docs/grub.texi

\input texinfo  @c -*-texinfo-*-
@setfilename grub.info
@include version.texi

@c Unify all our little indices for now.
@syncodeindex fn cp
@syncodeindex vr cp
@syncodeindex ky cp
@syncodeindex pg cp
@syncodeindex tp cp

@dircategory Kernel
@direntry
* GRUB: (grub).                 The GRand Unified Bootloader.
@end direntry

@ifinfo
Copyright @copyright{} 1996 Erich Boleyn
Copyright @copyright{} 1999 Free Software Foundation, Inc.

Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.

@ignore
Permission is granted to process this file through TeX and print the
results, provided the printed document carries a copying permission
notice identical to this one except for the removal of this paragraph
(this paragraph not being relevant to the printed manual).

@end ignore

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that
the entire resulting derived work is distributed under the terms of a
permission notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions.
@end ifinfo

@c @setchapternewpage odd
@settitle GRUB Manual
@titlepage
@finalout
@title The GRUB Manual
@author Gordon Matzigkeit
@author OKUJI Yoshinori
@page

@vskip 0pt plus 1filll
Copyright @copyright{} 1996 Erich Boleyn
Copyright @copyright{} 1999 Free Software Foundation, Inc.

Permission is granted to make and distribute verbatim copies of
this manual provided the copyright notice and this permission notice
are preserved on all copies.

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that
the entire resulting derived work is distributed under the terms of a
permission notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions.
@end titlepage

@c The Top node should not appear in TeX.
@ifnottex

@node Top
@top GRUB

This file documents GNU GRUB, the Grand Unified Bootloader.  This
edition documents version @value{VERSION}.
@end ifnottex


@menu
* Introduction::                Capturing the spirit of GRUB.
* Installing::                  How to install GRUB on your computer.
* Using::                       Booting your operating system.
* Filesystems::                 Filesystem syntax and semantics.
* Troubleshooting::             Error messages produced by GRUB.
* The grub shell::              The @command{grub} command.
* Hacking::                     Implementation details.
* Index::                       Index.

@detailmenu
 --- The Detailed Node Listing ---

Introduction

* History::                     From maggot to house fly.
* Features::                    How GRUB is different.
* Role of a bootloader::        Judging a system by its bootloader.

How to install GRUB on your computer

* Boot floppy::                 Creating a GRUB boot floppy.
* Automated install::           Installation via @code{install}.

Booting your operating system

* Command line::                The flexible command-line interface.
* Menu::                        The simple menu interface.
* Menu entry editor::           Editing a menu entry.
* Commands::                    The list of available commands.

Filesystem syntax and semantics

* Device syntax::               How to specify devices.
* Filename syntax::             How to specify files.

Error messages reported by GRUB

* Stage1 errors::               Errors reported by the Stage 1.
* Stage1.5 errors::             Errors reported by the Stage 1.5.
* Stage2 errors::               Errors reported by the Stage 2.

The @command{grub} command

* Basic usage::                 How to use the grub shell.
* Installation under UNIX::     How to install GRUB via @command{grub}.

Implementation details

* Memory map::                  The memory map of the various
                                components.
* Embedded data::               Embedded variables in GRUB.
* Memory detection::            How to detect all installed @sc{ram}.
* Low-level disk I/O::          INT 13H disk I/O interrupts.
* MBR::                         The structure of Master Boot Record.
* Partition table::             The format of partition table.
* Filesystem interface::        The generic interface for the fs code.

@end detailmenu
@end menu


@node Introduction
@chapter Introduction

Briefly, a @dfn{bootloader} is the first software program that runs when
a computer starts.  It is responsible for loading and transferring
control to the operating system @dfn{kernel} software (such as the Linux
or GNU Hurd kernel).  The kernel, in turn, initializes the rest of the
operating system (usually GNU).

@menu
* History::                     From maggot to house fly.
* Features::                    How GRUB is different.
* Role of a bootloader::        Judging a system by its bootloader.
@end menu


@node History
@section History of GRUB

GRUB originated in 1995 when Erich Boleyn was trying to boot the GNU
Hurd with the University of Utah's Mach 4 microkernel (now known as GNU
Mach).  Erich and Brian Ford designed the Multiboot Standard
(@pxref{Top, Multiboot Standard, Motivation, multiboot, The Multiboot
Standard}), because they were determined not to add to the large number
of mutually-incompatible PC boot methods.

Erich then began modifying the FreeBSD bootloader so that it would
understand Multiboot.  He soon realized that it would be a lot easier
to write his own bootloader from scratch than to keep working on the
FreeBSD bootloader, and so GRUB was born.

Erich added many features to GRUB, but other priorities prevented him
from keeping up with the demands of its quickly-expanding user base.  In
1999, Gordon Matzigkeit and OKUJI Yoshinori adopted GRUB as an official
GNU package, and opened its development by making the latest sources
available via anonymous CVS.@footnote{The repository is
@code{:pserver:anoncvs@@anoncvs.gnu.org:/gd/gnu/anoncvsroot}, module
@code{grub}.  Just hit return when prompted for a password.}


@node Features
@section GRUB features

The primary requirement for GRUB is that it be compliant with the
@dfn{Multiboot Standard}, which is described in @ref{Top, Multiboot
Standard, Motivation, multiboot, The Multiboot Standard}.

The other goals, listed in approximate order of importance, are:

@itemize @bullet{}
@item
Basic functions must be straightforward for end-users.

@item
Rich functionality to support kernel experts and designers.

@item
Backward compatibility for booting FreeBSD, NetBSD, OpenBSD, and
Linux. Proprietary kernels (such as DOS, Windows NT, and OS/2) are
supported via a chain-loading function.
@end itemize

Except for specific compatibility modes (chain-loading and the Linux
@dfn{piggyback} format), all kernels will be started in much the same
state as in the Multiboot Standard.  Only kernels loaded at 1 megabyte
or above are presently supported. Any attempt to load below that
boundary will simply result in immediate failure and an error message
reporting the problem.

In addition to the requirements above, GRUB has the following features
(note that the Multiboot Standard doesn't require all the features that
GRUB supports):

@table @asis
@item Multiple Executable Formats
Supports many of the @dfn{a.out} variants plus @dfn{ELF}.  Symbol
tables are also loaded.

@item Supports Non-Multiboot Kernels
Supports many of the various free 32-bit kernels that lack Multiboot
compliance (primarily FreeBSD, NetBSD, OpenBSD, and
Linux).  Chain-loading of other bootloaders is also supported.

@item Loads Multiples Modules
GRUB fully supports the Multiboot feature of loading multiple modules.

@item Configuration File
Supports a human-readable text configuration file with preset boot
commands.  The list of commands (@pxref{Commands}) are a superset of
those supported on the command line.  An example command file is provided
in @file{docs/menu.lst} in the source tree.

@item Menu Interface
A menu interface listing the preset boot commands, with a programmable
timeout, is available. There is no fixed limit on the number of boot
entries, and the current implementation has space for several hundred.

@item Flexible Command Line Interface
A fairly flexible command line interface, accessible from the menu,
is available to edit any preset commands, or write a new boot command
set from scratch.  If no command file is present, GRUB drops to
the command line.

The list of commands (@pxref{Commands}) are a subset of those supported
for command files. Editing commands closely resemble the Bash command
line (@pxref{Command Line Editing, Bash, Command Line Editing, features,
Bash Features}), with @key{TAB}-completion of commands, devices,
partitions, and files in a directory depending on context.

@item Multiple Filesystem Types
Supports multiple filesystem types transparently, plus a useful explicit
blocklist notation. The currently supported filesystem types are
@dfn{BSD FFS}, @dfn{DOS FAT16 and FAT32}, @dfn{Minix fs}, and
@dfn{Linux ext2fs}. @xref{Filesystems}, for more information.

@item Decompression Support
Can decompress files which were compressed by
@command{gzip}. This function is both automatic and transparent to the
user (i.e. all functions operate upon the uncompressed contents
of the specified files).  This greatly
reduces file size and loading time, a particularly major
benefit for floppies.@footnote{There are a few pathological cases
where loading a very badly organized ELF kernel might take
longer, but in practice this never happens.}

It is conceivable that some kernel modules should be loaded in a
compressed state, so a different module-loading command can be specified
to avoid uncompressing the modules.

@item Access Data on Any Installed Device
Supports reading data from any or all floppy or hard disk(s) recognized
by the BIOS, independent of the setting of the root partition.

@item Independent of Drive Geometry Translation
Unlike many other bootloaders, GRUB makes the particular drive
translation irrelevant.  A drive installed and running
with one translation may be converted to another translation without any
adverse effects or changes in GRUB's configuration.

@item Detects All Installed @sc{ram}
GRUB can generally find all the installed @sc{ram} on a PC-compatible
machine. It uses an advanced BIOS query technique for finding all
memory regions (@pxref{Memory detection}). As described on the Multiboot
Standard (@pxref{Top, Multiboot Standard, Motivation, multiboot, The
Multiboot Standard}), not all kernels make use of this information, but
GRUB provides it for those who do.

@item Supports Logical Block Address Mode
In traditional disk calls (called @dfn{CHS mode}), there is a geometry
translation problem, that is, the BIOS cannot access over 1024
cylinders, so the accessible space is limited to at least 508 MB and to
at most 8GB. GRUB can't universally solve this problem, as there is no
standard interface used in all machines. However, some newer machines
have the a new interface, Logical Block Address (@dfn{LBA}) mode.  GRUB
automatically detects if LBA mode is available and uses it if
available. In LBA mode, GRUB can access the entire disk.
@end table

Future directions might include an internal programming language for
supporting richer sets of boot options with control statements (which
would make GRUB its own kind of kernel).  Support for non-PC hardware
architectures is also planned.@footnote{There is already a port to the
NEC PC-98xx series. See
@url{http://www.kuis.kyoto-u.ac.jp/~kmc/proj/linux98/arch/i386/boot/grub98/},
for more information.}


@node Role of a bootloader
@section The role of a bootloader

The following is a quotation from Gordon Matzigkeit, a GRUB fanatic:

@quotation
Some people like to acknowledge both the operating system and kernel when
they talk about their computers, so they might say they use
``GNU/Linux'' or ``GNU/Hurd''.  Other people seem to think that the
kernel is the most important part of the system, so they like to call
their GNU operating systems ``Linux systems.''

I, personally, believe that this is a grave injustice, because the
@emph{bootloader} is the most important software of all.  I used to
refer to the above systems as either ``LILO''@footnote{The LInux LOader,
a bootloader that everybody uses, but nobody likes.} or ``GRUB''
systems.

Unfortunately, nobody ever understood what I was talking about; now I
just use the word ``GNU'' as a pseudonym for GRUB.

So, if you ever hear people talking about their alleged ``GNU'' systems,
remember that they are actually paying homage to the best bootloader
around@dots{} GRUB!
@end quotation

We, the GRUB maintainers, do not (usually) encourage Gordon's level of
fanaticism, but it helps to remember that bootloaders deserve
recognition.  We hope that you enjoy using GNU GRUB as much as we did
writing it.


@node Installing
@chapter How to install GRUB on your computer

Due to the nature of a @dfn{bootloader}, you need to install GRUB on
bootable media, such as a floppy disk. The installation can be performed
by @code{dd} or @code{rawrite} for a boot floppy, or the @code{install}
command at the GRUB command line (@pxref{Using}).

@menu
* Boot floppy::                 Creating a GRUB boot floppy.
* Automated install::           Installation via @code{install}.
@end menu


@node Boot floppy
@section Creating a GRUB boot floppy

@quotation
@strong{Caution:} This procedure will destroy any data currently stored
on the floppy.
@end quotation

If you install GRUB using this method, it will only have access to the
command line interface, since there is no filesystem in which to find a
configuration file. If you want to use the menu interface, see
@ref{Automated install}.

Under an UNIX-like operating system, such as GNU, use @code{dd} as
follows, where @file{/lib/grub/i386-pc} is the GRUB install directory
and @file{/dev/fd0} is the floppy device:

@example
$ cd /lib/grub/i386-pc
$ dd if=stage1 of=/dev/fd0 bs=512 count=1
1+0 records in
1+0 records out
$ dd if=stage2 of=/dev/fd0 bs=512 seek=1
67+1 records in
67+1 records out
$
@end example

Under DOS-based systems, such as Windows, use @code{copy} and
@code{rawrite}:

@example
copy /b stage1 + stage2 grub.raw
rawrite grub.raw a:
@end example

@code{rawrite} is available as a part of the installation tools that
come with many GNU and GNU/Linux distributions.


@node Automated install
@section Installation via the @code{install} command

@quotation
@strong{Caution:} Installing GRUB's stage1 in this manner will erase the
normal boot-sector used by an OS.  GRUB can boot GNU Mach, Linux,
FreeBSD, NetBSD, and OpenBSD directly, so this may be
desired.  Generally, it is a good idea to back up the first sector of the
partition on which you are installing GRUB's stage1.  This isn't as
important if you are installing GRUB on the first sector of a hard disk,
since it's easy to reinitialize it (by running @code{FDISK /MBR} from DOS).
@end quotation

GRUB has a command called @code{install} which is described in
@xref{Using}. The purpose of this section is to give examples and
describe how to use the command in different situations.

First, make a GRUB boot floppy (@pxref{Boot floppy}).  This is simply a
way to get the process started easily; any bootable copy of the same
version GRUB will work fine.

Then, make a @file{/boot/grub} directory (@file{\boot\grub} under DOS)
in the @dfn{install partition}.  Place the GRUB @file{stage2} file, any
optional @file{stage1.5} files, and the configuration file
(@file{menu.lst}) in that directory.

Now figure out how to use the @code{install} command appropriately, and
you're done!

Examples of how to use the @code{install} command:

@itemize @bullet
@item
@strong{Make a hard disk bootable with GRUB's stage2 on PC partition
number 2:} Make a directory in the partition called @file{/boot/grub},
place the @file{stage2} (and if desired, the @file{menu.lst}
configuration file), then run the following command at GRUB's command
line (after booting from the GRUB floppy):

@example
install (fd0)+1 (hd0) (hd0,1)/boot/grub/stage2 p
@end example

This tells GRUB to grab the first sector of the floppy and use it as the
stage1, create a blocklist using the file @file{/boot/grub/stage2} on
the first hard disk (partition number 2), merge them together, set the
load address correctly for a stage2 (0x8000), save the @dfn{install
partition} in the first sector of the stage2 (the @samp{p} at the end),
then write the result to the first sector of the hard disk.

@item
@strong{Same as above, but place the stage1 on the floppy, then have
it start the stage2 on the hard disk:} The difference here is you're
telling GRUB's stage1 to read from the first hard disk no matter where
the stage1 was loaded from:

@example
install (fd0)+1 d (fd0) (hd0,1)/boot/grub/stage2 p
@end example

By default, GRUB loads the stage2 from the disk where the stage1 was
installed.  The @samp{d} option near the beginning of this command line
forces GRUB to load the stage2 from the disk where it was installed.
Also, the @dfn{destination device} is changed to place the finished
stage1 on the floppy disk.

@item
@strong{Same as above, but place the stage1.5 in the first cylinder of
the hard disk, and load the stage2 via the stage1.5:} Each of Stage 1.5s
supports only one filesystem, so choose a stage1.5 that supports the
filesystem where stage2 is located. Here it is assumed that the
filesystem is ext2fs.

First, copy @file{e2fs_stage1_5} to the first sector after the MBR
(@pxref{MBR}):

@example
dd if=stage2/e2fs_stage1_5 of=/dev/hda bs=512 seek=1
@end example

Second, specify the stage1.5 argument in blocklist format:

@example
install (fd0)+1 (hd0) (hd0)1+1 p (hd0,1)/boot/grub/stage2
@end example

@item
@strong{Installing from an @emph{install directory} to the second hard
disk:} Here we're loading the stage1 from a file on the first hard disk,
installing stage2 from the first BSD @samp{a} partition on the second
hard disk, and setting the stage2's @dfn{configuration file} to
@file{(hd1,a)/grubdir/configfile}:

@example
install (hd0,1)/boot/grub/stage1 (hd1) (hd1,a)/boot/grub/stage2 p /grubdir/configfile
@end example
@end itemize

You can automate these steps by using a GRUB floppy with a filesystem
and a configuration file which contains entries such as:

@example
# Start of entry:
title GNU/Linux installation

# Installation command:
install (fd0)+1 (hd0) (hd0,1)/boot/grub/stage2 p

# Actually boot here:
root (hd0,1)
kernel /zImage root=/dev/hda2
@end example

@dots{} then have the install script continue from there after boot of
the OS.


@node Using
@chapter Booting your operating system

GRUB has both a simple menu interface for choosing preset entries from a
configuration file, and a highly flexible command line for performing
any desired combination of boot commands.

GRUB looks for its configuration file as soon as it is loaded.  If one
is found, then the full menu interface is activated using whatever
entries were found in the file.  If you choose the `command line' menu
option, or if the configuration file was not found, then GRUB drops to
the command line interface.

@menu
* Command line::                The flexible command line interface.
* Menu::                        The simple menu interface.
* Menu entry editor::           Editing a menu entry.
* Commands::                    The list of available commands.
@end menu


@node Command line
@section The flexible command line interface

The command line interface provides a prompt and after it an editable
text area much like a command line in Unix or DOS. Each command is
immediately executed after it is entered @footnote{However, this
behavior will be changed in the future version, in a user-invisible
way.}. The commands are a subset of those available in the configuration
file, used with exactly the same syntax.

@c FIXME: The list of available keys should be listed in @table, and should be
@c explained exactly. Current explanation is obscure.
Cursor movement and editing of the text on the line can be done via a
subset of the functions available in the BASH shell (@kbd{C-f} forward,
@kbd{C-b} backward, @kbd{C-a} beginning of line, @kbd{C-e} end of line,
@kbd{C-k} delete to end, @kbd{C-u} delete to beginning; the PC left and
right arrow keys, @key{HOME}, @key{DEL}, and @key{END} work as well).

When typing commands interactively, if the cursor is within or before
the first word in the command-line, pressing the @key{TAB} key will
display a listing of the available commands, and if the cursor is after
the first word, the @key{TAB} will provide a completion listing
of disks, partitions, and filenames depending on the context.


@node Menu
@section The simple menu interface

The menu interface is quite easy to use.  Its commands are both
reasonably intuitive and described on screen.

Basically, the menu interface provides a list of @dfn{boot entries} to
the user to choose from.  Use the arrow keys to select the entry of
choice, then press @key{RET} to run it.  An optional timeout is
available to boot the default entry (the first one if not set), which is
aborted by pressing any key.

Commands are available to enter a bare command line by pressing @key{c}
(which operates exactly like the non-config-file version of GRUB, but
allows one to return to the menu if desired by pressing @key{ESC}) or to
edit any of the @dfn{boot entries} by pressing @key{e}.


@node Menu entry editor
@section Editing a menu entry

The menu entry editor looks much like the main menu interface, but the
lines in the menu are individual commands in the selected entry instead
of entry names.

If an @key{ESC} is pressed in the editor, it aborts all the changes made
to the configuration entry and returns to the main menu interface.

When a particular line is selected, the editor places the user at a
special version of the GRUB command line to edit that line.  When the
user hits @key{RET}, GRUB replaces the line in question in the boot
entry with the changes (unless it was aborted via @key{ESC},
in which case the changes are thrown away).


@node Commands
@section The list of available commands

In this section, we list the available commands, both in the
configuration file and in the command line.

The configuration file should follow these rules:

@enumerate
@item
The configuration-file-specific commands have to be used before any
others.

@item
A multiboot kernel must be loaded before any modules.

@item
A kernel must be loaded before either the configuration file entry ends,
or any @samp{boot} command is issued.
@end enumerate

The semantics are as follows:

@itemize @bullet
@item
The files @emph{must} be in plain-text format.

@item
@samp{#} at the beginning of a line in a configuration file means it is
only a comment.

@item
Options are separated by spaces.

@item
All numbers can be either decimal or hexadecimal. A hexadecimal number
must be preceded by @samp{0x}, and is case-insensitive.

@item
Extra options or text at the end of the line is ignored unless otherwise
specified.

@item
Bad commands are added to the current entry, except before entries
start, where they are ignored.
@end itemize

Commands usable in configuration files only.

@table @code
@item default @var{num}
Set the default entry to entry number @var{num} (if not specified, it is
0, the first entry).

@item fallback @var{num}
Go into unattended boot mode: if the default boot entry has any errors,
instead of waiting for the user to do anything, it immediately starts
over using the @var{num} entry (same numbering as the @code{default}
command). This obviously won't help if the machine was rebooted by a
kernel that GRUB loaded.

@item password @var{passwd} @var{new-config-file}
Disable all interactive editing control (menu entry editor and
command line). If the password @var{passwd} is entered, it loads the
@var{new-config-file} as a new config file and restarts the GRUB Stage
2.

@item timeout @var{sec}
Set a timeout, in @var{sec} seconds, before automatically booting the
default entry (normally the first entry defined).

@item title @var{name}@dots{}
Start a new boot entry, and set its name to the contents of the rest of
the line, starting with the first non-space character.
@end table

Commands usable both interactively and in configuration files:

@table @code
@item boot
This boots the OS/chain-loader which has been loaded. Only necessary if
running the fully interactive command line (it is implicit at the end of
a config-file entry).

@item color @var{normal} [@var{highlight}]
Change the menu colors. The color @var{normal} is used for most
lines in the menu, and the color @var{highlight} is used to highlight the
line where the cursor points. If you omit @var{highlight}, then the
inverted color of @var{normal} is used for the highlighted line. You
must specify an integer for a color value, where bits 0-3 represent
the foreground color, bits 4-6 represent the background color, and
bit 7 indicates that the foreground blinks.

These are the possible values and the meanings:

@table @asis
@item 0
black

@item 1
blue

@item 2
green

@item 3
cyan

@item 4
red

@item 5
magenta

@item 6
brown

@item 7
light gray

@item
@strong{These below can be specified only for the foreground.}

@item 8
dark gray

@item 9
light blue

@item A
light green

@item B
light cyan

@item C
light red

@item D
light magenta

@item E
yellow

@item F
white
@end table

The background is represented by 3 bits, so you cannot specify more than
7 for it.

This command can be used in the configuration file and on the
command line, so you may write something like this in your configuration
file:

@example
# Set default colors (light gray / blue, black / light gray).
color 0x17 0x70

# Change the colors.
title OS-BS like
color 0x16 0x60
@end example

@item chainloader @var{file}
Load @var{file} as a chain-loader. Like any other file loaded by the
filesystem code, it can use the blocklist notation to grab the first
sector of the current partition with @samp{+1}.

@item configfile @var{file}
Load @var{file} as the configuration file.

@item displaymem
Display what GRUB thinks the system address space map of the machine is,
including all regions of physical @sc{ram} installed. GRUB's
@dfn{upper/lower memory} display uses the standard BIOS interface for
the available memory in the first megabyte, or @dfn{lower memory}, and a
synthesized number from various BIOS interfaces of the memory starting
at 1MB and going up to the first chipset hole for @dfn{upper memory}
(the standard PC @dfn{upper memory} interface is limited to reporting a
maximum of 64MB).

@item fstest
Toggle filesystem test mode.

Filesystem test mode, when turned on, prints out data corresponding to
all the device reads and what values are being sent to the low-level
routines. The format is @samp{<@var{partition-offset-sector},
@var{byte-offset}, @var{byte-length}>} for high-level reads inside a
partition, and @samp{[@var{disk-offset-sector}]} for low-level sector
requests from the disk.

Filesystem test mode is turned off by any use of the @command{install}
or @command{testload} commands.

@item help [@var{pattern} @dots{}]
Display helpful information about builtin commands. If you do not
specify @var{pattern}, this command lists the short documents of all
available commands, and, if you specify one or more @var{pattern}s, it
displays long documents of the commands which match @var{pattern}.

@item hide @var{partition}
Hide @var{partition} by setting the @dfn{hidden} bit in its partition
type code.

@item impsprobe
Probe the Intel Multiprocessor Specification 1.1 or 1.4 configuration
table and boot the various CPUs which are found into a tight loop.

@item initrd @var{file} @dots{}
Load an initial ramdisk for a Linux format boot image and set the
appropriate parameters in the Linux setup area in memory.

@item install @var{stage1-file} [d] @var{dest-device} @var{file} [@var{addr}] [p] [@var{config-file}]
This command is fairly complex, and for detailed examples one should
look at @ref{Automated install}. In short, it will perform a full
install presuming the stage2 or stage1.5@footnote{They're loaded the
same way, so we will refer to the stage1.5 as a stage2 from now on.}
is in its final install location.

In slightly more detail, it will load @var{stage1-file}, validate that
it is a GRUB stage1 of the right version number, install a blocklist for
loading @var{file} as a stage2.  If the option @samp{d} is present, the
stage1 will always look for the actual disk @var{file} was installed on,
rather than using the booting drive.  The stage2 will be loaded at
address @var{addr}, which must be @samp{0x8000} for a true stage2, and
@samp{0x2000} for a stage1.5. If @var{addr} is not present, GRUB will
determine the address automatically. It then writes the completed stage1
to the first block of the device @var{dest-dev}. If the options @samp{p}
or @var{config-file} are present, then it reads the first block of
stage2, modifies it with the values of the partition @var{file} was
found on (for @samp{p}) or places the string @var{config-file} into the
area telling the stage2 where to look for a configuration file at boot
time. This command preserves the DOS BPB (and for hard disks, the
partition table) of the sector the stage1 is to be installed into.

@item kernel @var{file} @dots{}
Attempt to load the primary boot image (Multiboot a.out or @sc{elf},
Linux zImage or bzImage, FreeBSD a.out, or NetBSD a.out) from
@var{file}. The rest of the line is passed verbatim as the @dfn{kernel
command line}. Any modules must be reloaded after using this command.

@item makeactive
Set the active partition on the root disk to GRUB's root partition.
This command is limited to @emph{primary} PC partitions on a hard disk.

@item module @var{file} @dots{}
Load a boot module @var{file} for a Multiboot format boot image (no
interpretation of the file contents are made, so that user of this
command must know what the kernel in question expects). The rest of the
line is passed as the @dfn{module command line}, like the
@command{kernel} command.

@item modulenounzip @var{file} @dots{}
The same as @command{module}, except that automatic decompression is
disabled.

@item pause @var{message}@dots{}
Print the @var{message}, then wait until a key is pressed. Note that
placing @key{^G} (ASCII code 7) in the message will cause the speaker to
emit the standard beep sound, which is useful when prompting the user to
change floppies.

@item quit
Exit from the GRUB shell @command{grub} (@pxref{The grub shell}). This
command is ignored in the native Stage 2.

@item read @var{addr}
Read a 32-bit value from memory at address @var{addr} and
display it in hex format.

@item root @var{device} [@var{hdbias}]
Set the current @dfn{root partition} to the device @var{device}, then
attempt to mount it to get the partition size (for passing the partition
descriptor in @code{ES:ESI}, used by some chain-loaded bootloaders), the
BSD drive-type (for booting BSD kernels using their native boot format),
and correctly determine the PC partition where a BSD sub-partition is
located. The optional @var{hdbias} parameter is a number to tell a BSD
kernel how many BIOS drive numbers are on controllers before the current
one. For example, if there is an IDE disk and a SCSI disk, and your
FreeBSD root partition is on the SCSI disk, then use a @samp{1} for
@var{hdbias}.

@item rootnoverify @var{device} [@var{hdbias}]
Similar to @command{root}, but don't attempt to mount the
partition. This is useful for when an OS is outside of the area of the
disk that GRUB can read, but setting the correct root partition is still
desired. Note that the items mentioned in @command{root} above which
derived from attempting the mount will @emph{not} work correctly.

@item testload @var{file}
Read the entire contents of @var{file} in several different ways and
compares them, to test the filesystem code. The output is somewhat
cryptic , but if no errors are reported and the final @samp{i=@var{X},
filepos=@var{Y}} reading has @var{X} and @var{Y} equal, then it is
definitely consistent, and very likely works correctly subject to a
consistent offset error. If this test succeeds, then a good next step is
to try loading a kernel.

@item unhide @var{partition}
Unhide @var{partition} by clearing the @dfn{hidden} bit in its partition
type code.

@item uppermem @var{kbytes}
Force GRUB to assume that only @var{kbytes} kilobytes of upper memory
are installed.  Any system address range maps are discarded.

@strong{Caution:} This should be used with great caution, and should
only be necessary on some old machines. GRUB's BIOS probe can pick up
all @sc{ram} on all new machines the author has ever heard of. It can
also be used for debugging purposes to lie to an OS.
@end table


@node Filesystems
@chapter Filesystem syntax and semantics

GRUB uses a special syntax for specifying disk drives which can be
accessed by BIOS.  Because of BIOS limitations, GRUB cannot distinguish
between IDE, ESDI, SCSI, or others.   You must know yourself which BIOS
device is equivalent to which OS device.

@menu
* Device syntax::               How to specify devices.
* Filename syntax::             How to specify files.
@end menu


@node Device syntax
@section How to specify devices

The device syntax is like this:

@example
@code{(@var{bios-device}[,@var{part-num}][,@var{bsd-subpart-letter}])}
@end example

@samp{[]} means the parameter is optional. @var{bios-device} should be
either @samp{fd} or @samp{hd} followed by a digit, like @samp{fd0}.
But you can also set @var{bios-device} to a hexadecimal or a decimal,
which is a BIOS drive number, so the following are equivalent:

@example
(hd0)
(0x80)
(128)
@end example

@var{part-num} represents the partition number of @var{bios-device},
starting from zero, and @var{bsd-subpart-letter} represents the BSD
disklabel subpartition, such as @samp{a} or @samp{e}.

A shortcut for specifying BSD subpartitions is
@code{(@var{bios-device},@var{bsd-subpart-letter})}, in this case, GRUB
searches for the first PC partition containing a BSD disklabel, then
finds the subpartition @var{bsd-subpart-letter}. Here is an example:

@example
(hd0,a)
@end example


@node Filename syntax
@section How to specify files

There are two ways to specify files, by @dfn{absolute filename} and by
@dfn{blocklist}.

An absolute filename resembles a Unix absolute filename, using @samp{/}
for the directory separator (not @samp{\} as in DOS).  One example of an
absolute filename is @samp{/boot/grub/menu.lst}.

A blocklist is used for specifying a file that doesn't appear in the
filesystem, like a chainloader. The syntax is
@code{[@var{offset}]+@var{length}[,[@var{offset}]+@var{length}]@dots{}}.
Here is an example:

@example
@code{0+100,200+1,300+300}
@end example

This represents that GRUB should read blocks 0 through 99, block 200,
and blocks 300 through 600.  If you omit an offset, then GRUB assumes
the offset is zero.


@node Troubleshooting
@chapter Error messages reported by GRUB

This chapter describes error messages reported by GRUB when you
encounter trouble.

@menu
* Stage1 errors::               Errors reported by the Stage 1.
* Stage1.5 errors::             Errors reported by the Stage 1.5.
* Stage2 errors::               Errors reported by the Stage 2.
@end menu


@node Stage1 errors
@section Errors reported by the Stage 1

The general way that the Stage 1 handles errors is to print an error
string and then halt. Pressing @kbd{@key{CTRL}-@key{ALT}-@key{DEL}} will
reboot.

The following is a comprehensive list of error messages for the Stage 1:

@table @asis
@item Hard Disk Error
The stage2 or stage1.5 is being read from a hard disk, and the attempt
to determine the size and geometry of the hard disk failed.

@item Floppy Error
The stage2 or stage1.5 is being read from a floppy disk, and the attempt
to determine the size and geometry of the floppy disk failed. It's listed
as a separate error since the probe sequence is different than for hard
disks.

@item Read Error
A disk read error happened while trying to read the stage2 or stage1.5.

@item Geom Error
The location of the stage2 or stage1.5 is not in the portion of the disk
supported directly by the BIOS read calls.  This could occur because the
BIOS translated geometry has been changed by the user or the disk is
moved to another machine or controller after installation, or GRUB was
not installed using itself (if it was, the Stage 2 version of this error
would have been seen during that process and it would not have completed
the install).
@end table

@c FIXME: Gord stopped here

@node Stage1.5 errors
@section Errors reported by the Stage 1.5

The general way that the Stage 1.5 handles errors is to print an error
number in the form @code{Error: @var{num}} and then halt. Pressing
@kbd{@key{CTRL}-@key{ALT}-@key{DEL}} will reboot.

The error numbers correspond to the @ref{Stage2 errors} in the listed
sequence.


@node Stage2 errors
@section Errors reported by the Stage 2

The general way that the Stage 2 handles errors is to abort the
operation in question, print an error string, then (if possible) either
continue based on the fact that an error occurred or wait for the user to
deal with the error.

The following is a comprehensive list of error messages for the Stage 2
(error numbers for the Stage 1.5 are listed before the colon in each
description):

@table @asis
@item 1 : Bad filename (must be absolute filename or blocklist)
This error is returned if a filename is requested which doesn't fit the
syntax/rules listed in the @ref{Filesystems}.

@item 2 : Bad file or directory type
This error is returned if a file requested is not a regular file, but
something like a symbolic link, directory, or FIFO.

@item 3 : Bad or corrupt data while decompressing file
This error is returned the run-length decompression code gets an
internal error. This is usually from a corrupt file.

@item 4 : Bad or incompatible header on compressed file
This error is returned if the file header for a supposedly compressed
file is bad.

@item 5 : Partition table invalid or corrupt
This error s returned if the sanity checks on the integrity of the
partition table fail. This is a bad sign.

@item 6 : Bad or corrupt version of stage1/stage2
This error is returned if the install command is pointed to incompatible
or corrupt versions of the stage1 or stage2. It can't detect corruption
in general, but this is a sanity check on the version numbers, which
should be correct.

@item 7 : Loading below 1MB is not supported
This error is returned if the lowest address in a kernel is below the
1MB boundary. The Linux zImage format is a special case and can be
handled since it has a fixed loading address and maximum size.

@item 8 : Cannot boot without kernel loaded
This error is returned if GRUB is told to execute the boot sequence
without having a kernel to start.

@item 9 : Unknown boot failure
This error is returned if the boot attempt did not succeed for reasons
which are unknown.

@item 10 : Unsupported Multiboot features requested
This error is returned when the Multiboot features word in the Multiboot
header requires a feature that is not recognized. The point of this is
that the kernel requires special handling which GRUB is likely unable to
provide.

@item 11 : Device string unrecognizable
This error is returned if a device string was expected, and the string
encountered didn't fit the syntax/rules listed in the @ref{Filesystems}.

@item 12 : Invalid device requested
This error is returned if a device string is recognizable but does not
fall under the other device errors.

@item 13 : Invalid or unsupported executable format
This error is returned if the kernel image being loaded is not
recognized as Multiboot or one of the supported native formats (Linux
zImage or bzImage, FreeBSD, or NetBSD).

@item 14 : Filesystem compatibility error, can't read whole file
Some of the filesystem reading code in GRUB has limits on the length of
the files it can read. This error is returned when the user runs into
such a limit.

@item 15 : File not found
This error is returned if the specified filename cannot be found, but
everything else (like the disk/partition info) is OK.

@item 16 : Inconsistent filesystem structure
This error is returned by the filesystem code to denote an internal
error caused by the sanity checks of the filesystem structure on disk
not matching what it expects. This is usually caused by a corrupt
filesystem or bugs in the code handling it in GRUB.

@item 17 : Cannot mount selected partition
This error is returned if the partition requested exists, but the
filesystem type cannot be recognized by GRUB.

@item 18 : Selected cylinder exceeds maximum supported by BIOS
This error is returned when a read is attempted at a linear block
address beyond the end of the BIOS translated area. This generally
happens if your disk is larger than the BIOS can handle (512MB for
(E)IDE disks on older machines or larger than 8GB in general).

@item 19 : Must load Linux kernel before initrd
This error is returned if the initrd command is used before loading a
Linux kernel. Similar to the above error, it only makes sense in that
case anyway.

@item 20 : Must load Multiboot kernel before modules
This error is returned if the module load command is used before loading
a Multiboot kernel. It only makes sense in this case anyway, as GRUB has
no idea how to communicate the presence of location of such modules to a
non-Multiboot-aware kernel.

@item 21 : Selected disk does not exist
This error is returned if the device part of a device- or full filename
refers to a disk or BIOS device that is not present or not recognized by
the BIOS in the system.

@item 22 : No such partition
This error is returned if a partition is requested in the device part of
a device- or full filename which isn't on the selected disk.

@item 23 : Error while parsing number
This error is returned if GRUB was expecting to read a number and
encountered bad data.

@item 24 : Attempt to access block outside partition
This error is returned if a linear block address is outside of the disk
partition. This generally happens because of a corrupt filesystem on the
disk or a bug in the code handling it in GRUB (it's a great debugging
tool).

@item 25 : Disk read error
This error is returned if there is a disk read error when trying to
probe or read data from a particular disk.

@item 26 : Too many symbolic links
This error is returned if the link count is beyond the maximum
(currently 5), possibly the symbolic links are looped.

@item 27 : Unrecognized command
This error is returned if an unrecognized command is entered into the
command line or in a boot sequence section of a configuration file and
that entry is selected.

@item 28 : Selected item won't fit into memory
This error is returned if a kernel, module, or raw file load command is
either trying to load its data such that it won't fit into memory or it
is simply too big.

@item 29 : Disk write error
This error is returned if there is a disk write error when trying to
write to a particular disk. This would generally only occur during an
install of set active partition command.
@end table


@node The grub shell
@chapter The command @command{grub}

This chapter documents the grub shell @command{grub}.

@menu
* Basic usage::                 How to use the grub shell.
* Installation under UNIX::     How to install GRUB via @command{grub}.
@end menu


@node Basic usage
@section Introduction into the grub shell

You can use the command @command{grub} for installing GRUB under your
operating systems and for a testbed when you add a new feature into GRUB
or when fix a bug. @command{grub} is almost the same as Stage 2, and, in
fact, it shares the source code with Stage 2 and you can use the same
commands in @command{grub}. It is emulated by replacing BIOS calls with
UNIX system calls and libc functions.

The command @command{grub} accepts the following options:

@table @code
@item --help
Print a summary of the command line options and exit.

@item --version
Print the version number of GRUB and exit.

@item --verbose
Print some verbose messages for debugging purpose.

@item --config-file=@var{file}
Read the configuration file @var{file} instead of
@file{/boot/grub/menu.lst}. The format is the same as the normal GRUB
syntax. See @ref{Filesystems}, for more information.

@item --boot-drive=@var{drive}
Set the stage2 @var{boot_drive} to @var{drive}. This argument should be
an integer (decimal, octal or hexadecimal).

@item --install-partition=@var{par}
Set the stage2 @var{install_partition} to @var{par}. This argument
should be an integer (decimal, octal or hexadecimal).

@item --no-config-file
Do not use the configuration file even if it can be read.

@item --no-curses
Do not use the curses interface even if it is available.

@item --batch
This option has the same meaning as @samp{--no-config-file --no-curses}.

@item --read-only
Disable writing to any disk.

@item --hold
Wait until a debegger will attach. This option is useful when you want
to debug the startup code.
@end table


@node Installation under UNIX
@section How to install GRUB via @command{grub}

The installation procedure is the same as under the @dfn{native} Stage
2. See @ref{Automated install} for more information. The command
@command{grub}-specific information is described here.

What you should be careful about is @dfn{buffer cache}. @command{grub}
makes use of raw devices instead of filesystems that your operating
systems serve, so there exists a potential problem that some cache
inconsistency may corrupt your filesystems. What we recommend is:

@itemize @bullet
@item
If you can unmount drives to which GRUB may write any amount of data,
unmount them before running @command{grub}.

@item
If a drive cannot be unmounted but can be mounted with the read-only
flag, mount it in read-only mode. That should be secure.

@item
If a drive must be mounted with the read-write flag, make sure that any
activity is not being done on it during running the command
@command{grub}.

@item
Reboot your operating system as soon as possible. Probably that is not
required if you follow these rules above, but reboot is the most secure
way.
@end itemize

In addition, enter the command @command{quit} when you finish the
installation. That is @emph{very important} because @command{quit} makes
the buffer cache consistent. Do not push @key{C-c}.

If you want to install GRUB non-interactively, specify @samp{--batch}
option in the command line. This is a simple example:

@example
#!/bin/sh

/sbin/grub --batch <<EOT 1>/dev/null 2>/dev/null
root (hd0,0)
install /boot/grub/stage1 (hd0) /boot/grub/stage2 p
quit
EOT
@end example


@node Hacking
@chapter Implementation details

This chapter describes the GRUB internals so that developers can
understand the implementation and start to hack GRUB. Of course, the
source code has the complete information, so refer to it when you are
not satisfied with this documentation.

@menu
* Memory map::                  The memory map of the various
                                components.
* Embedded data::               Embedded variables in GRUB.
* Memory detection::            How to detect all installed @sc{ram}.
* Low-level disk I/O::          INT 13H disk I/O interrupts.
* MBR::                         The structure of Master Boot Record.
* Partition table::             The format of partition table.
* Filesystem interface::        The generic interface for the fs code.
@end menu


@node Memory map
@section The memory map of various components

GRUB is broken into 2 distinct components, or @dfn{stages}, which are
loaded at different times in the boot process. The Stage 1 has to know
where to find Stage 2, and the Stage 2 has to know where to find its
configuration file (if Stage 2 doesn't have a configuration file, it
drops into the command line interface and waits for a user command).

Here is the memory map of the various components
@footnote{Currently GRUB does not use the extended memory for itself,
since it is used to load an operating system. But we are planning to use
it for GRUB itself in the future by @dfn{lazy loading}. Ask okuji for
more information.}:

@table @asis
@item 0 to 4K-1
Interrupt & BIOS area

@item down from 8K-1
16-bit stack area

@item 8K to (ebss1.5)
Stage 1.5 (optionally) loaded here by Stage 1

@item 0x7c00 to 0x7dff
Stage 1 loaded here by the BIOS

@item 0x7e00 to 0x7e08
Scratch space used by Stage 1

@item 32K to (ebss2)
Stage 2 loaded here by Stage 1.5 or Stage 1

@item (middle area)
Heap used for random memory allocation

@item down from 416K-1
32-bit stack area

@item 416K to 448K-1
Filesystem info buffer (when reading a filesystem)

@item 448K to 479.5K-1
BIOS track read buffer

@item 479.5K to 480K-1
512 byte fixed SCRATCH area

@item 480K to 511K-1
General storage heap
@end table


@node Embedded data
@section Embedded variables in GRUB

GRUB's stage1 and stage2 have embedded variables whose locations are
well-defined, so that the installation can patch the binary file
directly without recompilation of the modules.

In stage1, these are defined (The number in the parenthesis of each
entry is an offset number):

@table @asis
@item @dfn{major version} (0x1bc)
This is the major version byte (should be 2).

@item @dfn{minor version} (0x1bd)
This is the minor version byte (should be 1).

@item @dfn{stage2 start address} (0x1b8)
This is the data for the @code{ljmp} command to the starting address of
the component loaded by the stage1. The format is two 2-byte words: the
first is the IP, and the second is the CS segment register (remember,
we're in x86 real-mode@dots{} 16-bit instruction pointer and segment
registers shifted left by 4bits).

A @dfn{stage1.5} should be loaded at address 0x2000, and a @dfn{stage2}
should be loaded at address 0x8000. Both use a CS of 0.

@item @dfn{firstlist} (0x1b05)
This is the @emph{ending} address of the blocklist data area.

The trick here is that it is actually read backward, and the first
8-byte blocklist is not read here, but after the pointer is decremented
8 bytes, then after reading it, it decrements again, reads, decrements,
reads, etc. until it is finished.  The terminating condition is when the
number of sectors to be read in the next blocklist is 0.

The format of a blocklist can be seen from the example in the code just
before the @code{firstlist} label.  (note that it is always from the
beginning of the disk, and @emph{not} relative to the partition
boundaries)

@item @dfn{loading drive} (0x1b05)
This is the BIOS drive number to load the blocklists from. If the number
is 0xff, then load from the booting drive.
@end table

In stage1.5 and stage2 (these are all defined at the beginning of
@file{shared_src/asm.S}):

@table @asis
@item @dfn{major version} (0x6)
This is the major version byte (should be 2).

@item @dfn{minor version} (0x7)
This is the minor version byte (should be 0).

@item @dfn{install_partition} (0x8)
This is an unsigned long representing the partition on the currently
booted disk which GRUB should expect to find it's data files and treat
as the default root partition.

The format of is exactly the same as the @dfn{partition} part (the
@dfn{disk} part is ignored) of the data passed to an OS by a
Multiboot-compliant bootloader in the @dfn{boot_device} data element,
with one exception.

The exception is that if the first level of disk partitioning is left as
0xFF (decimal 255, which is marked as no partitioning being used), but
the second level does have a partition number, it looks for the first
BSD-style PC partition, and finds the numbered BSD sub-partition in it.
The default @dfn{install_partition} 0xFF00FF, would then find the first
BSD-style PC partition, and use the @samp{a} partition in it, and
0xFF01FF would use the @samp{b} partition, etc.

If an explicit first-level partition is given, then no search is
performed, and it will expect that the BSD-style PC partition is in the
appropriate location, else a @samp{no such partition} error will be
returned.

If a @dfn{stage1.5} is being used, it will pass its own
@dfn{install_partition} to any @dfn{stage2} it loads, therefore
overwriting the one present in the @dfn{stage2}.

@item @dfn{stage2_id} (0xc)
This is the @dfn{stage1.5} or @dfn{stage2} identifier.

@item @dfn{version_string} (0xd)
This is the @dfn{stage1.5} or @dfn{stage2} version string. It isn't
meant to be changed, simply easy to find.

@item @dfn{config_file} (after the terminating zero of @dfn{version_string})
This is the location, using the GRUB filesystem syntax, of the config
file.  It will, by default, look in the @dfn{install_partition} of the
disk GRUB was loaded from, though one can use any valid GRUB filesystem
string, up to and including making it look on other disks.

The bootloader itself doesn't search for the end of
@dfn{version_string}, it simply knows where @dfn{config_file} is, so the
beginning of the string cannot be moved after compile-time.  This should
be OK, since the @dfn{version_string} is meant to be static.

The code of stage2 starts again at offset 0x70,	so @dfn{config_file}
string obviously can't go past there.  Also, remember to terminate the
string with a 0.
@end table


@node Memory detection
@section How to detect all installed @sc{ram}

There are three BIOS calls which return the information of installed
@sc{ram}. GRUB uses these calls to detect all installed @sc{ram} and
which address range should be treated by operating systems.

@menu
* Query System Address Map::    INT 15H, AX=E820h interrupt call.
* Get Large Memory Size::       INT 15H, AX=E801h interrupt call.
* Get Extended Memory Size::    INT 15H, AX=88h interrupt call.
@end menu


@node Query System Address Map
@subsection INT 15H, AX=E820h interrupt call

Real mode only.

This call returns a memory map of all the installed @sc{ram}, and of
physical memory ranges reserved by the BIOS. The address map is returned
by making successive calls to this API, each returning one "run" of
physical address information.  Each run has a type which dictates how
this run of physical address range should be treated by the operating
system.

If the information returned from INT 15h, AX=E820h in some way differs
from INT 15h, AX=E801h (@pxref{Get Large Memory Size}) or INT 15h AH=88h
(@pxref{Get Extended Memory Size}), then the information returned from
E820h supersedes what is returned from these older interfaces. This
allows the BIOS to return whatever information it wishes to for
compatibility reasons.

Input:

@multitable @columnfractions .15 .25 .6
@item @code{EAX} @tab Function Code @tab E820h

@item @code{EBX} @tab Continuation @tab Contains the @dfn{continuation
value} to get the next run of physical memory. This is the value
returned by a previous call to this routine. If this is the first call,
@code{EBX} must contain zero.

@item @code{ES:DI} @tab Buffer Pointer @tab Pointer to an Address Range
Descriptor structure which the BIOS is to fill in.

@item @code{ECX} @tab Buffer Size @tab The length in bytes of the
structure passed to the BIOS. The BIOS will fill in at most @code{ECX}
bytes of the structure or however much of the structure the BIOS
implements. The minimum size which must be supported by both the BIOS
and the caller is 20 bytes. Future implementations may extend this
structure.

@item @code{EDX} @tab Signature @tab @samp{SMAP} - Used by the BIOS to
verify the caller is requesting the system map information to be
returned in @code{ES:DI}.
@end multitable

Output:

@multitable @columnfractions 0.15 0.25 0.6
@item @code{CF} @tab Carry Flag @tab Non-Carry - indicates no error

@item @code{EAX} @tab Signature @tab @samp{SMAP} - Signature to verify
correct BIOS revision.

@item @code{ES:DI} @tab Buffer Pointer @tab Returned Address Range
Descriptor pointer. Same value as on input.

@item @code{ECX} @tab Buffer Size @tab Number of bytes returned by the
BIOS in the address range descriptor. The minimum size structure
returned by the BIOS is 20 bytes.

@item @code{EBX} @tab Continuation @tab Contains the continuation value
to get the next address descriptor. The actual significance of the
continuation value is up to the discretion of the BIOS. The caller must
pass the continuation value unchanged as input to the next iteration of
the E820h call in order to get the next Address Range Descriptor. A
return value of zero means that this is the last descriptor. Note that
the BIOS indicate that the last valid descriptor has been returned by
either returning a zero as the continuation value, or by returning
carry.
@end multitable

The Address Range Descriptor Structure is:

@multitable @columnfractions 0.25 0.3 0.45
@item Offset in Bytes @tab Name @tab Description

@item 0 @tab @dfn{BaseAddrLow} @tab Low 32 Bits of Base Address

@item 4 @tab @dfn{BaseAddrHigh} @tab High 32 Bits of Base Address

@item 8 @tab @dfn{LengthLow} @tab Low 32 Bits of Length in Bytes

@item 12 @tab @dfn{LengthHigh} @tab High 32 Bits of Length in Bytes

@item 16 @tab @dfn{Type} @tab Address type of this range
@end multitable

The @dfn{BaseAddrLow} and @dfn{BaseAddrHigh} together are the 64 bit
@dfn{BaseAddress} of this range. The @dfn{BaseAddress} is the physical
address of the start of the range being specified.

The @dfn{LengthLow} and @dfn{LengthHigh} together are the 64 bit
@dfn{Length} of this range. The @dfn{Length} is the physical contiguous
length in bytes of a range being specified.

The @dfn{Type} field describes the usage of the described address range
as defined in the table below:

@multitable @columnfractions 0.1 0.35 0.55
@item Value @tab Pneumonic @tab Description

@item 1 @tab @dfn{AddressRangeMemory} @tab This run is available
@sc{ram} usable by the operating system.

@item 2 @tab @dfn{AddressRangeReserved} @tab This run of addresses is in
use or reserved by the system, and must not be used by the operating
system.

@item Other @tab @dfn{Undefined} @tab Undefined - Reserved for future
use. Any range of this type must be treated by the OS as if the type
returned was @dfn{AddressRangeReserved}.
@end multitable

The BIOS can use the @dfn{AddressRangeReserved} address range type to
block out various addresses as @emph{not suitable} for use by a
programmable device.

Some of the reasons a BIOS would do this are:

@itemize @bullet
@item
The address range contains system @sc{rom}.

@item
The address range contains @sc{ram} in use by the @sc{rom}.

@item
The address range is in use by a memory mapped system device.

@item
The address range is for whatever reason are unsuitable for a
standard device to use as a device memory space.
@end itemize

Here is the list of assumptions and limitations:

@enumerate
@item
The BIOS will return address ranges describing base board memory and ISA
or PCI memory that is contiguous with that base board memory.

@item
The BIOS @emph{will not} return a range description for the memory
mapping of PCI devices. ISA Option @sc{rom}'s, and ISA plug & play
cards. This is because the OS has mechanisms available to detect them.

@item
The BIOS will return chipset defined address holes that are not being
used by devices as reserved.

@item
Address ranges defined for base board memory mapped I/O devices (for
example APICs) will be returned as reserved.

@item
All occurrences of the system BIOS will be mapped as reserved. This
includes the area below 1 MB, at 16 MB (if present) and at end of the
address space (4 GB).

@item
Standard PC address ranges will not be reported. Example video memory at
A0000 to BFFFF physical will not be described by this function. The
range from E0000 to EFFFF is base board specific and will be reported as
suits the bas board.

@item
All of lower memory is reported as normal memory. It is OS's
responsibility to handle standard @sc{ram} locations reserved for
specific uses, for example: the interrupt vector table (0:0) and the
BIOS data area (40:0).
@end enumerate

Here we explain an example address map. This sample address map
describes a machine which has 128 MB @sc{ram}, 640K of base memory and
127 MB extended.  The base memory has 639K available for the user and 1K
for an extended BIOS data area.  There is a 4 MB Linear Frame Buffer
(LFB) based at 12 MB.  The memory hole created by the chipset is from 8
M to 16 M.  There are memory mapped APIC devices in the system.  The IO
Unit is at FEC00000 and the Local Unit is at FEE00000.  The system BIOS
is remapped to 4G - 64K.

Note that the 639K endpoint of the first memory range is also the base
memory size reported in the BIOS data segment at 40:13.

Key to types: @dfn{ARM} is AddressRangeMemory, @dfn{ARR} is
AddressRangeReserved.

@multitable @columnfractions 0.15 0.1 0.1 0.65
@item Base (Hex) @tab Length @tab Type @tab Description

@item 0000 0000 @tab 639K @tab ARM @tab Available Base memory -
typically the same value as is returned via the INT 12 function.

@item 0009 FC00 @tab 1K @tab ARR @tab Memory reserved for use by the
BIOS(s). This area typically includes the Extended BIOS data area.

@item 000F 0000 @tab 64K @tab ARR @tab System BIOS.

@item 0010 0000 @tab 7M @tab ARM @tab Extended memory, this is not
limited to the 64MB address range.

@item 0080 0000 @tab 8M @tab ARR @tab Chipset memory hole required to
support the LFB mapping at 12 MB.

@item 0100 0000 @tab 120M @tab ARM @tab Base board @sc{ram} relocated
above a chipset memory hole.

@item FE00 0000 @tab 4K @tab ARR @tab IO APIC memory mapped I/O at
FEC00000. Note the range of addresses required for an APIC device may
vary from base OEM to OEM.

@item FEE0 0000 @tab 4K @tab ARR @tab Local APIC memory mapped I/O at
FEE00000.

@item FFFF 0000 @tab 64K @tab ARR @tab Remapped System BIOS at end of
address space.
@end multitable

The following code segment is intended to describe the algorithm needed
when calling the Query System Address Map function. It is an
implementation example and uses non standard mechanisms.

@example
E820Present = FALSE;
Regs.ebx = 0;
do
  @{
    Regs.eax = 0xE820;
    Regs.es = SEGMENT (&Descriptor);
    Regs.di = OFFSET (&Descriptor);
    Regs.ecx = sizeof (Descriptor);
    Regs.edx = 'SMAP';

    _int (0x15, Regs);

    if ((Regs.eflags & EFLAGS_CARRY) || Regs.eax != 'SMAP')
      @{
        break;
      @}

    if (Regs.ecx < 20 || Regs.ecx > sizeof (Descriptor))
      @{
        /* bug in bios - all returned descriptors must be at
           least 20 bytes long, and can not be larger than
           the input buffer.  */
        break;
      @}

    E820Present = TRUE;
    .
    .
    .
    Add address range Descriptor.BaseAddress through
    Descriptor.BaseAddress + Descriptor.Length
    as type Descriptor.Type
    .
    .
    .
  @}
while (Regs.ebx != 0);

if (! E820Present)
  @{
    .
    .
    .
    call INT 15H, AX E801h and/or INT 15H, AH=88h to obtain old style
    memory information
    .
    .
    .
  @}
@end example


@node Get Large Memory Size
@subsection INT 15H, AX=E801h interrupt call

Real mode only.

Originally defined for EISA servers, this interface is capable of
reporting up to 4 GB of @sc{ram}.  While not nearly as flexible as
E820h, it is present in many more systems.

Input:

@multitable @columnfractions 0.15 0.25 0.6
@item @code{AX} @tab Function Code @tab E801h.
@end multitable

Output:

@multitable @columnfractions 0.15 0.25 0.6
@item @code{CF} @tab Carry Flag @tab Non-Carry - indicates no error.

@item @code{AX} @tab Extended 1 @tab Number of contiguous KB between 1
and 16 MB, maximum 0x3C00 = 15 MB.

@item @code{BX} @tab Extended 2 @tab Number of contiguous 64KB blocks
between 16 MB and 4GB.

@item @code{CX} @tab Configured 1 @tab Number of contiguous KB between 1
and 16 MB, maximum 0x3c00 = 15 MB.

@item @code{DX} @tab Configured 2 @tab Number of contiguous 64KB blocks
between 16 MB and 4 GB.
@end multitable

Not sure what this difference between the @dfn{Extended} and
@dfn{Configured} numbers are, but they appear to be identical, as
reported from the BIOS.

It is possible for a machine using this interface to report a memory
hole just under 16 MB (Count 1 is less than 15 MB, but Count 2 is
non-zero).


@node Get Extended Memory Size
@subsection INT 15H, AX=88h interrupt call

Real mode only.

This interface is quite primitive.  It returns a single value for
contiguous memory above 1 MB.  The biggest limitation is that the value
returned is a 16-bit value, in KB, so it has a maximum saturation of
just under 64 MB even presuming it returns as much as it can.  On some
systems, it won't return anything above the 16 MB boundary.

The one useful point is that it works on every PC available.

Input:

@multitable @columnfractions 0.15 0.25 0.6
@item @code{AH} @tab Function Code @tab 88h
@end multitable

Output:

@multitable @columnfractions 0.15 0.25 0.6
@item @code{CF} @tab Carry Flag @tab Non-Carry - indicates no error.

@item @code{AX} @tab Memory Count @tab Number of contiguous KB above 1
MB.
@end multitable


@node Low-level disk I/O
@section INT 13H disk I/O interrupts

In the PC world, living with the BIOS disk interface is definitely a
nightmare. This section documents how awful the chaos is and how GRUB
deals with the BIOS disks.

@menu
* CHS Translation::             CHS addressing and LBA addressing.
* CHS mode disk I/O::           INT 13H, AH=0xh interrupt call.
* LBA mode disk I/O::           INT 13H, AH=4xh interrupt call.
@end menu


@node CHS Translation
@subsection CHS addressing and LBA addressing

CHS --- Cylinder/Head/Sector --- is the traditional way to address
sectors on a disk. There are at least two types of CHS addressing; the
CHS that is used at the INT 13H interface and the CHS that is used at
the ATA device interface. In the MFM/RLL/ESDI and early ATA days the CHS
used at the INT 13H interface was the same as the CHS used at the device
interface.

Today we have CHS translating BIOS types that can use one CHS at the INT
13H interface and a different CHS at the device interface. These two
types of CHS will be called the logical CHS or @dfn{L-CHS} and the
physical CHS or @dfn{P-CHS} in this section. L-CHS is the CHS used at
the INT 13H interface and P-CHS is the CHS used at the device interface.

The L-CHS used at the INT 13 interface allows up to 256 heads, up to
1024 cylinders and up to 63 sectors. This allows support of up to 8GB
drives. This scheme started with either ESDI or SCSI adapters many years
ago.

The P-CHS used at the device interface allows up to 16 heads up to 65535
cylinders, and up to 63 sectors. This allows access to 2^28 sectors
(136GB) on an ATA device.  When a P-CHS is used at the INT 13H interface
it is limited to 1024 cylinders, 16 heads and 63 sectors. This is where
the old 528MB limit originated.

LBA --- Logical Block Address --- is another way of addressing sectors
that uses a simple numbering scheme starting with zero as the address of
the first sector on a device. The ATA standard requires that cylinder 0,
head 0, sector 1 address the same sector as addressed by LBA 0. LBA
addressing can be used at the ATA interface if the ATA device supports
it. LBA addressing is also used at the INT 13H interface by the AH=4xH
read/write calls.

ATA devices may also support LBA at the device interface. LBA allows
access to approximately 2^28 sectors (137GB) on an ATA device.

A SCSI host adapter can convert a L-CHS directly to an LBA used in the
SCSI read/write commands. On a PC today, SCSI is also limited to 8GB
when CHS addressing is used at the INT 13H interface.

First, all OS's that want to be co-resident with another OS (and that is
all of the PC based OS's that I know of) @emph{must} use INT 13H to
determine the capacity of a hard disk. And that capacity information
@emph{must} be determined in L-CHS mode. Why is this?  Because:

@enumerate
@item
FDISK and the partition tables are really L-CHS based.

@item
MS/PC DOS uses INT 13H AH=02H and AH=03H to read and write the disk and
these BIOS calls are L-CHS based.

@item
The boot processing done by the BIOS is all L-CHS based.
@end enumerate

During the boot processing, all of the disk read accesses are done in
L-CHS mode via INT 13H and this includes loading the first of the OS's
kernel code or boot manager's code.

Second, because there can be multiple BIOS types in any one system, each
drive may be under the control of a different type of BIOS. For example,
drive 80H (the first hard drive) could be controlled by the original
system BIOS, drive 81H (the second drive) could be controlled by a
option @sc{rom} BIOS and drive 82H (the third drive) could be controlled
by a software driver. Also, be aware that each drive could be a
different type, for example, drive 80H could be an MFM drive, drive 81H
could be an ATA drive, drive 82H could be a SCSI drive.

Third, not all OS's understand or use BIOS drive numbers greater than
81H. Even if there is INT 13H support for drives 82H or greater, the OS
may not use that support.

Fourth, the BIOS INT 13H configuration calls are:

@table @asis
@item AH=08H, Get Drive Parameters
This call is restricted to drives up to 528MB without CHS translation
and to drives up to 8GB with CHS translation. For older BIOS with no
support for >1024 cylinders or >528MB, this call returns the same CHS as
is used at the ATA interface (the P-CHS). For newer BIOS's that do
support >1024 cylinders or >528MB, this call returns a translated CHS
(the L-CHS). The CHS returned by this call is used by FDISK to build
partition records.

@item AH=41H, Get BIOS Extensions Support
This call is used to determine if the IBM/Microsoft Extensions or if the
Phoenix Enhanced INT 13H calls are supported for the BIOS drive number.

@item AH=48H, Extended Get Drive Parameters
This call is used to determine the CHS geometries, LBA information and
other data about the BIOS drive number.
@end table

An ATA disk must implement both CHS and LBA addressing and must at any
given time support only one P-CHS at the device interface. And, the
drive must maintain a strick relationship between the sector addressing
in CHS mode and LBA mode. Quoting @cite{the ATA-2 document}:

@example
@group
LBA = ( (cylinder * heads_per_cylinder + heads )
        * sectors_per_track ) + sector - 1

where heads_per_cylinder and sectors_per_track are the current
translation mode values.
@end group
@end example

This algorithm can also be used by a BIOS or an OS to convert a L-CHS to
an LBA.

This algorithm can be reversed such that an LBA can be converted to a
CHS:

@example
@group
cylinder = LBA / (heads_per_cylinder * sectors_per_track)
    temp = LBA % (heads_per_cylinder * sectors_per_track)
    head = temp / sectors_per_track
  sector = temp % sectors_per_track + 1
@end group
@end example

While most OS's compute disk addresses in an LBA scheme, an OS like DOS
must convert that LBA to a CHS in order to call INT 13H.

The basic problem is that there is no requirement that a CHS translating
BIOS followed these rules. There are many other algorithms that can be
implemented to perform a similar function. Today, there are at least two
popular implementions: the Phoenix implementation (described above) and
the non-Phoenix implementations. Because a protected mode OS that does
not want to use INT 13H must implement the same CHS translation
algorithm. If it doesn't, your data gets scrambled.

In the perfect world of tomorrow, maybe only LBA will be used. But today
we are faced with the following problems:

@itemize @bullet
@item
Some drives >528MB don't implement LBA.

@item
Some drives are optimized for CHS and may have lower performance when
given commands in LBA mode. Don't forget that LBA is something new for
the ATA disk designers who have worked very hard for many years to
optimize CHS address handling. And not all drive designs require the use
of LBA internally.

@item
The L-CHS to LBA conversion is more complex and slower than the bit
shifting L-CHS to P-CHS conversion.

@item
DOS, FDISK and the MBR are still CHS based --- they use the CHS returned
by INT 13H AH=08H.  Any OS that can be installed on the same disk with
DOS must understand CHS addressing.

@item
The BIOS boot processing and loading of the first OS kernel code is done
in CHS mode --- the CHS returned by INT 13H AH=08H is used.

@item
Microsoft has said that their OS's will not use any disk capacity that
can not also be accessed by INT 13H AH=0xH.
@end itemize

These are difficult problems to overcome in today's industry
environment. The result: chaos.


@node CHS mode disk I/O
@subsection INT 13H, AH=0xh interrupt call

Real mode only. These functions are the traditional CHS mode disk
interface. GRUB calls them only if LBA mode is not available.

INT 13H, AH=02h reads sectors into memory.

Input:

@multitable @columnfractions .15 .85
@item @code{AH} @tab 02h

@item @code{AL} @tab The number of sectors to read (must be non-zero).

@item @code{CH} @tab Low 8 bits of cylinder number.

@item @code{CL} @tab Sector number in bits 0-5, and high 2 bits of
cylinder number in bits 6-7.

@item @code{DH} @tab Head number.

@item @code{DL} @tab Drive number (bit 7 set for hard disk).

@item @code{ES:BX} @tab Data buffer.
@end multitable

Output:

@multitable @columnfractions .15 .85
@item @code{CF} @tab Set on error.

@item @code{AH} @tab Status.

@item @code{AL} @tab The number of sectors transferred (only valid if CF
set for some BIOSes).
@end multitable

INT 13H, AH=03h writes disk sectors.

Input:

@multitable @columnfractions .15 .85
@item @code{AH} @tab 03h

@item @code{AL} @tab The number of sectors to write (must be non-zero).

@item @code{CH} @tab Low 8 bits of cylinder number.

@item @code{CL} @tab Sector number in bits 0-5, and high 2 bits of
cylinder number in bits 6-7.

@item @code{DH} @tab Head number.

@item @code{DL} @tab Drive number (bit 7 set for hard disk).

@item @code{ES:BX} @tab Data buffer.
@end multitable

Output:

@multitable @columnfractions .15 .85
@item @code{CF} @tab Set on error.

@item @code{AH} @tab Status.

@item @code{AL} @tab The number of sectors transferred (only valid if CF
set for some BIOSes).
@end multitable

INT 13H, AH=08h returns drive parameters. For systems predating the IBM
PC/AT, this call is only valid for hard disks.

Input:

@multitable @columnfractions .15 .85
@item @code{AH} @tab 08h

@item @code{DL} @tab Drive number (bit 7 set for hard disk).
@end multitable

Output:

@multitable @columnfractions .15 .85
@item @code{CF} @tab Set on error.

@item @code{AH} @tab 0.

@item @code{AL} @tab 0 on at least some BIOSes.

@item @code{BL} @tab Drive type (AT/PS2 floppies only).

@item @code{CH} @tab Low 8 bits of maximum cylinder number.

@item @code{CL} @tab Maximum sector number in bits 0-5, and high 2 bits
of maximum cylinder number in bits 6-7.

@item @code{DH} @tab Maxiumum head number.

@item @code{DL} @tab The number of drives.

@item @code{ES:DI} @tab Drive parameter table (floppies only).
@end multitable

GRUB does not use this call for floppies, but attempts to read the first
sector from the last head of the last cylinder to determine the maximum
head number and the maximum cylinder number.


@node LBA mode disk I/O
@subsection INT 13H, AH=4xh interrupt call

Real mode only. These functions are IBM/MS INT 13 Extensions to support
LBA mode. GRUB uses them if available so that it can read/write over 8GB
area.

INT 13, AH=41h checks if LBA is supported.

Input:

@multitable @columnfractions 0.15 0.85
@item @code{AH} @tab 41h.

@item @code{BX} @tab 55AAh.

@item @code{DL} @tab Drive number.
@end multitable

Output:

@multitable @columnfractions 0.15 0.85
@item @code{CF} @tab Set on error.

@item @code{AH} @tab Major version of extensions (01h for 1.x, 20h for
2.0 / EDD-1.0, 21h for 2.1 / EDD-1.1 and 30h for EDD-3.0) if successful,
otherwise 01h (the error code of @dfn{invalid function}).

@item @code{BX} @tab AA55h if installed.

@item @code{AL} @tab Internal use.

@item @code{CX} @tab API subset support bitmap (see below).

@item @code{DH} @tab Extension version.
@end multitable

The bitfields for the API subset support bitmap are:

@multitable @columnfractions 0.15 0.85
@item Bit(s) @tab Description

@item 0 @tab Extended disk access functions (AH=42h-44h, 47h, 48h)
supported.

@item 1 @tab Removable drive controller functions (AH=45h, 46h, 48h,
49h, INT 15H, AH=52h) supported.

@item 2 @tab Enhanced disk drive (EDD) functions (AH=48h, 4Eh)
supported.

@item 3-15 @tab Reserved (0).
@end multitable

INT 13, AH=42h reads sectors into memory.

Input:

@multitable @columnfractions .15 .85
@item @code{AH} @tab 42h.

@item @code{DL} @tab Drive number.

@item @code{DS:SI} @tab Disk Address Packet (see below).
@end multitable

Output:

@multitable @columnfractions .15 .85
@item @code{CF} @tab Set on error.

@item @code{AH} @tab 0 if successful, otherwise error code.
@end multitable

The format of @dfn{Disk Address Packet} is:

@multitable @columnfractions 0.15 0.15 0.7
@item Offset (hex) @tab Size (byte) @tab Description

@item 00 @tab 1 @tab 10h (The size of packet).

@item 01 @tab 1 @tab Reserved (0).

@item 02 @tab 2 @tab The number of blocks to transfer (max 007F for
Phoenix EDD).

@item 04 @tab 4 @tab Transfer buffer (SEGMENT:OFFSET).

@item 08 @tab 8 @tab Starting absolute block number.
@end multitable

INT 13, AH=43h writes disk sectors.

Input:

@multitable @columnfractions 0.15 0.85
@item @code{AH} @tab 43h.

@item @code{AL} @tab Write flags (In version 1.0 and 2.0, bit 0 is the
flag for @dfn{verify write} and other bits are reserved (0). In version
2.1, 00h and 01h indicates @dfn{write without verify}, and 02h indicates
@dfn{write with verify}.

@item @code{DL} @tab Drive number.

@item @code{DS:SI} @tab Disk Address Packet (see above).
@end multitable

Output:

@multitable @columnfractions 0.15 0.85
@item @code{CF} @tab Set on error.

@item @code{AH} @tab 0 if successful, otherwise error code.
@end multitable

INT 13, AH=48h returns drive parameters. GRUB only makes use of the
total number of sectors, and ignore the CHS information, because only
L-CHS makes sense. @xref{CHS Translation}, for more information.

Input:

@multitable @columnfractions 0.15 0.85
@item @code{AH} @tab 48h.

@item @code{DL} @tab Drive number.

@item @code{DS:SI} @tab Buffer for drive parameters (see below).
@end multitable

Output:

@multitable @columnfractions 0.15 0.85
@item @code{CF} @tab Set on error.

@item @code{AH} @tab 0 if successful, otherwise error code.
@end multitable

The format of drive parameters is:

@multitable @columnfractions 0.25 0.15 0.6
@item Offset (hex) @tab Size (byte) @tab Description

@item 00 @tab 2 @tab The size of buffer. Before calling this function,
set to the maximum buffer size, at least 1Ah. The size actually filled
is returned (1Ah for version 1.0, 1Eh for 2.x and 42h for 3.0).

@item 02 @tab 2 @tab Information flags (see below).

@item 04 @tab 4 @tab The number of physical cylinders.

@item 08 @tab 4 @tab The number of physical heads.

@item 0C @tab 4 @tab The number of physical sectors per track.

@item 10 @tab 8 @tab The total number of sectors.

@item 18 @tab 2 @tab The bytes per sector.

@comment Add an empty row for readability...
@item @tab @tab

@item @strong{v2.0 and later} @tab @tab

@item 1A @tab 4 @tab EDD configuration parameters.

@comment Add an empty row for readability...
@item @tab @tab

@item @strong{v3.0} @tab @tab

@item 1E @tab 2 @tab Signature BEDD to indicate presence of Device Path
information.

@item 20 @tab 1 @tab The length of Device Path information, including
signature and this byte (24h for version 3.0).

@item 21 @tab 3 @tab Reserved (0).

@item 24 @tab 4 @tab ASCIZ name of host bus (@samp{ISA} or @samp{PCI}).

@item 28 @tab 8 @tab ASCIZ name of interface type (@samp{ATA},
@samp{ATAPI}, @samp{SCSI}, @samp{USB}, @samp{1394} or @samp{FIBRE}).

@item 30 @tab 8 @tab Interface Path.

@item 38 @tab 8 @tab Device Path.

@item 40 @tab 1 @tab Reserved (0).

@item 41 @tab 1 @tab Checksum of bytes 1Eh-40h (2's complement of sum,
which makes the 8 bit sum of bytes 1Eh-41h equal to 00h).
@end multitable

The information flags are:

@multitable @columnfractions 0.15 0.85
@item Bit(s) @tab Description

@item 0 @tab DMA boundary errors handles transparently.

@item 1 @tab CHS information is valid.

@item 2 @tab Removable drive.

@item 3 @tab Write with verify supported.

@item 4 @tab Drive has change-line support (required if drive is
removable).

@item 5 @tab Drive can be locked (required if drive is removable).

@item 6 @tab CHS information set to maximum supported values, not
current media.

@item 7-15 @tab Reserved (0).
@end multitable


@node MBR
@section The structure of Master Boot Record

A Master Boot Record (@dfn{MBR}) is the sector at cylinder 0, head 0,
sector 1 of a hard disk. A MBR-like structure must be created in each of
partitions by the FDISK program.

At the completion of your system's Power On Self Test (@dfn{POST}), INT
19H is called. Usually INT 19 tries to read a boot sector from the first
floppy drive@footnote{Which drive is read first depends on your BIOS
settings.}. If a boot sector is found on the floppy disk, that boot
sector is read into memory at location 0000:7C00 and INT 19H jumps to
memory location 0000:7C00. However, if no boot sector is found on the
first floppy drive, INT 19H tries to read the MBR from the first hard
drive. If an MBR is found it is read into memory at location 0000:7C00
and INT 19H jumps to memory location 0000:7C00. The small program in the
MBR will atempt to locate an active (bootable) partition in its
partition table@footnote{This behavior is DOS MBR's, and GRUB ignores
the active flag.}. The small program in the boot sector must locate the
first part of the operating system's kernel loader program (or perhaps
the kernel itself or perhaps a @dfn{boot manager program}) and read that
into memory.

INT 19H is also called when the @key{CTRL}-@key{ALT}-@key{DEL} keys are
used. On most systems, @key{CTRL}-@key{ALT}-@key{DEL} causes an short
version of the POST to be executed before INT 19H is called.

The stuff is:

@table @asis
@item Offset 0000
The address where the MBR code starts.

@item Offset 01BE
The address where the partition table starts (@pxref{Partition table}).

@item Offset 01FE
The signature, AA55.
@end table

However, the first 62 bytes of a boot sector are known as the BIOS
Parameter Block (@dfn{BPB}), so GRUB cannot use these bytes for its own
purpose.

If an active partition is found, that partition's boot record is read
into 0000:7C00 and the MBR code jumps to 0000:7C00 with @code{SI}
pointing to the partition table entry that describes the partition being
booted. The boot record program uses this data to determine the drive
being booted from and the location of the partition on the disk.

The first byte of an active partition table entry is 80. This byte is
loaded into the @code{DL} register before INT 13H is called to read the
boot sector. When INT 13H is called, @code{DL} is the BIOS device
number. Because of this, the boot sector read by this MBR program can
only be read from BIOS device number 80 (the first hard disk). This is
one of the reasons why it is usually not possible to boot from any other
hard disk.


@node Partition table
@section The format of partition table

@menu
* Partition basics::            Overview the partition table.
* Partition types::             The list of the @dfn{type} code.
* Partition entry format::      The format of the table entry.
* Partition table rules::       Some basic rules for Partition table.
@end menu


@node Partition basics
@subsection Overview the partition table

FDISK creates all partition records (sectors). The primary purpose of a
partition record is to hold a partition table. The rules for how FDISK
works are unwritten but so far most FDISK programs seem to follow the
same basic idea.

First, all partition table records (sectors) have the same format. This
includes the partition table record at cylinder 0, head 0, sector 1 --
what is known as the Master Boot Record (MBR). The last 66 bytes of a
partition table record contain a partition table and a 2 byte
signature. The first 446 bytes of these sectors usually contain a
program but only the program in the MBR is ever executed (so extended
partition table records could contain something other than a program in
the first 466 bytes). For more information, see @ref{MBR}.

Second, extended partitions are @emph{nested} inside one another and
extended partition table records form a @dfn{linked list}. I will
attempt to show this in a diagram at @ref{Partition entry format}.

Each partition table entry is 16 bytes and contains things like the
start and end location of a partition in CHS, the start in LBA, the size
in sectors, the partition @dfn{type} and the @dfn{active} flag. Older
versions of FDISK may compute incorrect LBA or size values. And when
your computer boots itself, only the CHS fields of the partition table
entries are used (another reason LBA doesn't solve the >528MB
problem). The CHS fields in the partition tables are in L-CHS format,
see @ref{CHS Translation}.


@node Partition types
@subsection The list of the @dfn{type} code

There is no central clearing house to assign the codes used in the one
byte @dfn{type} field. But codes are assigned (or used) to define most
every type of file system that anyone has ever implemented on the x86
PC: 12-bit FAT, 16-bit FAT, HPFS, NTFS, etc. Plus, an extended partition
also has a unique type code.

In the FDISK program @samp{sfdisk}, the following list is assumed:

@table @asis
@item 00
Empty

@item 01
DOS 12-bit FAT

@item 02
XENIX /

@item 03
XENIX /usr

@item 04
DOS 16-bit FAT <32M

@item 05
DOS Extended

@item 06
DOS 16-bit FAT >=32M

@item 07
HPFS / NTFS

@item 08
AIX boot or SplitDrive

@item 09
AIX data or Coherent

@item 0A
OS/2 Boot Manager

@item 0B
Windows95 FAT32

@item 0C
Windows95 FAT32 (LBA)

@item 0E
Windows95 FAT16 (LBA)

@item 0F
Windows95 Extended (LBA)

@item 10
OPUS

@item 11
Hidden DOS FAT12

@item 12
Compaq diagnostics

@item 14
Hidden DOS FAT16

@item 16
Hidden DOS FAT16 (big)

@item 17
Hidden HPFS/NTFS

@item 18
AST Windows swapfile

@item 24
NEC DOS

@item 3C
PartitionMagic recovery

@item 40
Venix 80286

@item 41
Linux/MINIX (sharing disk with DRDOS)

@item 42
SFS or Linux swap (sharing disk with DRDOS)

@item 43
Linux native (sharing disk with DRDOS)

@item 50
DM (disk manager)

@item 51
DM6 Aux1 (or Novell)

@item 52
CP/M or Microsoft SysV/AT

@item 53
DM6 Aux3

@item 54
DM6

@item 55
EZ-Drive (disk manager)

@item 56
Golden Bow (disk manager)

@item 5C
Priam Edisk (disk manager)

@item 61
SpeedStor

@item 63
GNU Hurd or Mach or Sys V/386 (such as ISC UNIX)@footnote{But the reason
why they decided that 63 means GNU Hurd is not known. Do not use 63 for
GNU Hurd.}

@item 64
Novell Netware 286

@item 65
Novell Netware 386

@item 70
DiskSecure Multi-Boot

@item 75
PC/IX

@item 77
QNX4.x

@item 78
QNX4.x 2nd part

@item 79
QNX4.x 3rd part

@item 80
MINIX until 1.4a

@item 81
MINIX / old Linux

@item 82
Linux swap

@item 83
Linux native@footnote{This is not true. Use 83 for ext2fs even if the
owner OS is GNU/Hurd.}

@item 84
OS/2 hidden C: drive

@item 85
Linux extended

@item 86
NTFS volume set

@item 87
NTFS volume set

@item 93
Amoeba

@item 94
Amoeba BBT

@item A0
IBM Thinkpad hibernatoin

@item A5
BSD/386

@item A7
NeXTSTEP 486

@item B7
BSDI fs

@item B8
BSDI swap

@item C1
DRDOS/sec (FAT-12)

@item C4
DRDOS/sec (FAT-16, < 32M)

@item C6
DRDOS/sec (FAT-16, >= 32M)

@item C7
Syrinx

@item DB
CP/M or Concurrent CP/M or Concurrent DOS or CTOS

@item E1
DOS access or SpeedStor 12-bit FAT extended partition

@item E3
DOS R/O or SpeedStor

@item E4
SpeedStor 16-bit FAT extended partition < 1024 cyl.

@item F1
SpeedStor

@item F2
DOS 3.3+ secondary

@item F4
SpeedStor large partition

@item FE
SpeedStor >1024 cyl. or LANstep

@item FF
Xenix Bad Block Table
@end table

@node Partition entry format
@subsection The format of the table entry

The 16 bytes of a partition table entry are used as follows:

@example
@group
    +--- Bit 7 is the active partition flag, bits 6-0 are zero.
    |
    |    +--- Starting CHS in INT 13 call format.
    |    |
    |    |      +--- Partition type byte.
    |    |      |
    |    |      |    +--- Ending CHS in INT 13 call format.
    |    |      |    |
    |    |      |    |        +-- Starting LBA.
    |    |      |    |        |
    |    |      |    |        |       +-- Size in sectors.
    |    |      |    |        |       |
    v <--+--->  v <--+-->     v       v

   0  1  2  3  4  5  6  7  8 9 A B  C D E F
   DH DL CH CL TB DL CH CL LBA..... SIZE....

   80 01 01 00 06 0e be 94 3e000000 0c610900  1st entry

   00 00 81 95 05 0e fe 7d 4a610900 724e0300  2nd entry

   00 00 00 00 00 00 00 00 00000000 00000000  3rd entry

   00 00 00 00 00 00 00 00 00000000 00000000  4th entry
@end group
@end example

Bytes 0-3 are used by the small program in the Master Boot Record to
read the first sector of an active partition into memory. The @dfn{DH},
@dfn{DL}, @dfn{CH} and @dfn{CL} above show which x86 register is loaded
when the MBR program calls INT 13H AH=02h to read the active partition's
boot sector. For more information, see @ref{MBR}.

These entries define the following partitions:

@enumerate
@item
The first partition, a primary partition DOS FAT, starts at CHS 0H,1H,1H
(LBA 3EH) and ends at CHS 294H,EH,3EH with a size of 9610CH sectors.

@item
The second partition, an extended partition, starts at CHS 295H,0H,1H
(LBA 9614AH) and ends at CHS 37DH,EH,3EH with a size of 34E72H sectors.

@item
The third and fourth table entries are unused.
@end enumerate


@node Partition table rules
@subsection Some basic rules for Partition table.

Keep in mind that there are @emph{no} written rules and @emph{no}
industry standards on how FDISK should work but here are some basic
rules that seem to be followed by most versions of FDISK:

@enumerate
@item
In the MBR there can be 0-4 @dfn{primary} partitions, OR, 0-3 primary
partitions and 0-1 extended partition entry.

@item
In an extended partition there can be 0-1 @dfn{secondary} partition
entries and 0-1 extended partition entries.

@item
Only 1 primary partition in the MBR can be marked @dfn{active} at any
given time.

@item
In most versions of FDISK, the first sector of a partition will be
aligned such that it is at head 0, sector 1 of a cylinder. This means
that there may be unused sectors on the track(s) prior to the first
sector of a partition and that there may be unused sectors following a
partition table sector.

For example, most new versions of FDISK start the first partition
(primary or extended) at cylinder 0, head 1, sector 0. This leaves the
sectors at cylinder 0, head 0, sectors 2...n as unused sectors. This
same layout may be seen on the first track of an extended partition.
See example 2 below.

Also note that software drivers like Ontrack's Disk Manager depend on
these unused sectors because these drivers will @dfn{hide} their code
there (in cylinder 0, head 0, sectors 2...n). This is also a good place
for boot sector virus programs to hang out.

@item
The partition table entries (slots) can be used in any order. Some
versions of FDISK fill the table from the bottom up and some versions of
FDISK fill the table from the top down. Deleting a partition can leave
an unused entry (slot) in the middle of a table.

@item
And then there is the @dfn{hack} that some newer OS's (OS/2 and Linux)
use in order to place a partition spanning or passed cylinder 1024 on a
system that does not have a CHS translating BIOS. These systems create a
partition table entry with the partition's starting and ending CHS
information set to all FFH. The starting and ending LBA information is
used to describe the location of the partition. The LBA can be converted
back to a CHS --- most likely a CHS with more than 1024 cylinders. Since
such a CHS can't be used by the system BIOS, these partitions can not be
booted or accessed until the OS's kernel and hard disk device drivers
are loaded.  It is not known if the systems using this @dfn{hack} follow
the same rules for the creation of these type of partitions.
@end enumerate

There are @emph{no} written rules as to how an OS scans the partition
table entries so each OS can have a different method. For DOS, this
means that different versions could assign different drive letters to
the same FAT file system partitions.


@node Filesystem interface
@section The generic interface for the fs code

For any particular partition, it is presumed that only one of the
@dfn{normal} filesystems such as FAT, FFS, or ext2fs can be used, so
there is a switch table managed by the functions in
@file{disk_io.c}. The notation is that you can only @dfn{mount} one at a
time.

The blocklist filesystem has a special place in the system. In addition
to the @dfn{normal} filesystem (or even without one mounted), you can
access disk blocks directly (in the indicated partition) via the
blocklist notation. Using the blocklist filesystem doesn't effect any
other filesystem mounts.

The variables which can be read by the filesystem backend are:

@vtable @code
@item current_drive
Contain the current BIOS drive number (numbered from 0, if a floppy,
and numbered from 0x80, if a hard disk).

@item current_slice
Contain the current partition length, in sectors.

@item print_possibilities
True when the @code{dir} function should print the possible completions
of a file, and false when it should try to actually open a file of that
name.

@item FSYS_BUF
Point to a filesystem buffer which is 32K in size, to use in any way
which the filesystem backend desires.
@end vtable

The variables which need to be written by a filesystem backend are:

@vtable @code
@item filepos
Should be the current position in the file.

@strong{Caution:} the value of @var{filepos} can be changed out from
under the filesystem code in the current implementation. Don't depend on
it being the same for later calls into the back-end code!

@item filemax
Should be the length of the file.

@item disk_read_func
Should be set to the value of @samp{disk_read_hook} @emph{only} during
reading of data for the file, not any other fs data, inodes, FAT tables,
whatever, then set to @code{NULL} at all other times (it will be
@code{NULL} by default). If this isn't done corrently, then the
@command{testload} and @command{install} commands won't work
correctly.
@end vtable

The functions expected to be used by the filesystem backend are:

@ftable @code
@item devread
Only read sectors from within a partition. Sector 0 is the first sector
in the partition.

@item grub_read
If the backend uses the blocklist code (like the FAT filesystem backend
does), then @code{grub_read} can be used, after setting @var{block_file}
to 1.
@end ftable

The functions expected to be defined by the filesystem backend are
described at least moderately in the file @file{filesys.h}. Their usage
is fairly evident from their use in the functions in @file{disk_io.c},
look for the use of the @var{fsys_table} array.

@strong{Caution:} The semantics are such that then @samp{mount}ing the
filesystem, presume the filesystem buffer @var{FSYS_BUF} is corrupted,
and (re-)load all important contents. When opening and reading a file,
presume that the data from the @samp{mount} is available, and doesn't
get corrupted by the open/read (i.e. multiple opens and/or reads will be
done with only one mount if in the same filesystem).


@node Index
@unnumbered Index

@c Currently, we use only the Concept Index.
@printindex cp

@contents
@bye