user.tex

linux 的引导程序源码The Microsoft&reg Windows&reg Software Development Kit (SDK) provides the documentation

    \cline{2-3}
    & \multicolumn{2}{|l|}{MBR\hbox to 1.3in{\hfill\tt /dev/hda~}} \\
    \cline{3-3}
    & & MS-DOS\hfill\tt /dev/hda1 \\
    \cline{3-3}
    $\rightarrow$ & & Linux {\tt /}\hfill\tt /dev/hda2 \\
    \cline{2-3}
  \end{tabular}
$$
%%endskip
%:\begin{verbatim}
%;    +--------------------------+
%;    | MBR            /dev/hda  |
%;    | +------------------------|
%;    | | MS-DOS       /dev/hda1 |
%;    | |------------------------|
%;--> | | Linux /      /dev/hda2 |
%;    +--------------------------+
%:\end{verbatim}

In this example, the \craw{boot} variable could be omitted, because
the boot sector is on the root partition.


\subsubsection{One disk, Linux on a logical partition}

If no primary partition is available for Linux, but at least one logical
partition of an extended partition on the first hard disk contains a
Linux file system, the \LILO\ boot sector should be stored in the partition
sector of the extended partition and it should be booted by the original
master boot record or by a program like \name{BOOTACTV}.

%%beginskip
$$
  \begin{tabular}{r|c|c|c|}
    \cline{2-4}
    & \multicolumn{3}{|l|}{MBR\hbox to 1.3in{\hfill\tt /dev/hda~}} \\
    \cline{3-4}
    & & \multicolumn{2}{|l|}{MS-DOS\hfill\tt /dev/hda1} \\
    \cline{3-4}
    $\rightarrow$ & & \multicolumn{2}{|l|}{Extended\hfill\tt /dev/hda2} \\
    \cline{4-4}
    & & & Linux\hfill\tt /dev/hda5 \\
    \cline{4-4}
    & & & $\ldots$\hfill\tt /dev/hda6 \\
    \cline{2-4}
  \end{tabular}
$$
%%endskip
%:\begin{verbatim}
%;    +--------------------------+
%;    | MBR            /dev/hda  |
%;    | +------------------------|
%;    | | MS-DOS       /dev/hda1 |
%;    | |------------------------|
%;--> | | Extended     /dev/hda2 |
%;    | | +----------------------|
%;    | | | Linux      /dev/hda5 |
%;    | | |----------------------|
%;    | | | ...        /dev/hda6 |
%;    +--------------------------+
%:\end{verbatim}

Because many disk partitioning programs refuse to make an extended
partition (in our example \path{/dev/hda2}) active, you might have to
use \name{activate}, which comes with the \LILO\ distribution.

OS/2 BootManager should be able to boot \LILO\ boot sectors from logical
partitions. The installation on the extended partition itself is not
necessary in this case.


\subsubsection{Two disks, Linux (at least partially) on the first disk}

This case is equivalent to the configurations where only one disk
is in the system. The Linux boot sector resides on the first hard
disk and the second disk is used later in the boot process.

Only the location of the boot sector matters -- everything
else (\path{/boot/boot.b},
\path{/boot/map}, the root file system, a swap partition, other
Linux file systems, etc.) can be located anywhere on the second disk,
provided that the constraints described in section \ref{broken} are met.


\subsubsection{Two disks, Linux on second disk, first disk has an extended
  partition}

If there is no Linux partition on the first disk, but there is an
extended partition, the \LILO\ boot sector can be stored in the partition
sector of the extended partition and it should be booted by the original
master boot record or by a program like \name{BOOTACTV}.

%%beginskip
$$
  \begin{tabular}{r|c|c|c|c|c|c|}
    \multicolumn{1}{r}{}
    & \multicolumn{3}{c}{\bf First disk} &
      \multicolumn{1}{r}{\qquad}
      & \multicolumn{2}{c}{\bf Second disk} \\
    \cline{2-4}\cline{6-7}
    & \multicolumn{3}{|l|}{MBR\hbox to 1.3in{\hfill\tt /dev/hda~}} &
      & \multicolumn{2}{|l|}{MBR\hbox to 1.3in{\hfill\tt /dev/hdb~}} \\
    \cline{3-4}\cline{7-7}
    & & \multicolumn{2}{|l|}{MS-DOS\hfill\tt /dev/hda1} &
      & & Linux\hfill\tt /dev/hdb1 \\
    \cline{3-4}\cline{7-7}
    $\rightarrow$ & & \multicolumn{2}{|l|}{Extended\hfill\tt /dev/hda2} &
      & & $\ldots$\hfill\tt /dev/hdb2 \\
    \cline{4-4}
    & & & $\ldots$\hfill\tt /dev/hda5 & & & \\
    \cline{4-4}
    & & & $\ldots$\hfill\tt /dev/hda6 & & & \\
    \cline{2-4}\cline{6-7}
  \end{tabular}
$$
%%endskip
%:\begin{verbatim}
%4             FIRST DISK                     SECOND DISK
%4    +--------------------------+    +--------------------------+
%4    | MBR            /dev/hda  |    | MBR            /dev/hdb  |
%4    | +------------------------|    | +------------------------|
%4    | | MS-DOS       /dev/hda1 |    | | Linux        /dev/hdb1 |
%4    | |------------------------|    | |------------------------|
%4--> | | Extended     /dev/hda2 |    | | ...          /dev/hdb2 |
%4    | | +----------------------|    | |                        |
%4    | | | ...        /dev/hda5 |    | |                        |
%4    | | |----------------------|    | |                        |
%4    | | | ...        /dev/hda6 |    | |                        |
%4    +--------------------------+    +--------------------------+
%:\end{verbatim}

The program \name{activate}, that accompanies \LILO, may have to be used to
set the active marker on an extended partition, because MS-DOS' \name{FDISK}
and some older version of Linux \name{fdisk} refuse to do that. (Which is
generally a good idea.)


\subsubsection{Two disks, Linux on second disk, first disk has no
  extended partition}

If there is neither a Linux partition nor an extended partition on the first
disk, then there's only one place left, where a \LILO\ boot sector could be
stored: the master boot record.

In this configuration, \LILO\ is responsible for booting all other operating
systems too.

%%beginskip
$$
  \begin{tabular}{r|c|c|c|c|c|}
    \multicolumn{1}{r}{}
    & \multicolumn{2}{c}{\bf First disk} &
      \multicolumn{1}{r}{\qquad}
      & \multicolumn{2}{c}{\bf Second disk} \\
    \cline{2-3}\cline{5-6}
    $\rightarrow$ & \multicolumn{2}{|l|}{MBR\hbox to 1.3in{
      \hfill\tt /dev/hda~}} &
      & \multicolumn{2}{|l|}{MBR\hbox to 1.3in{\hfill\tt /dev/hdb~}} \\
    \cline{3-3}\cline{6-6}
    & & MS-DOS\hfill\tt /dev/hda1 &
      & & Linux\hfill\tt /dev/hdb1 \\
    \cline{3-3}\cline{6-6}
    & & $\ldots$\hfill\tt /dev/hda2 &
      & & $\ldots$\hfill\tt /dev/hdb2 \\
    \cline{2-3}\cline{5-6}
  \end{tabular}
$$
%%endskip
%:\begin{verbatim}
%4             FIRST DISK                     SECOND DISK
%4    +--------------------------+    +--------------------------+
%4--> | MBR            /dev/hda  |    | MBR            /dev/hdb  |
%4    | +------------------------|    | +------------------------|
%4    | | MS-DOS       /dev/hda1 |    | | Linux        /dev/hdb1 |
%4    | |------------------------|    | |------------------------|
%4    | | ...          /dev/hda2 |    | | ...          /dev/hdb2 |
%4    +--------------------------+    +--------------------------+
%:\end{verbatim}

You should back up your old MBR before installing \LILO\ and verify that
\LILO\ is able to boot your other operating system(s) before relying on
this approach.

The line \verb"boot = /dev/hda2" in \path{/etc/lilo.conf} would have to be
changed to \verb"boot = /dev/hda" in this example.


\subsubsection{More than two disks}

On systems with more than two disks, typically only the first two can
be accessed. The configuration choices are therefore the same as with
two disks.

When attempting to access one of the extra disks, \LILO\ displays a
warning message (\raw{Warning: BIOS drive 0x\meta{number} may not be
accessible}) but does not abort. This is done in order to allow
the lucky few whose BIOS (or controller-BIOS) does support more than
two drives to make use of this feature. By all others, this warning
should be considered a fatal error.

Note that the two disks restriction is only imposed by the BIOS.
Linux normally has no problems using all disks once it is booted.


\subsubsection{\path{/boot} on a DOS partition}
\label{viados}

Recent kernels support all the functions \LILO\ needs to map files also
on MS-DOS (or UMSDOS) file systems. Since DOS partitions tend to occupy
exactly the places where BIOS restrictions (see section \ref{broken})
are invisible, they're an ideal location for \path{/boot} if the native
Linux file systems can't be used because of BIOS problems.

In order to accomplish this, the DOS partition is mounted read-write,
a directory (e.g. \path{/dos/linux}) is created, all files from
\path{/boot} are moved to that directory, \path{/boot} is replaced by
a symbolic link to it, the kernels are also moved to the new directory,
their new location is recorded in \path{/etc/lilo.conf}, and finally
\path{/sbin/lilo} is run.

From then on, new kernels must always be copied into that directory on
the DOS partition before running \path{/sbin/lilo}, e.g. when recompiling
a kernel, the standard procedure changes from

\begin{verbatim}
# make zlilo
\end{verbatim}

to

\begin{verbatim}
# make zImage
# mv /dos/linux/vmlinuz /dos/linux/vmlinuz.old
# mv arch/i386/boot/zImage /dos/linux/vmlinuz
# /sbin/lilo
\end{verbatim}

\emphasize{WARNING:} De-fragmenting such a DOS partition is likely to
make Linux or even the whole system unbootable. Therefore, the DOS
partition should either not be de-fragmented, or a Linux boot disk
should be prepared (and tested) to bring up Linux and to run
\path{/sbin/lilo} after the de-fragmentation.\footnote{Setting the
``system'' attribute from DOS on the critical files (e.g. everything
in \path{C:$\backslash$LINUX}) may help to protect them from being
rearranged. However, the boot floppy should still be ready, just in case.}


\newpage
\section{The boot prompt}
\label{cmdopt}

Immediately after it's loaded, \LILO\ checks whether one of the following
is happening:

\begin{itemize}
  \item any of the \key{Shift}, \key{Control} or \key{Alt} keys is
    pressed.
  \item \key{CapsLock} or \key{ScrollLock} is set.
\end{itemize}

If this is the case, \LILO\ displays the \verb"boot:" prompt and waits for
the name of a boot image (i.e. Linux kernel or other operating system).
Otherwise, it boots the default boot image\footnote{%
The default boot image is either the first boot image, the image specified
with the \craw{default} variable, or the image that
has been selected at the boot prompt.}
or -- if a
delay has been specified -- waits for one of the listed activities
until that amount of time has passed.

At the boot prompt, the name of the image to boot can be entered. Typing
errors can be corrected with \key{BackSpace}, \key{Delete},
\key{Ctrl U} and \key{Ctrl X}. A list of known images can be obtained by
pressing \key{?} or \key{Tab}.

If \key{Enter} is pressed and no file name has been entered, the default
image is booted.


\subsection{Boot command-line options}

\LILO\ is also able to pass command-line options to the kernel.
Command-line options are words that follow the name of the boot
image and that are separated by spaces.

Example:
\begin{verbatim}
boot: linux single root=200
\end{verbatim}

This document only gives an overview of boot options. Please consult Paul
Gortmaker's BootPrompt-HOWTO for a more complete and more up to date list.
You can get it from
\path{ftp://sunsite.unc.edu/pub/Linux/docs/HOWTO/BootPrompt-HOWTO.gz}
or from one of the many mirror sites.


\subsubsection{Standard options}

Recent kernels recognize a large number of options, among them are \raw{debug},
\raw{no387}, \raw{no-hlt},
\raw{ramdisk=\meta{size}}, \raw{reserve=\meta{base},\meta{size},$\ldots$},
\raw{root=\meta{device}}, \raw{ro}, and \raw{rw}.
All current \name{init} programs also recognize the option \raw{single}.
The options \raw{lock} and \raw{vga} are processed by the boot loader itself.
Boot command-line options are always case-sensitive.

\raw{single} boots the system in single-user mode. This bypasses most
system initialization procedures and directly starts a root shell on the
console. Multi-user mode can typically be entered by exiting the single-user
shell or by rebooting.

\raw{root=\meta{device}} changes the root device. This overrides
settings that may have been made in the boot image and on the
\LILO\ command line. \meta{device} is either the hexadecimal device number
%%beginskip
\footnote{%
This is a list of device numbers of some frequently used devices: \\
\begin{tabular}{lllllll}
\qquad
& {\tt /dev/fd0} & 200\qquad & \tt /dev/hda1 & 301\qquad & /dev/sda1 & 801 \\
& {\tt /dev/fd1} & 201 & \tt /dev/hda2 & 302 & /dev/sda2 & 802 \\
& \multicolumn{2}{c}{$\cdots$} & \multicolumn{2}{c}{$\cdots$} &
\multicolumn{2}{c}{$\cdots$} \\
& & & \tt /dev/hdb1 & 341 & /dev/sdb1 & 811 \\
& & & \tt /dev/hdb2 & 342 & /dev/sdb2 & 812 \\
& \multicolumn{2}{c}{} & \multicolumn{2}{c}{$\cdots$} &
\multicolumn{2}{c}{$\cdots$} \\
\end{tabular}}
%%endskip
or the full path name of the device, e.g. \verb"/dev/hda3".%
\footnote{The device names are hard-coded in the kernel. Therefore, only
the ``standard'' names are supported and some less common devices may not
be recognized. In those cases, only numbers can be used.}