no such restriction for the RAM disk.
It is important that you {\tt rdev} the kernel to the correct diskette
type. For a RAM disk on a separate diskette, use the following commands:
\begin{verbatim}
rdev /dev/fd0 /dev/fd0u1722
rdev -r /dev/fd0 49152
\end{verbatim}
For the RAM disk on the same diskette there are several solutions:
\begin{itemize}
\item Format the diskette with 21 sectors per track, but write the
kernel with 18 sectors per track. The three extra sectors per track
are effectively wasted, but it works. Of course the RAM disk is
written with 21 sectors per track. The following commands have been tested:
\begin{verbatim}
fdformat /dev/fd0u1722
dd if=zImage of=/dev/fd0H1440
dd if=root.img.gz of=/dev/fd0u1722 bs=1k seek=600
rdev /dev/fd0H1440 /dev/fd0u1722
rdev -r /dev/fd0H1440 16984
\end{verbatim}
\item Change the boot sector program of the Linux kernel so it expects
21 sectors instead of 18. This should be simple, but I have not
tested it.
\item Put a small minix file system at the start of the diskette and
boot the kernel with LILO from that minix file system. The rest of
the diskette holds the RAM disk and is not used by the minix file system.
\end{itemize}
These bootable disks are limited to 21 sectors per track. The 24
sectors per track formats use fewer, but bigger sectors, which the
boot loader does not understand. The absolute maximum amount of Linux
to put on a single 3.5'' HD diskette can be achieved as follows:
\begin{itemize}
\item Format the diskette in two parts, one for the kernel with 21
sectors per track, one for the RAM disk with 24 sectors per
track. {\tt superformat} can format a range of tracks so you can
make that disk by using it twice with different parameters.
\item Put the kernel with the modified boot sector in the first area.
\item Put the RAM disk in the second area,
\item Calculate the RAM disk offset with respect to the 24 sectors per
track disk. Give appropriate {\tt rdev} commands.
\end{itemize}
Success with your copy-protected Linux boot diskette!
\subsection{LILO}
LILO is one of the first boot loaders for Linux that was capable of
booting from the hard disk, dating back to 1992\footnote{In fact {\tt
shoelace} is even older, but who has heard of that?}. It's still the
market leader and over the years it was updated with new features such
as a nice boot menu and the use of the new BIOS functions that
eliminate the 1024 cylinder limit. The basic working principle of
LILO has stayed the same. This means that most work is done by
the installer and not by the boot time code. LILO is the only boot
loader that supports booting from a RAID system.
All Linux distributions already have LILO installed, but not
necessarily the latest version. In this example we will download the
\htmladdnormallinkfoot{latest version of LILO}{
http://brun.dyndns.org/pub/linux/lilo}. At the time of writing, the
latest version of LILO was 22.3.1, Unpack the source tarball and
make. This is easy as long as you have a sufficiently recent version
of {\tt bin86} installed. The file {\tt README.common.problems} tells
you where to get it. We won't install LILO for now, but run it from
the source directory where it was built.
If you have read the \htmladdnormallinkfoot{``Getting Linux into Small
Machines''}{http://www.xs4all.nl/\~{}lennartb/linux.html} article, the
following may be familiar to you, but we will do a few things
differently:
\begin{itemize}
\item We will install on an overformatted diskette. This is not
necessary for the current kernel and initial RAM disk, but for
larger files, this may become necessary.
\item We will use the minix file system instead of ext2. This has less
overhead. The booted kernel need not support minix file systems, the
host system must.
\item We will use a boot menu.
\end{itemize}
First create the file {\tt lilo-initrd.conf} with the following
contents:
\begin{verbatim}
prompt
timeout=100
boot=/dev/fd0u1722
map=/mnt/boot/map
disk=/dev/fd0u1722
geometric
install=/mnt/boot/boot.b
image=/mnt/boot/zImage
label=linux
root=/dev/ram
initrd=/mnt/boot/root.img.gz
image=/mnt/boot/zImage
label=noram
root=/dev/fd0u1722
\end{verbatim}
Note the following:
\begin{itemize}
\item Specify {\tt /dev/fd0u1722} as the diskette device.
\item Using this version of LILO and an overformatted diskette, I was unable
to use the {\tt compact} option. This option works for normally
formatted diskettes. The speed penalty for not using this
option is not as great as with LILO 21.4.
\item Using the {\tt prompt} and {\tt timeout} options we cause the
menu to appear immediately and the default option to be booted after
10 seconds.
\end{itemize}
Finally type the following commands to format the diskette, make a
minix file system on it, copy the files and run LILO:
\begin{verbatim}
fdformat /dev/fd0u1722
mkfs.minix /dev/fd0u1722
mount /dev/fd0u1722 /mnt
mkdir /mnt/boot
cp zImage /mnt/boot
cp root.img.gz /mnt/boot
cp lilo-22.3.1/boot-menu.b /mnt/boot/boot.b
lilo-22.3.1/lilo -C lilo-initrd.conf
umount /mnt
\end{verbatim}
Instead of {\tt boot-menu.b} there is still the old-style {\tt
boot-text.b}, giving exactly the same interface as old LILO versions
and {\tt boot-bmp.b} to show a bitmapped screen. You can try all
three of course.
\subsection{GNU GRUB}
\htmladdnormallinkfoot{GNU GRUB}{http://www.gnu.org/software/grub/}
is the boot loader of the GNU project. It was the first boot loader to
use the new BIOS features to boot beyond 1024 cylinders and it had a
menu interface before LILO. Also it could detect more then 64MB of RAM
when other systems could not yet do this.
In an ideal world all operating system kernels would conform to a
certain standard so that any kernel could be loaded by any boot
loader. The {\em Multiboot} standard is such a proposed standard, but
very few operating systems conform to it (primarily the GNU HURD
operating system). GRUB is designed to conform to the Multiboot
standard. Besides Multiboot, it can boot Linux and BSD kernels and
chainload anything else. For BSD kernels the latest features of the
latest version are not supported, but Linux support is up to date.
GRUB is huge compared to both LILO and SYSLINUX. This is mainly
because it can read most file system types by itself. Besides it
serves as its own installer and it has many useful and less useful
commands. You can recompile it with some file systems and other
features configured out to save space.
Just for fun and to show how flexible GNU GRUB really is, we will
install the Multiboot compliant \htmladdnormallinkfoot{GRUB Invaders
game}{http://www.erikyyy.de/invaders/} as well. Skip this if you don't
like it. Unpack the tar archive and just use the {\tt invaders}
binary.
Download the latest version of GRUB (0.92), unpack it and type the
following commands in the source directory:
\begin{verbatim}
./configure
make
\end{verbatim}
Create the file {\tt menu.lst} with the following contents.
\begin{verbatim}
root (fd0)
title Linux with RAM disk
kernel /boot/zImage
initrd /boot/root.img.gz
title Linux without RAM disk
kernel /boot/zImage root=/dev/fd0
title GRUB Invaders
kernel /boot/invaders
\end{verbatim}
Create a diskette and copy the usual stuff to it. The fact that we
create a DOS diskette is arbitrary. We could have performed the same
procedure with almost any other file system that Linux supports. The
files {\tt stage1} and {\tt stage2} are components of grub that are
traditionally installed in the directory {\tt /boot/grub} on the boot
device. The configuration file is {\tt menu.lst} is also copied to
this directory. As opposed to LILO, GRUB reads the configuration file
at boot time.
\begin{verbatim}
fdformat /dev/fd0H1440
mkfs.msdos /dev/fd0
mount -t vfat /dev/fd0 /mnt
mkdir /mnt/boot
mkdir /mnt/boot/grub
cp zImage /mnt/boot
cp root.img.gz /mnt/boot
cp invaders/invaders /mnt/boot
cp grub-0.92/stage1/stage1 /mnt/boot/grub
cp grub-0.92/stage2/stage2 /mnt/boot/grub
cp menu.lst /mnt/boot/grub
umount /dev/fd0
\end{verbatim}
Now we only need to make this diskette bootable. There are three ways
to do it:
\begin{itemize}
\item Boot from an existing GRUB diskette or hard disk and install
GRUB onto the diskette using the GRUB command line.
\item Use the GRUB shell (this is what we will do). This is a program
that runs under Linux and emulates the GRUB command line. You can do
everything but boot a kernel.
\item Use grub-install, a non-interactive tool to perform the task
from shell scripts.
\end{itemize}
Run the GRUB shell:
\begin{verbatim}
grub-0.92/grub/grub
\end{verbatim}
In the GRUB shell type the following commands:
\begin{verbatim}
root (fd0)
setup (fd0)
quit
\end{verbatim}
The {\tt setup} command installs a boot sector program on the given
device. Note that GRUB can install its own boot sector, but it cannot
write, and hence cannot copy, the required files {\tt stage1} and {\tt
stage2} to the target device. This must have been done before by an
operating system.
After you have installed GRUB onto the diskette, you can copy other
kernels to it and edit the {\tt menu.lst} file as much as you like,
without the need to reinstall. Only the file {\tt stage2} may not be
moved.
When you boot the GRUB diskette, you get a menu with three
entries. You can do any of the following:
\begin{itemize}
\item Just select a menu entry and press ENTER.
\item Press the {\tt e} key over a menu entry and edit that
entry. This is useful for the second entry (Linux without
RAM disk) to edit the {\tt root} device for that command.
\item Press the {\tt c} key to go to the command line. The command
line is very powerful.
\begin{itemize}
\item The {\tt cat} command shows the contents of any text
file. Example, the following command shows the {\tt /etc/passwd}
file on {\tt /dev/hda1}:
\begin{verbatim}
cat (hd0,0)/etc/passwd
\end{verbatim}
While you type the command, pressing the TAB key shows you which
files are available, just as in the {\tt bash} shell.
\item The kernel command enables you to select any Linux kernel
lying around on the hard disk and boot from it. Example:
\begin{verbatim}
kernel (hd0,2)/boot/vmlinuz root=/dev/hda3 read-only
boot
\end{verbatim}
This boots a kernel on {\tt /dev/hda3}. This works, even if you hosed
your LILO.
\end{itemize}
\end{itemize}
\subsection{SYSLINUX}
\htmladdnormallinkfoot{SYSLINUX}{http://syslinux.zytor.com} is a
boot loader specially suited for diskettes. It boots only from DOS
formatted disks (could be hard disk partitions as well). It supports
custom help screens to introduce the user into using the diskette.
DOS diskettes have the advantage that they are not considered
defective by most PC users and therefore they less likely thrown away
or reformatted. Like GRUB, SYSLINUX reads its configuration file at
boot time.
In theory SYSLINUX can be installed in a FAT partition on the hard
disk and it can boot Linux kernels from there and it can chain load MSDOS and
similar operating systems. However, SYSLINUX is not the natural choice
for booting Linux from the hard disk.
First obtain the latest version of SYSLINUX. (version 1.75 at the time
of writing). Unpack and run {\tt make}. Note that you need a recent
version of {\tt nasm} to rebuild the boot loader from source and that
the binaries are already included, so you do not really need nasm just
to install this boot loader.
Create a text file {\tt syslinux.cfg} with the following contents:
\begin{verbatim}
prompt 1
timeout 100
say Available options: linux and noram
label linux
kernel zimage
append initrd=root.gz
label noram
kernel zimage
\end{verbatim}
This file specifies the usual options for loading a kernel with an
initial RAM disk and a kernel without an initial RAM disk (where a
root device should be specified on the command line).
Next format a disk and copy all relevant files to it. An overformatted
disk can be used and an MSDOS file system {\em must} be used. The {\tt
syslinux} command installs the boot sector program.
\begin{verbatim}
fdformat /dev/fd0u1722
mkfs.msdos /dev/fd0u1722
cd syslinux-1.75
syslinux /dev/fd0u1722
cd ..
mount -t msdos /dev/fd0u1722 /mnt
cp zImage /mnt/zimage
cp root.img.gz /mnt/root.gz
cp syslinux.cfg /mnt
umount /mnt
\end{verbatim}
You can copy other kernels and initial RAM disk and edit
syslinux.cfg as much as you like without any form of re-installation,
just as you can do with GRUB. The GRUB binary on the example disk was
95k while the SYSLINUX binary is only 7k. That's the difference.
\subsection{LOADLIN}
LOADLIN is a program that can boot Linux from DOS. Since the advent of
Windows ME and Windows XP, it is no longer natural that PC users have
DOS installed on their PCs. Windows versions up to and including
Windows 98 came with a DOS version that was usable for LOADLIN. In my
opinion LOADLIN is primarily of historical interest for this
reason. The primary reasons to use LOADLIN were the following:
\begin{itemize}
\item Getting devices working under Linux after loading their DOS
device drivers. Some PC devices (mostly sound cards) were supposed
to be fully compatible with a standard product (that Linux had a
device driver for), but only after they were initialized by their
DOS device drivers. So you needed to run a DOS device driver first
before Linux would be able to access that device. Therefore it made
sense to boot Linux from DOS.
\item Booting from a Linux CD without the need to create boot
floppies. Bootable CDs are a comparatively recent invention and most
BIOS-es did not support it before 1998. Without the ability to boot
directly from the CD you either needed boot floppies or you could
start Linux on the CD from DOS using LOADLIN.
\item Installing Linux without the need to tamper with the hard
disk. With LOADLIN it is not necessary to install anything into the
MBR in order to get Linux booted from the hard disk. A loadlin entry
could easily be added to the DOS boot menu in CONFIG.SYS.
Linux could even be installed without repartitioning the disk.
For this very purpose the UMSDOS file system was added to the
Linux kernel. It still exists, though it sees very little usage
today. The Linux file systems lived in the LINUX directory on the
DOS partition and it contained special files to translate long file
names into DOS 8+3 file names and to specify attributes such as
users, groups, permissions and device files. Linux could be
installed just by unpacking a bunch of ZIP files, just as one would
install a large DOS application. LOADLIN made booting Linux just as easy
as running a DOS application, hence it fit nicely into the DOS
mindset.
\end{itemize}
Loadlin can be downloaded from
\htmladdnormallinkfoot{Hans Lermen's home page}{http://elserv.ffm.fgan.de/~lermen}.
Download the file {\tt lodln16.tgz} and unpack it. We will use only
the file {\tt loadlin.exe}.
In order to run LOADLIN, we need DOS. Create a bootable DOS diskette with just
the bare system and {\tt COMMAND.COM} on it. For those of you who have either
Windows NT/ME/2000/XP or no
Microsoft OS at all, you can use
\htmladdnormallinkfoot{Freedos}{http://www.freedos.org}. Download the file
{\tt FDB8\_144.DSK} (the rawrite image) and transfer it to a diskette
\begin{verbatim}
dd if=FDB8_144.DSK of=/dev/fd0
\end{verbatim}
Next remove everything but {\tt KERNEL.SYS} and {\tt COMMAND.COM} from
this diskette.
\begin{verbatim}
mount /dev/fd0 /mnt
cd /mnt
rm -rf config.sys *.bat kernel32.sys kernel16.sys readme\
docs fdos games install util
cd
umount /mnt
\end{verbatim}
Now that we have a minimal DOS diskette, copy {\tt LOADLIN.EXE}, the
linux kernel and the initial RAM disk to it:
\begin{verbatim}
mount /dev/fd0 /mnt
cd /home/lennartb/myboot
cp loadlin.exe /mnt
cp zImage /mnt/zimage
cp root.img.gz /mnt/root.gz
umount /mnt
\end{verbatim}
Next try to boot from the diskette. You should boot into DOS. As there
is no {\tt AUTOEXEC.BAT}, you will be prompted for date and
time. Press enter twice. Now you should be at the \verb'A>' prompt.
Type the following line to boot Linux:
\begin{verbatim}
loadlin zimage initrd=root.gz
\end{verbatim}
Of course you can put this line in a batch file or even in {\tt
AUTOEXEC.BAT}.
\subsection{Where Can Boot Loaders Live}
Table \ref{bltab} summarizes where boot loaders can be installed.
\begin{itemize}
\item The Linux kernel boot sector can only be installed on a raw
diskette with no file system, it cannot be installed on a hard disk
and not on a diskette with a file system.
\item GRUB is the only boot loader that can live both on a DOS file
system and on a different type of file system (ext2, minix). LILO
cannot live on a DOS file system, SYSLINUX lives on a DOS file
system only.
\item The Linux kernel boot sector and GRUB do not work on diskettes
with a nonstandard number of sectors per track. LILO and SYSLINUX do
boot from diskettes with 21 sectors per track. The 24-sector formats
work with no boot loader.
\item The overhead is estimated for a boot diskette. It consists of
the file system overhead (indicated by FS) and the second stage boot
program. For LILO we have to add the map file, for GRUB and SYSLINUX
we have to add the configuration file. For LOADLIN we have to add
DOS as well.
\end{itemize}
\begin{table}
{\small
\begin{tabular}{|l|c|c|c|c|c|}
\hline
Boot Loader&DOS FS&Other FS & Overformatted FD & Hard disk&Overhead\\
\hline
Kernel boot sector & No & No & No & No & 0\\
LILO & No & Yes & Yes & Yes & FS + 15k\\
GRUB & Yes & Yes & No & Yes & FS + 96k\\
SYSLINUX & Yes & No & Yes & Yes & FS + 8k\\
LOADLIN & Yes & No & Yes & Yes & FS + DOS + 32k\\
\hline
\end{tabular}}
\caption{\label{bltab}Where boot loaders can be installed}
\end{table}
\section{RAM Disks}
\label{ramsect}
The Linux kernel needs to mount a root file system before it can execute
any processes. For rescue diskettes and CD's it is often quite natural
to load the root file system into a RAM disk at boot time and mount
the root file system on the RAM disk. This has the following advantages:
\begin{itemize}
\item The RAM disk can be stored compressed on the disk, so more
programs can be stored on a diskette.
\item The diskette or CD can be removed from the drive after the RAM
disk is loaded, freeing the drive for other purposes, such as
restoring backups or loading other utilities.
\item Programs load faster from a RAM disk than from a diskette.
\end{itemize}
\subsection{Living without a RAM disk}
For a Linux installation on a hard disk it is quite natural not to use
a RAM disk, but for a Linux installation that is booted from a
diskette (without depending on a root file system elsewhere) there is
only one valid reason not to use a RAM disk: you don't have enough RAM
for the RAM disk. As a rule of thumb, if you have less than 6--8MB of
RAM. you will probably need to boot without a RAM disk, though it can
be achieved with 4MB if you really strip things down.
If you boot from a diskette without a RAM disk, with the root file
system on a diskette, the kernel will always prompt for an extra
diskette, so the kernel and the root file systems can live on separate
diskettes. The root file system cannot be compressed. It is of course
possible to put both on a single diskette. In that case
the boot diskette contains the root file system and in that root file
system the kernel lives, as would be the case with most hard disk
installations. On the boot diskette you need to install a boot loader
that can live inside that file system, which rules out SYSLINUX, but
both GRUB and LILO should do.
Installing Linux on a machine from diskette without a RAM disk is a
tricky process, which involves the following steps:
\begin{itemize}
\item Boot from a diskette without a RAM disk. The diskette may not be
removed from the drive.
\item Partition the hard disk, initialize the swap partition with {\tt
mkswap} and create a file system on the hard disk.
\item Copy the contents of the root file system from the diskette to
the hard disk.
\item Reboot and select the root device on the hard disk (that you had
just copied). This time the diskette drive is free.
\item Copy all your data to the hard disk partition.
\end{itemize}
\subsection{RAM disk devices}
A RAM disk is just another block device, except that the buffers are
never written to a real device, so the data exists only in RAM. The
standard kernel comes with 16 RAM disk devices. Initially no memory is
allocated to a RAM disk, but buffers will be added dynamically as data
is written to it. The maximum size of a RAM disk is 4MB by default and
can be modified at boot time with the {\tt ramdisk\_size} command line
option. In LILO you could add the following command to the
configuration file to double the size:
\begin{verbatim}
append ramdisk_size=8192
\end{verbatim}
Note that this specifies a maximum size, smaller RAM disks take less space.
See the kernel source file {\tt drivers/block/rd.c} how the RAM disk
is initialized. There are three possibilities:
\begin{itemize}
\item A RAM disk is not initialized at boot time by default. On a
system that was booted without a RAM disk, you can put data into the
RAM disk block device and mount it. The following commands create an
ext2 file system and mount that:
\begin{verbatim}
mke2fs /dev/ram2 2880
mount /dev/ram2 /mnt
\end{verbatim}
\item A RAM disk can be loaded at boot time from a device. In this
case the kernel checks if the data is a valid {\tt gzip} data stream
or a valid file system (only a few types allowed). Compressed RAM
disk images are decompressed by the kernel.
\item A RAM disk can be loaded at boot time from the memory area where
the boot loader put the initial RAM disk image. It is loaded in almost the
same way (the read routines pull the data from memory instead of a
device), but the main initialization code will treat the initial RAM
disk differently with regard to mounting.
\end{itemize}
\subsection{Loading a RAM disk at boot time}
The feature to load a RAM disk from a diskette at boot time has
existed almost from the beginning (it existed in 1992). Since then,
some features have been added or modified. It is possible to load a
RAM disk image from the boot diskette at a specific offset and to load
it from a different diskette.
There are two ways to specify the RAM disk parameters.
\begin{itemize}
\item
Use the {\tt rdev} command to set the boot parameters. See section
\ref{lbssect} for details.
\item Set the {\tt ramdisk\_start}, {\tt load\_ramdisk} and {\tt
prompt\_ramdisk} parameters from the kernel command line. This is
how to do it in LILO:
\begin{verbatim}
root=/dev/fd0
append ramdisk_start=0 load_ramdisk=1 prompt_ramdisk=1
\end{verbatim}
\end{itemize}
It is possible to put a file system on the first part of the diskette
(and install LILO there) and to use the rest of the diskette for the
compressed RAM disk image. Use a size argument to the mkfs command to
specify the number of blocks for the file system.
It is also possible to put a (non-compressed) file system on the boot
diskette and load the entire boot diskette as a RAM disk at boot time. Even
the kernel, the LILO second stage boot loader and the map file would
be loaded into the RAM disk. Advantage is that the same diskette can
be used with or without a RAM disk.
\subsection{The initial RAM disk}
The initial RAM disk is loaded by the boot loader. There are two uses
for the initial RAM disk:
\begin{itemize}
\item The primary purpose of the initial RAM disk is mostly of
interest to maintainers of Linux distributions. The initial RAM disk
makes it possible to install a minimal kernel image on the hard disk
without the file system or hard disk device drivers compiled in. At
boot time these drivers are loaded as modules, not from the hard
disk, but from an initial RAM disk. After doing that, the kernel
would switch to the real root device (the hard disk). Without an
initial RAM disk you would need the hard disk drivers (there are
many possible SCSI host adapters) and the file system drivers
compiled into the kernel, so you either have a huge kernel or many
different kernels to choose from.
Such an initial RAM disk needs a file system with the following:
\begin{itemize}
\item A {\tt /dev} directory with some essential devices.
\item Some directories for mount points.
\item A shell with some binaries. The \htmladdnormallinkfoot{{\tt
busybox}}{http://www.busybox.net} program would be all that is
needed.
\item The required kernel modules.
\item A startup script that loads the required kernel modules.
\end{itemize}
\item The initial RAM disk can also be used for rescue, install or
special-purpose diskettes, much as you would use the other type of
RAM disk. You simply do not use the feature that you can switch back
to the real root device. This is the type of diskette described in
my ``Getting Linux into Small Machines'' article.
\end{itemize}
There is no reason why you should not be able to install an
initial RAM disk from a second diskette, but most boot loaders do not
support this. This is a limitation of the boot loader, not of
Linux. Using the {\tt pause} command, GRUB can do this, as the
following part of a configuration file shows:
\begin{verbatim}
title Linux with RAM disk on separate diskette
kernel (fd0)/boot/vmlinuz
pause Insert the second diskette.
initrd (fd0)/boot/initrd.gz
\end{verbatim}
\section{Making Diskette Images without Diskettes}
So far we have been preparing boot diskettes by writing to real
diskettes. This sounds like the most logical way to do it, but there
can be reasons why we want to prepare an {\em image file} of a
diskette without using real diskettes. Several reasons could be:
\begin{itemize}
\item Create a diskette image for a diskette drive you do not have.
\item Automate boot image creation (several diskette images for a
distribution).
\item Create a diskette image for bootable CD-ROM
\item Create a diskette image for a PC emulator
\end{itemize}
Basically we could create a diskette image as follows:
\begin{itemize}
\item Create an all zero image file using {\tt dd}.
\item Create a file system onto the image file.
\item Mount the image file using the loop option and copy all files to
it.
\item Install the boot loader onto the image file.
\end{itemize}
The last part is the trickiest especially for LILO. It's fairly
trivial for SYSLINUX and using the {\tt device} command it can be done
with GRUB. There is also another trick for SYSLINUX and GRUB (it does
not work with LILO):
\begin{itemize}
\item Start with an image file with just the boot loader installed and
an empty file system and copy that image file each time you create
another image. This image file may be extracted from a real diskette
just once.
\item Mount the image file using the loop option and copy all files to
it.
\end{itemize}
On
\htmladdnormallinkfoot{Timo's Rescue CD Page}{http://rescuecd.sourceforge.net/}
there is a good explanation of how to create 2.88MB diskette images
for a bootable CD-ROM, using all boot loaders. I could not explain it
better. Of course these recipes apply also to other types of disk images.
\section{Hard Disk Installation}
For a booting Linux on the hard disk there are basically the following
choices:
\begin{itemize}
\item Use an OS neutral MBR boot sector program (such as the one from
DOS) and let it boot a Linux boot loader in a primary
partition. Some such boot sector programs let the user select one of
several operating systems at boot time.
\item Boot a Linux boot loader (Linux or GRUB) directly from the MBR.
\item Chainload a Linux boot loader through the boot loader of another
OS (the NT Boot Manager). One could even chainload Linux from
another instance of a Linux boot loader. There could be an instance
of GRUB that is booted from the MBR and GRUB could chainload several
instances of LILO, each on its own partition.
\item Boot Linux from a diskette. The diskette may contain just the
boot sector (it loads the second stage of the boot loader from the
hard disk), the entire boot loader or even the boot loader plus the
kernel.
\item Boot Linux through DOS using LOADLIN.
\end{itemize}
Which option you use, depends on several factors:
\begin{itemize}
\item What other operating systems run on the same
machine? Chainloading Linux from the Windows NT boot loader would
make no sense on a machine without NT.
\item What partitions are used on the disk? Is a RAID used? As far as
I know, LILO is the only Linux boot loader that boots from a RAID.
\item What is the relative importance of each operating system?
\item Perceived reliability of each boot loader.
\item Personal preferences.
\end{itemize}
One factor that may complicate boot loader installation is that each
operating system has its own naming and numbering scheme for hard disks and
partitions and so does GRUB. The BIOS assigns a number to each
disk. The first disk is 0x80, the second disk is 0x81, etc. When you
install LILO, things go without problems most of the time, but there
are situations in which it may be necessary to specify the BIOS code
of a drive.
The naming convention of hard disks in GRUB can be confusing:
\begin{itemize}
\item The whole hard disk (if you want to install GRUB in the MBR) is
called {\tt (hd0)}, {\tt (hd1)} etc.
\item Partitions are counted from zero, so partition 1 on the first disk
is called {\tt (hd0,0)}. Extended partitions are counted from 4,
like they are counted from 5 in Linux.
\item Disks are also counted from zero in the order that BIOS assigns
them. On most systems {\tt /dev/hda} corresponds to {\tt (hd0)}
(except on systems with only SCSI disks), but
what Linux disk will correspond to {\tt (hd1)}, depends on what you
connect to the EIDE ports. If the primary slave is a hard disk, this
disk will be {\tt /dev/hdb} in Linux and {\tt (hd1)} in GRUB. If the
primary slave is a CD-ROM and the secondary master is a hard disk,
the hard disk will be {\tt /dev/hdc} in Linux and {\tt (hd1)} in
GRUB.
\end{itemize}
Table \ref{partexampl1} shows the disk and partition numbers on a
system with two hard disks, connected as master and slave to the
primary IDE controller. The first disk contains two primary partitions
(DOS and Linux) and two logical partitions (DOS and Linux swap). The
second disk contains one primary DOS partition.
\begin{table}
{\small
\begin{tabular}{|l|c|c|c|c|}
\hline
Disk & Linux & BIOS code (LILO) & GRUB disk name&DOS drive\\
\hline
IDE Primary Master & /dev/hda & 0x80 & (hd0) & -\\
Primary DOS & /dev/hda1& - & (hd0,0) & C:\\
Primary Linux & /dev/hda2& - & (hd0,1) & -\\
Extended & /dev/hda3& - & (hd0,2) & -\\
Logical DOS & /dev/hda5& - & (hd0,4) & E:\\
Linux swap & /dev/hda6& - & (hd0,5) & -\\
\hline
IDE Primary Slave & /dev/hdb & 0x81 & (hd1) & -\\
Primary DOS & /dev/hdb1& - & (hd1,0) & D:\\
\hline
\end{tabular}}
\caption{\label{partexampl1}Disk and partition names (first example)}
\end{table}
Table \ref{partexampl2} shows the disk and partition numbers on a
system with three hard disks, one on the primary IDE controller and one
on the secondary IDE controller and one SCSI.
\begin{table}
{\small
\begin{tabular}{|l|c|c|c|c|}
\hline
Disk & Linux & BIOS code (LILO) & GRUB disk name&DOS drive\\
\hline
IDE Primary Master & /dev/hda & 0x80 & (hd0) & -\\
Primary DOS & /dev/hda1& - & (hd0,0) & C:\\
Primary Linux & /dev/hda2& - & (hd0,1) & -\\
\hline
IDE Secondary Master& /dev/hdc & 0x81 & (hd1) & -\\
Primary DOS & /dev/hdc1& - & (hd1,0) & D:\\
\hline
First SCSI & /dev/sda & 0x82 & (hd2) & -\\
Primary DOS & /dev/sda1& - & (hd2,0) & E:\\
\hline
\end{tabular}}
\caption{\label{partexampl2}Disk and partition names (second example)}
\end{table}
If you use FreeBSD as well, things get even more complicated. BSD
divides a normal disk partition into several sub-partitions. Linux can
only access them if support for BSD disk labels is compiled into the
kernel. Needless to say that Linux, BSD and GRUB all have different
names for these sub-partitions.
In GRUB you have to use GRUB partition names for files accessed by GRUB,
but Linux partition names for arguments on the kernel command
line. The following is a very typical {\tt kernel} command line in GRUB:
\begin{verbatim}
kernel (hd0,1)/vmlinuz root=/dev/hda2
\end{verbatim}
Finally on some (mostly older) systems, disk geometry may cause
problems. As long as everything uses LBA, no problems are expected. If
your BIOS does not support LBA, boot loaders using the BIOS have no
way to reach beyond 1024 cylinders. {\tt nuni} {\em may} be a solution
in such a case, but most of the time you will need to create a small
partition near the start of the disk and boot the kernel from there.
\section{CD-ROM Installation}
Most Linux distributions come on a CD-ROM and there are various rescue
disk images available that you can burn to a CD-ROM. In the near
future, most PC's won't even have a floppy drive, so the CD-ROM is the
only choice.
Most modern BIOS-es support the El Torito standard to boot from a
CD-ROM. There are two different ways to boot Linux from a CD-ROM:
\begin{itemize}
\item Emulated diskette images are used in most cases. One file on the
CD-ROM is a diskette image file. Several diskette sizes are
supported, but 1.44MB is the most common. The BIOS loads the first
512-bit block into memory as if it were the boot sector of a
diskette. Next it redirects all INT 0x13 calls for the first floppy
drive to that file on the CD-ROM. Generally a program that boots
from a diskette and that only uses BIOS calls to access the
diskette, will work without change when booted from an emulated
diskette image. The boot loaders LILO, GRUB and SYSLINUX should just
work when booted this way.
As an initial RAM disk on a diskette is loaded by the boot loader
and the boot loader uses BIOS, this will also work from an emulated
diskette image on a CD-ROM. This does not apply to a RAM disk that
is loaded by the kernel (the non-initrd type), because the kernel
does not use BIOS.
The LILO, GRUB and SYSLINUX boot diskettes described in this
document should all work from CD-ROM, provided you create them on
normal 1.44MB diskettes and not on overformatted 1.72MB diskettes.
It is not possible to use overformatted diskettes on a bootable
CD-ROM (21 sectors per track 82 tracks), but it is possible to use
2.88MB diskette images. Real 2.88MB diskettes and their drives are
very rare, but as diskette images on bootable CD-ROM's this format has
gained popularity.
\item It is possible to boot a program without diskette
emulation. ISOLINUX is a program that can be loaded in such a
way. It comes with the SYSLINUX package and can find files
everywhere on an ISO9660 format CD-ROM. It has a configuration file
that is almost identical to that of SYSLINUX. Using this method
there is no practical limit to the number of different kernels or to
the size of an initial RAM disk.
\end{itemize}
As soon as the Linux kernel is booted, it runs with the RAM disk as
the root file system. From there it is easy to mount the CD-ROM. The
CD-ROM can be filled with utilities, packages to install or whatever.
In theory it is even possible to mount the CD-ROM as the root file
system, but then it becomes impossible to swap disks while the system
is running.
Both types of bootable CD-ROM can be created with {\tt mkisofs}. The
resulting ISO image can then be burnt to a CD-ROM using {\tt cdrecord}
or another CD writing program, even under Windows. I've tried both
types using CD-RW disks.
Let's start with an emulated diskette. In the example I used the GRUB
diskette, as it was already formatted at 1.44M (the image from my home
page uses LILO and it should work fine too). First create a directory
tree for the ISO image.
\begin{verbatim}
mkdir iso
mkdir iso/boot
mkdir iso/data
\end{verbatim}
Next copy some files to the {\tt data} subdirectory and copy the
diskette to the {\tt boot} directory.
\begin{verbatim}
cp somefile iso/boot/data
dd if=/dev/fd0 of=iso/boot/boot.img
\end{verbatim}
Now create the ISO image.
\begin{verbatim}
mkisofs -o emuboot.iso -b boot/boot.img \
-c boot/boot.catalog -r iso
\end{verbatim}
Finally burn it to a CD-ROM. Use the appropriate device ID.
\begin{verbatim}
cdrecord dev=0,1,0 -eject -pad -data emuboot.iso
\end{verbatim}
This CD is bootable and from the booted Linux you should be able to
mount the CD-ROM to access the data files. As my CD-ROM is on {\tt
/dev/hdd} and the RAM disk image does not include this device, I had to
{\tt mknod} it first:
\begin{verbatim}
mknod /dev/hdd b 22 64
mount -r -t iso9660 /dev/hdd /mnt
\end{verbatim}
Next create a bootable Linux CD-ROM with ISOLINUX. First create the
file {\tt isolinux.cfg} with the following contents:
\begin{verbatim}
prompt 1
timeout 100
say Available options: linux and noram
label linux
kernel zimage
append initrd=root.gz
label noram
kernel zimage
\end{verbatim}
In fact this file is identical to {\tt syslinux.cfg} on the SYSLINUX diskette.
Remove the old boot directory from the {\tt iso} directory tree and
replace it with the {\tt isolinux} directory containing the kernel,
the RAM disk image and the boot files:
\begin{verbatim}
rm -rf iso/boot
mkdir iso/isolinux
cp zImage iso/isolinux/zimage
cp root.img.gz iso/isolinux/root.gz
cp syslinux-1.75/isolinux.bin iso/isolinux
cp isolinux.cfg iso/isolinux
\end{verbatim}
Next create the ISO image and burn it to CD as usual:
\begin{verbatim}
mkisofs -o isoboot.iso -b isolinux/isolinux.bin \
-c isolinux/boot.catalog \
-no-emul-boot -boot-load-size 4 \
-boot-info-table -r iso
\end{verbatim}
Older BIOS-es cannot boot from CD-ROM. There are two solutions:
\begin{itemize}
\item Any 1.44MB El Torito boot image can do double duty as a diskette
image. The DOS program {\tt RAWRITE.EXE} or Linux {\tt dd} can
transfer such a file to a diskette and we can boot from the diskette
with no problems.
\item If the machine runs DOS and DOS can read the CD-ROM, LOADLIN
should be able to boot Linux from it.
\end{itemize}
Most Linux distributions provide both RAWRITE and LOADLIN on the
CD-ROM and some even provide a real boot diskette.
\section{Conclusions}
There exist many boot loaders for Linux and no single boot loader is
the best in all circumstances.
I would recommend to use the following boot loaders:
\begin{itemize}
\item For general purpose hard disk installations, use either LILO or
GRUB.
\begin{itemize}
\item GRUB has more features for recovery, but these can be
dangerous as well. Of course you can protect the command line with
a password.
\item LILO comes standard with all distributions. If you just want
to install Linux (maybe together with DOS/Windows), use LILO. On the
other hand, if you install different Linux distributions, different
operating systems and do frequent kernel compilation, GRUB is
absolutely for you.
\item LILO can boot from RAID systems while GRUB can't.
\end{itemize}
\item SYSLINUX is ideal for diskette installation. It is compact, once
it is installed, you can just copy files to DOS disks and it supports
overformatted diskettes.
\item For CD-ROM installations we could use diskette images with
SYSLINUX, but ISOLINUX may be worth a try.
\item A diskette with just GRUB installed and nothing else, can be a great
rescue tool for machines that went unbootable. Using the GRUB
diskette, you can mix and match kernels and initial RAM disk images
from other diskettes and you can boot any kernel on any hard disk partition.
\end{itemize}
\end{document}