From: [email protected] (I Reid)
Subject: Beginner's Guide: Installation
Date: 2 Apr 92 22:18:33 GMT

==============================================================
This is an updated release of my guide to installing Linux. It mostly corrects
one or two errata but there are one or two new bits. It is being reposted
because the original is disappearing from many news feeds and I think there is
still a demand for it.

A plea for errors
=================

I've been fortunate.... for the most part installation of Linux has been easy
and completely trouble free for me. If you have had problems with installation
then please tell me what they were, where they were, what was on the screen,
how you fixed them etc etc.

The other thing to be said is....

I will be trying to keep this document up to date as new developments occur
(e.g. the new fdisk command) but there is probably not too much to be done
here. Is there any aspect of Linux which you feel isn't well covered in the
documentation available? Please let me know and I'll see if I want to work on
it (well, I have to enjoy myself doing it!).

Iain [[email protected]]

p.s. Mail on any subject is always welcome and I'll try to help with specific
problems if I can. Please check the FAQ and the other documentation if
possible 'cos many of the common problems are covered there.

p.p.s A rough contents list is

Section 1: First tests with floppies, rawrite etc
Section 2: Installing to hard disk (including non destructive partitioning)
==============================================================

A Beginner's Guide to Linux
---------------------------
Purpose
=======

This document is intended to be a complete beginner's guide to the
installation and running of Linux (a free Unix clone) and, as such,
tries to make as few assumptions as possible. However, some familiarity
with Unix at an "ordinary user" level (i.e. no sysadmin experience) is
assumed/required but I'll try to keep to a "hand-holding" level for anything
out of the ordinary. I don't intend this as a technical reference
guide, merely something to get you started and keep you clear of trouble!

I hope to cover everything you need to get "up and running"
but if not please let me know so that I can change things. Similarly,
if there is anything you think is a "must know", a useful tip, or a
waste of space please let me know so I can do something about it.
Email is best 'cos I can't guarantee to see all the stuff posted to
alt/comp.os.linux.

Hope this helps and happy hacking,

Iain Reid [[email protected]]

This document is quite big (and growing all the time). Despite this you might
find it helpful to print it out. I appreciate that this is damaging to
rainforests, aardvarks etc but then again, do you really want to be stuck in
the middle of something tricky with nothing to refer to if everything starts
to go wrong?


Errors: Where possible I have included details of potential errors and
troublespots (with solutions where known). These are marked as

[Error n]

where n is a number. A description of the error and solution etc can
be found at the end of each section. I badly need feedback on these
(list of other problems etc 'cos this bit is a bit sparse at the moment)

==============================================================
Required abilities
==================

To get Linux and keep up to date with new releases you will need
to be able to get files from your nearest Linux archive site,
uncompress some of them and transfer them to a DOS filesystem. If you
don't know how to do this you should ask someone at your local site to
show you because the exact method used is highly dependant on your
local hardware and software.

All the files and utilities mentioned should be available from the
Linux archive sites.
==============================================================

Ok, ready to start.... First and foremost....

BACKUP EVERYTHING!!!!!!

I don't expect you to have problems but if you do then why risk losing
everything? You'll need to do this before installing to hard disk
anyway so why not do it now?

Section 1. Preliminary testing (otherwise known as 10 steps to happiness)
=========================================================================

Q. Will Linux run on my hardware?
A. If you have a 386Sx or above with 2 or more megabytes of memory
then the chances are that Linux will run but the easiest way to
check is to try running from floppy.

You will need - the current bootimage (bootimage-0.95a.Z)
- the current rootimage (rootimage-0.95a.Z)
- rawrite.exe (a utility to write the boot & root images
to floppy from within DOS)
- Two formatted floppies (2 x 1.2 Mb or 2 x 1.44 Mb but
both must be the same capacity)

It's up to you whether you use 1.2 Mb or 1.44 Mb disks but you
should use whichever type you can boot your machine from. For example,
my disk drives (under DOS) are

A: 3.5" 1.44 Mb floppy drive [ bootable ]
B: 5.25" 1.2 Mb floppy drive [ NOT bootable]
C: 110 Mb IDE Hard Disk [ bootable ]

In my BIOS setup program I can choose to boot from drive A or C so I
would choose 1.44 Mb floppy disks.

What to do:

1. Get the three files listed above from your nearest archive site.
2. Uncompress the boot and root images and transfer them plus
rawrite.exe to DOS
3. Run rawrite.exe and you should see something like the following
(type bootimage or rootimage where prompted and use whichever
drive you can boot from)

c:\> rawrite
Rawrite 1.2 - Write disk file to raw floppy diskette

Enter source file name: bootimage
Enter destination drive: a
Please insert a formatted diskette into drive A: and press -ENTER-:
Number of sectors per track for this disk is 18
Writing Image to drive A:. Press ^C to abort.
Track: nn Head: n Sectors: nn

4. Save bootimage (about 200 Kbytes) and rootimage (about 1100
Kbytes) to separate disks using rawrite as above. You should
see some information about tracks and sectors and the numbers
should keep changing until it is finished and the command
prompt comes back. Bootimage is fairly quick to transfer but
rootimage is sloooooow so leave it running and have a quick (3 minute)
cup of tea or something.

5. Reboot your machine with the boot floppy (i.e. the one with
bootimage on it) in your floppy drive. If all goes well your
floppy should whir a bit and the message "Loading...." should
appear. The floppy will whir and click a bit more and some more
dots should appear. On my machine this takes about 6 or 7
seconds in total... it should be fairly quick. The following
message should appear

Press to see SVGA-modes available or any other key to continue.

and you may see something like EGAc in the top right corner of
your screen. Things get slightly hazy for me here because Linux
and my video card disagree about things but you can have a
quick explore here if you wish. I have no idea what is supposed
to happen, I just know that it doesn't happen for me :-(. Press
return or any other key as instructed and you should see all the text
on your screen shrink to half size (VGA 50 row mode) and something like
this

8 virtual consoles
4 pty's
Partition table ok.
Insert root floppy and press ENTER

[Error 1]

6. Put the root floppy (i.e. the one with the rootimage on it) in
the drive and press enter. You should see something like the
following although the numbers may be different)

44/1200 free blocks.
265/400 free inodes.
721 buffers = 738304 bytes buffer space
Free mem: 2621440 bytes
All systems functioning.


There will be a slight pause (2-3 seconds here) and then at the bottom the
screen you should see


Welcome to Linux 0.95a.

Login as 'root' to install. No password.

(none) login:

7. login as root and you should see

==============================================
Welcome to Linux version 0.95a
==============================================

See the files in the directory
/INSTALL for installation info
and scripts. - Jim Winstead
------------------------------
Questions? You should talk to
[email protected]

==============================================

#

8. Congratulations! Linux is now running (from floppy admittedly)
on your system. The '#' is your command prompt so you can have
a play now although there isn't very much on the root floppy.
It is worthwhile trying a few ls's (list files) and cd's (change
directory) and you should definitely read some of the documents in
INSTALL

[Error 2]

# cd /INSTALL
# ls

CHANGES-0.95a INSTALL-0.95a LDS-Standard LDSS.Tree
WARNING install mkdev mktree

# more INSTALL-0.95a

(q get you out of more.... make sure it's lower case)

Don't worry about your floppy drive bursting into life at odd
times (usually at least once every 30 seconds or so (see "sync"
later on)).... this is normal.

[ You can run entirely from floppy but as you can see it really isn't
much fun so, if you like what you see, onwards ever onwards, otherwise
please think again... Linux really is gorgeous to use once you get it
set up. Still not convinced? Oh well, bye then ]

9. You now know that Linux will run on your processor so the next
question is "Will it run on my hard disk?". To test this use
the fdisk command

# fdisk

Disk 0:
/dev/hda1: 61516 blocks active 16-bit DOS (>=32M)
/dev/hda2: 48090 blocks unknown partition type 0x80

The messages you get will vary according to what you have set
up on your hard disk but, if it tells you about any partitions
(I have 2 as shown), then Linux can read at least part of your
hard drive. This is good. If it barfs then things don't look
so good and, if you think Linux should run on your hardware
(see the info sheet at the archive sites), you should tell
Linus Torvalds (the creator of Linux) 'cos he needs to know
such things. Please be very sure that it is supposed to run on
your hardware before sending off that mailer tho' 🙂

10. Sync. This is an initially mysterious command which you must
ALWAYS, ALWAYS, ALWAYS run before you log out. What does it
do? Well, in common with many unix systems, Linux stores some
information in buffers in memory (remember those 721 buffers
at bootup?). Some of it is to do with what's where on the disk
so rebooting the machine without syncing (writing those
buffers to disk) could leave you in a BIG mess. That's why you
must ALWAYS sync. 'Nuff said?

# sync

Well, you've got this far so that means that 1. you can run Linux, 2.
your hard disk and Linux are on talking terms and 3. You're ready to
install to hard disk 🙂

p.s. type "exit" to leave ash.
==============================================================

Errors and solutions (for section 1)
====================================

1. Just after the "4 pty's" message appears and before the "Partition table
ok." your hard disk may make a descending tone 2-3 seconds long and the
message HD timeout appears. This is a recognised problem caused by failure
to reset the hard disk controller. It is not damaging and you can deal with
it once you are up and running. You have to tolerate it until you have
recompiled the kernel. Unfortunately you need the "damaged" kernel to build
a linux filesystem which is why this bug still exists.

2. There is a bug in ash (the shell supplied on rootimage 0.95a) in that it
doesn't seem to recognise delete when used on the first virtual console.
This is being worked on and will hopefully be fixed by the next release. A
solution is to run on the second virtual console. Press Alt-F2 to get it.
You should see a login prompt. Login as root again and check that you can
delete.

==============================================================

Section II: Installation to Hard Disk
=====================================

You've tried the basic tests in section I; you liked what you saw
and you want to install to hard disk. Well, there are a few things you
should know before you start...


Linux uses a Minix file system which is comaptible with Minix but not
with anything else (e.g. Dos). You will, therefore, have to give it its own
area on your hard disk and keep that area separate from everything else.
This is done by creating a separate PARTITION (area of disk) for
Linux. This is quite straightforward but is a fairly major thing so
I'll start with a little theory so that you know what you're doing and
why...

... There are two types of partition that Linux can use- primary and
extended. At the moment, however, the extended partition code is still
being written and tested so, in the interests of simplicity, I will
only cover primary partitions.


Q. What is a primary partition?
A. The first sector (512 bytes) of a hard disk contains the primary
boot record. This is a record of various pieces of information
about the physical characteristics of the hard disk, code for use
during bootup and information about partitions. There is a fixed
amount of space allocated for the partition information (enough for 4
entries) and so there is a limit to the number of divisions that a
hard disk may have (4). A primary partition is one of these 4 major
divisions of a hard disk.

This limit of 4 partitions on a disk was rather restrictive and so
secondary or extended partitions were introduced to bypass it. As
with most add ons, however, they are not as "clean" from a
programming point of view and special care is needed when dealing
with them.


Q. How do you create a primary partition?
A. Simple. You just edit the partition table in the primary boot
record to reflect your needs. Reformat any areas of the disk
affected by the partitioning and that's about it (in theory
anyway!)


Armed with that extensive tour of the intricacies of hard disk
partitions you are now ready for the main event. To make things
clearer I will use the example of a machine with one hard drive, a 107
Mb IDE drive which at the start has one partition (all disks must have
at least one) but which I want to divide into a 47 Mb Linux partition and
a 60 Mb Dos partition. The actual figures involved are irrelevant and
are meant as an example, not a guide.

Q. How much disk space should I give to Linux?
A. How high is the sky? This is very much a personal decision based on
your computing interests, whether you intend to run in single or
multi-user mode, whether you are planning to have Usenet stuff on your
disk etc etc. I can't give you a figure that is guaranteed to be what
you'll need. I doubt I could even come close but, even with a couple
of versions of the kernel compiled up, lots of source sitting around,
lots of big text files, lots of Usenet stuff, lots of utilities etc etc I
haven't gone beyond about 18 Mb (+ 8 Mb for swapping). At the moment I'm using
about 18 Mb and that's only because I've just de-archived my entire Linux
partition and there are lots of duplicate files to throw out.

Note: For serious kernel hacking, you will need an additional ~3Mb to
maintain the kernel source tree under RCS + source files and object
code created by the build.


Structure of the rest of this section
=====================================

There are a number of topics which you may want/need to read about
before partitioning. To keep the main bit reasonably brief I have put these
at the end of the SECTION as appendices. These are:

Appendix A: The DOS bootable floppy

Appendix B: Rootset.c

Appendix C: Non destructive partitioning (courtesy of Drew Eckhardt)


The rest is straightforward - Disk partitioning using pfdisk
- Installing Linux to hard drive
- Booting entirely from hard disk (not yet)
- Common problems and solutions (some)


Installing Linux to HD
======================

You will need: - the current bootimage (0.95a)
- the current rootimage (0.95a)
- tools to reinstall DOS, OS/2 etc after partitioning
- Hard drive specifics (number of cylinders, heads and
sectors- look at your BIOS setup for these)


===========================================================================:
: There is a diving (sub-aqua) rule which applies to doing this kind of :
: thing to your setup.... make sure you have at least 2 ways out of any :
: serious situation. What I mean by this is that you are about to kill :
: off your HD... you will not be able to boot up into DOS (or whatever) :
: on it so make sure that you have the tools to rebuild DOS on a write :
: protected floppy somewhere handy. You should also have the tools to do :
: this for Linux handy as well. What I used was :
: :
: 1. a bootable DOS floppy with: :
: :
: 1.1: pkunzip.exe to restore my DOS backups :
: 1.2: DOS fdisk and format :
: 1.3: rawrite.exe :
: 1.4: zipzap (a good binary file editor) :
: 1.5: Rootset.exe (see appendix B for src) :
: :
: :
: 2. Linux v0.95a rootimage rawritten to floppy (NOT WRITE-PROTECTED) :
: :
: 3. Linux v0.95a bootimage rawritten to a WRITE-PROTECTED floppy :
: :
: 4. Safe copies of the root and boot images stored as dos files (for :
: use with rawrite) :
: :
: :
: Escape plan 1: If all goes badly wrong you can use fdisk and format to :
: rebuild a dos disk partition. :
: :
: Escape plan 2: If you can perform a floppy only boot into Linux you :
: can use pfdisk to try to recover your partitions. :
: :
: Escape plan 3: You can't salvage dos but you can install linux to HD. :
: Use rootset.exe to complete the Linux installation then :
: fix things from Linux. :
: :
============================================================================


What to do:
===========

1. Backup everything. You have no choice in this.... you either backup
or say bye bye to DOS (or whatever else you have on your hard disk) (see
Appendix C if you don't like the sound of this restore and you are happy
guddling about with low-ish level Dos)

2. Boot up into Linux from floppy... login as root (see part 1 of this
guide for details)

[ Remember, one HD, 107 Mb -> 47 Mb Linux + 60 Mb DOS ]

3. Once logged in run pfdisk as follows

# pfdisk /dev/hda

This command means run pfdisk (a disk partitioner that comes on the
0.95a root disk) on /dev/hda (my entire primary hard disk). Some text
like this

==============================================================================
# pfdisk version 1.2.1 by Gordon W. Ross Aug. 1990
Modified by S. Lubkin Oct. 1991

Warning: The device "/dev/hda" is assumed to have:
cyls=306, heads=4, sectors=17 (an arbitrary guess).

This program was compiled for generic UNIX and therefore
does not know how to determine the disk parameters. You must
describe the disk geometry manually, using the 'G' command.
For help, enter: '?'
pfdisk>
==============================================================================

will appear if all goes well. The first thing to do is to tell pfdisk some
things about your hard disk using the 'g' command ('g' for geometry).
The format for this is g so for a disk
with 523 cylinders, 15 heads and 28 sectors (such as mine), the
command would be

pfdisk> g 523 15 28

Next you use the 'l' command ('l' for list) to see what is currently
set. You should see 4 entries (remember 4 primary partitions per hard
disk) but only one of them should have any numbers in it (one 107 Mb
partition) . For a partitioned disk this looks something like this

==============================================================================
# Partition table on device: /dev/hda
geometry 523 15 28 (cyls heads sectors)
# ID First(cyl) Last(cyl) Name # start, length (sectors)
1 6 0 292 DOSbi # 28, 123032
2 128 293 521 Minix # 123060, 96180
3 0 0 0 empty # 0, 0
4 0 0 0 empty # 0, 0
active: 1
pfdisk>
==============================================================================

Warning: Do not used edpart.exe (a Dos disk partitioner) on your disk 'cos it
completely wrecked my partition table (re-ordered everything and
wrote rubbish into some bits of the partition table). It is a good
disk partition editor for partitions smaller than 32 Mb but there are
better, friendlier ones around.

Make a note of the number in the ID field (labelled) for the single
partition. If you are partitioning a dos disk (as in the example
then this value will probably be 6. For other systems other values
apply.

Make a note of the maximum cylinder (521 on mine despite the fact
that the tech support people reckon 523 and the bios is set to 523)

Now for the clever bit..... I decided I wanted 2 partitions (47Mb
and 60 Mb) so I divide up my 523 cylinders in that ratio i.e.

(60/107) x 523 = 293 clusters for DOS (what a waste)
(47/107) x 523 = 228 clusters for Linux (near enough)
---
521 <-- make sure everything adds up!
---

then you use the "unnameable" command to partition your hard disk.
The format of this command is

n sys_id first_cluster last_cluster sys_name

where n ............ is the partition number
sys_id ....... is a magic number (ID in example above)
and sys_name ..... is optional and should not be used ('cos it
makes things fail (strange but true)


For DOS sys_id is 6 (and you should have seen that in the ID field
above), some other systems have their own code. Linux uses a Minix file
system so we use 128 (code for Minix) for the Linux sys_id and use whatever
the original ID was for the other partition (e.g. 6 for DOS).

For the example disk I would type the following

pfdisk> 2 6 0 292
pfdisk> 1 128 293 521

Explanation time..... partition 1 is now a Dos partition with 293
clusters (0 to 292 inclusive). Partition 2 is a Linux partition with
228 clusters.

Last bit... last but not least is the 'a' command ('a' for active)
which marks a partition as the one to boot from. I chose partiton 1
(DOS) 'cos Linux cannot boot entirely from hard disk without some
assistance (more later (eventually)). So ....

pfdisk> a 2

Finally, the 'wq' command ('wq' = write partition table and quit) so...

pfdisk> wq


4. That wasn't too bad was it? Now all you need to do is make a Linux
filesystem, install Linux and rebuild DOS. Filesystem first I think...

... at the command prompt use the fdisk command and you should see
something like this

# fdisk

Disk 0:
/dev/hda1: 61516 blocks active 16-bit DOS (>=32M)
/dev/hda2: 48090 blocks unknown partition type 0x80

What this tells you is that you have 2 partitions which, as far as
Linux is concerned, are known as /dev/hda1 and /dev/hda2.
/dev/hda2 is 48090 blocks long (1 block = 1 Kbyte thus this
partition is 48090/1024 or 46.9Mb) and is of type 0x80 (hex 80 =
decimal 128 (sound familiar yet?)). The other one is a 60 Mb Dos
partition.

Next you use the mkfs command (make filesystem) with the following
format

mkfs -c device blocks

the -c option specifies that it should perform some checking as it
works and the blocks figure is the one from fdisk (above). To make
a file system on this partition I would type

# mkfs -c /dev/hda1 48090

Your hard disk will be busy for a few minutes and you should get
some numbers popping up and changing on your screen. Don't worry
about these... you don't need to know anything about them (nor
could I tell you much about them :-))

[Error 1]

Ah, finally.... you are ready to install to HD! Thanks to Jim
Winstead's excellent installation scripts this is easy as pie so
without further ado type the following

# mount /dev/hda1 /mnt

to "attach" your newly created/formatted partition to the file
system (currently on the floppy) at the mnt directory in the root
directory. Remember to use whichever device you have formatted.

Now type

# cd /INSTALL
# ./install /mnt

to run the installation script and install the file system bits
and pieces in whatever is attached to the file system at /mnt.
This is of course part of your hard disk so you are now installing
properly to hard disk from the original file system on the floppy
disk. Clever huh?

When that finishes (you'll know when) you unmmount your partition
by using the umount command (e.g. umount /dev/hda1) and you should
sync (see part 1 of this guide) and logout (with the exit command).

WARNING: ALWAYS SYNC IF POSSIBLE AND LOG OUT PROPERLY. THIS IS UNIX
NOT MESSY-DOS AND THE 'OLD 3 FINGER SALUTE (CTRL-ALT-DEL) WAY
OF EXITING IS LIKELY TO LAND YOU IN TROUBLE SO DON'T DO IT!

5. Go back to DOS and edit the bootimage to reflect the partition
upon which you have installed Linux. How? Well, there are 2 main
methods:

1. Edit a fresh copy of the bootimage and rawrite it to a
FRESH floppy (i.e. not you existing bootimage 'cos you
never know when you might need that (pure floppy boot +
mount /dev/hd?? is a very useful way to sort out problems)).

2. Edit the bootimage disk (while keeping a copy- same idea as
at 1). I haven't come across any binary editors that can read
in rawritten disks but the Turbo C absread and abswrite
function seem to manage it (see Appendix C) so you could
write a small utility to make your life easier/safer (as I did)

Note: there is some debate as to which binary editor should be recommended. I
use Zipzap 7.15 and I'm very happy with it (it's shareware and available
on most major archives). Others like Norton Disk Editor. It's up to you
what you use.

The important bit, regardless of the method used, is to change the
bytes at offset 508 and 509 in the bootimage to store the minor and
major numbers for your new Linux partition.

For a hard disk the major number (offset 509) is always 3, and the
minor number (offset 508) is a combination of primary or secondary HD
drive and partition details (see FAQ for details). Here are the details

Device Major Minor Device Major Minor
------ ----- ----- ------ ----- -----
/dev/hda1 3 1 /dev/hdb1 3 65
/dev/hda2 3 2 /dev/hdb2 3 66
/dev/hda3 3 3 /dev/hdb3 3 67
/dev/hda4 3 4 /dev/hdb4 3 68
/dev/hda5 3 5 /dev/hdb5 3 69
/dev/hda6 3 6 /dev/hdb6 3 70
/dev/hda7 3 7 /dev/hdb7 3 71
/dev/hda8 3 8 /dev/hdb8 3 72


6. Boot up linux with your new boot floppy. It should give the
loading messages and then switch to booting and running from hard
disk.

Taaaaaa-Daaaaaa!


7. When you're happy with it logout and rebuild dos. I don't intend
to cover this in any detail 'cos most dos manuals already do
that. What you need to fo is to format your new dos partition using
format (on the dos bootable floppy you made). Something like

A:\ format /u /s /v:"Dos_5_0" c:

should do the trick. After that finishes you should have a
bootable Dos partition on your hard disk and you can reinstall
from your backups.
==============================================================================

Errors and solutions (for section 2)
====================================

1. If you have bad blocks on the first bit of your disk then mkfs will barf
and refuse to build a file system (happened to me!). The solution is to
make Linux your first partition. Those of you who are still awake will
notice that in the example above (my disk), my Dos partition is first on
the disk. Am I lying about this bad blocks thing? Nope! The reason is that
the mkfs suplied with 0.95a doesn't like bad blocks but the one with 0.95
(and I believe 0.12) doesn't care. I partitioned my disk with 0.95a,
sync'ed, exited, rebooted with 0.95 (an otherwise bug ridden release) and
ran mkfs. That done I rebooted with 0.95a. Cunning huh?
==============================================================================


Appendix A: The DOS bootable floppy
===========

Q. What is a "DOS bootable floppy"?
A. A DOS bootable floppy is a floppy disk containing just enough of DOS to be
able to be "boot up" and produce a command prompt. It consists of 4 things:

1. A DOS boot sector (first 512 bytes on the disk)
2. MSDOS.SYS (a hidden dos system file)
3. IO.SYS (another hidden dos system file)
4. COMMAND.COM (the DOS command interpreter)

Q. Do I need one?
A. If you run DOS at the moment and want to keep DOS somewhere on your hard
disk then you will need to create a DOS partition. Once created you will
need to reinstall DOS to this partition and a bootable floppy makes this
much, much easier.

Q. How do I make one?
A. You can create one in either of two ways using DOS commands.

1. Use the SYS command with a blank, DOS formatted floppy disk e.g.

# sys a:

will copy the "system" (the 4 things above) to a disk in drive A.

2. Use the FORMAT command with the /s and /u switches e.g.

# format /s /u /f:1.44 /v:"Dos_Boot" a:

will perform an unconditional format (checks for bad sectors etc) on the
1.44 Mb floppy disk in drive A, labels it Dos_Boot (11 characters maximum
allowed length) and copies the "system" on to it.

Now test it by resetting your machine with the boot disk in the floppy drive.
Your machine should start up (floppy is much slower than hard disk though)
and you should get a command prompt. Try some simple commands like DIR to
satisfy yourself that you are running from floppy only.

[ If your machine boots from hard disk and ignores the floppy you will need
to check your BIOS setup and enable "Floppy Seek at Boot" (AMI BIOS) or
equivalent option and try again. ]

Next you need to copy some things to your bootable floppy. What you copy is
entirely up to you but the following is the sort of thing I have on mine

4DOS - 4dos.com [ an excellent command.com replacement ]
- aliases.txt [ my command aliases/macros for 4dos ]

DOS - ansi.sys
- attrib.exe
- emm386.exe
- fdisk.exe
- format.com [ Absolutely essential! Do NOT forget ]
- himem.sys
- keyb.com

UTILS - pkunzip.exe [ for restoring backups ]

autoexec.bat
command.com [ Just in case ]
config.sys
filelist [ dir /skf c: > filelist before I backed up ]

Once you've got everything you think you want onto the floppy write protect
it (in case of foolishness) and THOROUGHLY TEST IT (i.e. make sure it boots,
formats, restores backups etc 'cos you're going to have to rely on it!)
=========================================================================

Appendix B
==========
/********************************************************/
/* Rootset.c - modify Linux bootdisk root device entry */
/* - quick Turbo C hack by [email protected] */
/********************************************************/


#include #include
#include
#include

void main(int ac, char **av)
{
char buf[512];


if (ac != 3) {
fprintf (stderr, "%s: update Linux bootimage root device details.\n",
av[0]);
fprintf (stderr, "Usage: %s \n", av[0]);
exit (-1);
}

/**************************************************************/
/* Don't remove this keypress bit 'cos it gives you a chance */
/* to run this program from your dos boot floppy, remove that */
/* disk, put your linux boot disk into the same drive, patch */
/* in the new root device details and save them. No HD or */
/* rawrite required! (v. handy if DOS is dead) */
/**************************************************************/

printf ("Insert Linux boot disk into drive A and press any key\n");
getch();

if (absread (0, 1, 0, &buf) != 0) {
perror ("Disk reading problem");
exit (-1);
}

printf ("Current rootdevice: Major %d Minor: %d\n", buf[509], buf[508]);

buf[508] = atoi (av[2]);
buf[509] = atoi (av[1]);

if (abswrite (0, 1, 0, &buf) != 0)
{
perror ("Disk writing problem");
exit (-1);
}
printf ("New rootdevice: Major %d Minor: %d\n", buf[509], buf[508]);

exit(0);
}
=========================================================================

Appendix C: Non destructive partitioning by Drew Eckhardt
==========

Reformatting your hard disk, and reinstalling all your DOS programs from
scratch can be a real nuissance, and is unecessary.

Breathe a sigh of relief : it is possible to non-destructively repartition
hard disks.

Notes : Before continuing, make sure you have a RECENT BACKUP.

I assume that you understand hex arithmatic, and are not
afraid of a little assembler or DEBUG.

Actually, a decent partition and disk editor will get you around
this - NU works gret.

Of course, the modern programmer doesn't use anything but a
source level debugger - so I've included some helpful hints
and the command syntaxes. However, there is no room for
handholding here : if you screw up, you might have to use
that backup. Don't do it unless you are confident in your
abilities.

Also, this procedure only works with NON-EXTENDED DOS partitions,
< 64K logical sectors, (DOS 4 large partitions add additional BPB
fields that I am unsure of - roughly the same procedure applies there
though. According to Townsend' Advanced MS DOS : Expert Techniques
for programmers

offset 26h will have the signature byte 29h if this is
the case, 20h a dword containing the number of sectors if
volume size > 64K sectors

I still use MS-Loss 3.3, with an ~82M partition under disk mangler
and fall into the tested category)

Large partitions, handled by a third party partition manager
and handled so that there are < 64K logical sectors
work - this was the case with my SCSI disk.

I will lay down what general procedures you need to know (
required to read / write the raw disk), as well as the
data structures we are dealing with. Then I will proceed
with the entire procedure, which applies the general procedures
in reading and modifying the data structures. If it looks like
a tech manual - it is. If you don't grok non-destructive
repartitioning, you shouldn't be doing it.

DEBUG has a Hexaritmatic command, h which will add and
subtract the two operands. You may find this useful.

IE : I have loaded sector 0 into memory at 0200, and wish to know
the address of the partition table at 1be.

-h 200 1be
03BE 0042

Where 03BE is the sum, and 042 the difference.

DEBUG prints a segment before the offset : note that your segments
will probably not match. The offset is what's important.

The 80x86 family is LITTLE ENDIAN. This means least significant
byte first - ie the internal representation of 0x12345678 would be
78 56 34 12. When dealing with multi-byte quantities, keep this
in mind.

When I say word, I mean word as in the Intel documentation :
16 bits. dword is 32 bits.

DISK BIOS addresses cylinder, head as zero based, sector
as 1 based. Same thing for the partition table.

DOS addresses sectors as 0 based, from the start of the
logical partition, and as logical sectors which may
consist of multiple physical sectors.

Unless otherwise noted, all numbers are hex.

You're better off using Norton Utilities - but Debug works fine
too.

This document is sort of a quick hack 8^)

Tools :

Required : DEBUG, and a disk defragmenter

Optional : partition editors (NOTE!!!! Make sure these DO NOT perform
any formatting, and allow you to edit partitions in the
REAL order they appear on the disk.), the Linux FDISK program,
utilities that save an image of the boot sector FATs, and
directory (IE Norton's Format Recover), a raw disk editor
(Norton Utilities NU)..

Procedures :
Editing memory with debug :

d ADDRESS l LENGTH

will dump memory

f ADDRESS l LENGTH values

will fill memory.


Reading and writing DOS logical sectors (using debug):

Reading is accomplished using the debug l command.

l ADDRESS DRIVE SECTOR COUNT

Where ADDRESS is the hex address of where to put the data,

DRIVE is a 0 based drive number (IE A:=0, B:=1, C:=2, etc. If there
is only one floppy drive, it is considered both A: and B: as it is
in DOS)

SECTOR is a zero based sector number. Of interest to us are
sector 0, the boot sector, and the sectors immediately following it -
the FAT's.


COUNT is the number of sectors to read.


So, to read in the first 10 sectors of the FAT on my E: partition (the
FAT starts at sector 1), storing them at 0200, I would enter

l 0200 4 0 A

Writing is done with the W command, which takes the same
parameters. Assuming I had edited the boot sector of E: at 0200 in memory,
and wanted to write it back to disk, I would type in

w 0200 4 0 1


Using Norton Utilities NU:

xplore disk, hoose Item, ector, dit/display

Unfortunately, absolute sector 0 falls outside of all partitions (this
is where the partition table is), and we need to use a different
procedure for it.

Reading / Writing absolute sector 0:
The following debug / assembler interaction shows how to read absolute disk
sector 0 , replace xx with 80 for hard disk 0, 81 for hard disk 1:

-a 0100
1984:0100 mov ax, 0201
1984:0103 mov bx, 0200
1984:0106 mov cx, 0001
1984:0109 mov dx, 00xx
1984:010C int 13
1984:010E int 20
1984:0110 ^C

-


-g = 0100

This will read sector 0 into DS:0200. To write it back,

-a 0100
1984:0100 mov ax, 0301
1984:0103 ^C


-g = 0100.


Using Norton Utilities NU :
Under xplore disk, choose hoose item, bsolute sector, dit / display



Structures :

1. The Partition table

The partition table resides at absolute sector 0 (ie
cylinder 0, head 0, sector 1) on all harddisks. It is accessed
by a short bootstrap loader on that sector, which reads the partition
table and then picks a partition from which to load the boot sector for
the operating system.


The partition table itself resides at offset 1be. It is 64 bytes (decimal)
in length, plus the two byte signature 55 AA. When dealing with the
partition table, make sure byte 40 (offset 1fe of the sector) is 55 and
byte 41 (offset 1ff of the sector) is
of the sector




The partition table is subdivided into FOUR 16 byte entries, fielded
as follows :
offset length field
0 byte bootable 80h = bootable, 0 = not
1 byte starting head number
2 word starting cylinder (and sector - sector is 1 based
high byte is low byte of cylinder, low byte low 6 bits is
sector, high 2 bits of low byte high 2 bits of cylinder)


typically, sector = 0.

4 byte system 1 = primary DOS, 12 bit, 4 = primary DOS 16 bit,
5 = extended DOS, 8 = NON-DOS (might be usable)
5 byte ending head number


6 word ending cylinder / sector

8 dword starting sector (relative to begining of disk - THIS IS ZERO
based)

C dword number of sectors



My partition table on drive 0 looks like :

-d 3be
1984:03BE 80 01 ..
1984:03C0 01 00 01 05 22 08 22 00-00 00 0A 07 00 00 00 00 ....".".........
1984:03D0 01 09 51 05 E2 2B 2C 07-00 00 E4 7F 02 00 00 00 ..Q.b+,...d.....
1984:03E0 00 00 00 00 00 00 00 00-00 00 00 00 00 00 00 00 ................
1984:03F0 00 00 00 00 00 00 00 00-00 00 00 00 00 00 55 AA ..............U*

Interpreting this, we can see that I have two partitions in use -
partition 1 and 2, the rest being blank.


My first partition is bootable (80)
FAT size is 12 bits (type = 01)
It starts at head 1, cylinder 0, sector 1, which is sector 22
(34 decimal) relative to the start of the disk.
It is 0000070a sectors in length (1802 decimal), and ends on
head 5, cylinder 8, sector 22 (34 decimal).


My second partition is non-bootable (00).
It is type 51 (Disk Mangler) meaning I need to
find out fat size some other way.

It starts at Head 0, cylinder 9, sector 1, which is sector 72c
(1836 decimal) relative to the start of the disk.
It is 27FE4 (163812 decimal) sectors in length, and ends on
head 5, cylinder 32b (811 decimal), sector 22 (34 decimal)..

See how I got ending cylinder? In Hex, each digit is a nibble. You can
easily convert to binary a nibble at a time. IE E2 becomes 1110 0010


The low 6 bits (sector) are 10 0010 = 22
The high 2 bits are 11 = 3

So high byte cylinder is 03, low byte is 2B so cylinder = 032b

Got it?



2. The Boot sector

The second important piece of data is the bootsector. There are a number
of fields we are interested in. I have ommitted the DOS 5 extended fields
(Can't give you an answer I'm 100% sure on), as well as fields unecessary
to our procedure.

Fields we are interested in :
Offset Size Field
b word bytes per LOGICAL sector - divide by 512 to get physical
to logical mapping

d byte sectors per cluster. Multiply by logical sector size and
divide by 1024 (decimal) to get K / cluster

13 word total number of LOGICAL sectors. This is one field
extended by DOS5.

18 word sectors per track

1a word heads

Clusters = sectors / sectors per cluster.

Since my first partition on drive 1 is fairly boring, we'll look at
partition 2 - E:

-l 0200 4 0 1
-d 0200
1984:0200 EB FE 90 4E 4F 53 59 53-54 45 4D 00 08 04 01 00 k~.NOSYSTEM.....
1984:0210 02 00 02 F9 9F F8 0A 00-22 00 06 00 2C 07 80 00 ...y.x.."...,...

The first three bytes are either a near jump over the actual BPB or
a short jump and a NOP. There follows an eight byte system ID, and then
the Good Stuff.

I have 800 (2048 decimal) bytes per logical sector, meaning 4 physical
sectors = 1 logical sector.
.
There are 4 logical sectors per cluster, meaning there is 8K in
each cluster, 16 logical sectors per cluster.

There are 9ff9 (40953 decimal) logical sectors in my partition, meaning
163812 decimal physical sectors (Notice how this number matches the
number in the partition table.)

There are 22 (34 decimal) sectors per track, and 6 heads.

3. The FAT

The final interesting data structure is the FAT. If you have some other
reliable means of insuring all data is at the begining of the partition
you can ignore this.

If there are > 4096 clusters, you definately have a 16 bit FAT.

The FAT will begin with an FFF8 signature - 16 bit or (FF8) 12 bit.
What follows
is one entry for each cluster of the disk - with FFFF (FFF) marking
end of file, FFF7 (FF7) marking a bad sector, 0000 (000) empty, basically
anything else a next cluster pointer.

Basically, all that were interested in is that there are roughly enough
0 entries all together, with only BAD marks and no end of file or next
cluster pointers.

12 bit entries are packed two per three bytes - low byte = low byte
first cluster, middle byte low nibble is high nibble for the first cluster,
middle byte high nibble is low nibble for the second cluster, high byte
is high byte for the second cluster.


Step 1:
Make a FULL backup.

1a:
If you have a program to save off an image of the boot sector,
do it now. Dumping the boot sector / partition sectors to
floppy disk (do an rcx:200 in debug, then a w FILENAME to
save to a file) would also accomplish the same thing.

This will save you the trouble of having to restore an entire
backup if you screw up either of these sectors.

Step 2:

Defragment. All data must be contiguous.


Step 3:
Decide where the partition goes, and how big it is.

The new partition must fullfil these requirements :

1. It resides at the end of the original partition.
2. It is situated such that the original partition has a boundary
that ends on a FULL logical sector (In my example, this is four
physical sectors).

In my case, I am simply starting testing on my already packed SCSI
disk, and want a ~1 megabyte partition, or 2048 sectors.

This partition ends at :

cylinder 32b (812 decimal), head 5, sector 22 (34 decimal)

If I divide by 34 sectors per track (available in the boot record),
I get 60 tracks. If I divide by 6 heads, I get 10 cylinders.

So, I will start my partition at

cylinder 802 decimal (322 hex)
sector 0
head 0
relative sector 802 * 34 * 6 decimal= 163608


I will adjust the size of my E: drive down by 510 logical (2040) (decimal)
physical sectors, and make this my new partition.

Note that this falls on the required logical sector boundary.

Likewise, I will reduce the size of my second partition by 2040 (decimal)
sectors, to 161772 (decimal) sectors and reduce the ending cylinder to
801 (decimal).

Step 4:

Verify that all data is contiguous, and we have enough free space
for the partition we are about to create.

Use a reliable disk space map utility, if none is available you will
have to do so manually. Again, NU can fill this role.


To do this manually, first determine FAT size

check - partition table
- number of clusters (roughly - (sectors - 1) / sectors per cluster)
if > 4096, definately 16 bit

After you've decided on 12 bits or 16 bits,

Find the first 0 cluster (The disk has been defragmented?)

Count consecutive 0 or BAD FAT entries (1 per 2 bytes on 16 bits,
3 per 2 bytes on 12 bits), multiply by K per cluster (as determined
from the boot sector - see data structures above), and be happy.

If there are intervening NEXT CLUSTER pointers, or END OF FILE (FFFF),
try a different defragmenter.

Also, not that some defragmenters will not move hidden files. You will
have to unhide these, defragment, and rehide these.



5. Adjust boot record of old DOS partition. Change # sectors field,
offset 13 hex (word), be sure to adjust in number of
LOGICAL SECTORS, not physical sectors (as determined from the
word size field at offset b divided by a 200 hex (512 decimal) byte
sector size)

After image for my example :
-l 0200 4 0 1
-d 0200
1563:0200 EB FE 90 4E 4F 53 59 53-54 45 4D 00 08 04 01 00 k~.NOSYSTEM.....
1563:0210 02 00 02 FB 9D F8 0A 00-22 00 06 00 2C 07 80 00 ...{.x.."...,...

6. Adjust partition table. Change ending head, cylinder, sector
field of old partition, total sectors of old partition,
add new partition.

Avoid overlap. Note that new partition table becomes active
on reboot.

After image for our example:
-d 03be
1563:03BE 80 01 ..
1563:03C0 01 00 01 05 22 08 22 00-00 00 0A 07 00 00 00 00 ....".".........
1563:03D0 01 09 51 05 E2 21 2C 07-00 00 EC 77 02 00 00 00 ..Q.b!,...lw....
1563:03E0 C1 22 FF 05 E2 2B 18 7F-02 00 F8 07 00 00 00 00 A"..b+....x.....
1563:03F0 00 00 00 00 00 00 00 00-00 00 00 00 00 00 55 AA ..............U*

Yes, I know 0xff is a bizarre identifier for the linux partition. That's
how I did my first one, that's how I did this one.

References :
Norton, Peter Norton's Programmer's Guide to the IBM PC
Townsend, Advanced MS-DOS Expert Techniques for Programmers

Hopefully this has been enlightening, and possibly even useful to some.

Mail comments to [email protected]
=========================================================================

Section III: What now?
======================

I was going to write about the maintainance and running of Linux, installation
of new software etc but I've been beaten to it. For details of all this sort
of thing please see Chuck Boyer's excellent "Beginner's Guide" which was
appeared in the alt.os.linux yesterday.
=========================================================================



From: [email protected] (I Reid)
Subject: HD booting: a solution
Date: 6 Apr 92 07:22:52 GMT

==============================================================
I finally got round to writing this part of section II of my guide to Linux
installation. It's a bit rough around the edges yet so give me some feedback
on it and I'll see what I can do with it.

Sorry for the delay,

Iain
==============================================================

Booting Linux from your hard disk
=================================

For a good general introduction to the principles of what happens when your
computer boots up you should read the read.me file that comes with the
freeware program bootany (more later) or any good Dos programmers's reference.
(For the latter I can recommend "The Waite Group's MS-DOS Developer's Guide").
You don't need to know very much about what happens but it may help you
understand things if you've read a little about what normally happens.

Remember the master boot record and primary partitions from earlier in the
section? If not, you should re-read the relevant bit (at the start of the
section) before continuing.

On a very simplistic level, the mechanism of hard disk boot up is as follows
(assuming only one bootable partition)

1. The boot sector of the hard disk (physically the first sector on the disk)
is read into memory at 0000:7C00 by the ROM BIOS.
2. A few basic validation checks are carried out to make sure that this is a
proper boot sector.
3. Part of the boot sector is moved within memory and from there validates the
partition table.
4. The boot sector on the one and only startable partition is read into memory
at 0000:7C00 and the executable components run. This is operating specific
and loads Dos, Linux or whatever OS you are using.

The situation with 2 primary partitions is slightly more complicated because
the ROM BIOS cannot cope with having two startable (bootable) partitions and
will boot neither if it finds this situation. So, how do we get round this?

The master primary boot record contains code to check the partition table and
choose the startable partition. If we can replace this code with some that
gives us the choice between say Dos and Linux then the boot sector of the
partition we choose will be loaded in and run. This boot sector contains the
code for the specific operating system.

As you can see, there are 2 distinct problems:

1. What do we put into the master boot record to let us choose partitions
2. What do we put into the Linux partition to let us boot it

Thankfully the solution is easier than the theory so, without further ado let
us press on. Starting with Linux.....

You will need: - Shoelace
- a compiled version of the bootimage (either the original
bootimage or from a kernel rebuild)

You should find Shoelace in any of the Linux archives as a compressed tar file
(i.e .tar.Z suffix). Transfer it to Linux, uncompress it and move it to /tmp
(creating /tmp with mkdir if necessary). Now cd to /tmp and you're ready to
start....

If you untar the shoelace tar with tar xvf shoe.tar (or whatever it's called)
you should see something like this
[Root]/tmp: tar xvf shoe.tar

shoe/
shoe/Readme
shoe/bootlace
shoe/src/
shoe/src/bpb.h
shoe/src/part.h
shoe/src/proto.h
shoe/src/shoelace.h
shoe/src/laceup.c
shoe/src/shoelace.cdif
shoe/config
shoe/disktab
shoe/laceup
shoe/readme
shoe/shoelace
shoe/winiboot

What this means is that shoe.tar has been extracted, the directory /tmp/shoe
has been created and within that a directory called src has also been
created. There are some files in each. If you go into the shoe directory and
do a directory listing with ls you should see

Readme config laceup shoelace winiboot
bootlace disktab readme src

You should read the readme file (note the small r in readme) because it
contains some important warnings which I do not intend to repeat. You should
read them and be aware of them. Much of the readme file is irrelevant although
you should try the floppy bootup it suggests if you are not happy with things.
The synopsis of it is that you should move laceup, bootlace, winiboot, config
and disktab to /etc and shoelace to /.

After that you need to move your bootimage to somewhere standard. I chose to
call it /vmlinux. The name you give it is up to you.

Next you need to edit /etc/config. Initially it contains a lot of entries that
are irrelevant to Linux (it is mainly a minix program). Most of these can be
happily deleted (mine looks like this)

==============================================================
# Shoelace configuration file

boot /vmlinux

==============================================================

The only important bit is the boot entry!

Next go to /etc (cd /etc) and type

# laceup /dev/hda2 wini

Be very careful here! What you are trying to do is to put the operating system
specific information into the boot sector of your LINUX PARTITION. In my case
this is /dev/hda2. Read and heed the warnings in the shoelace readme file
before you do this 'cos one false move and you could end up in a mess. You
should see something like this

[Root]/etc: laceup /dev/hda2 wini
Identification: ShoeLace
Sector size: 512 bytes
Block size: 2 sectors
Boot sectors: 2 sectors
Number of fats: 0
Root directory size: 0 entries
Volume size: 0 sectors
Media descriptor: 0xf8
Fat size: 0 sectors
Track size: 0 sectors
Cylinder size: 0 tracks
Hidden sectors: 0 sectors
Floppy table: 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00 0x00
[Root]/etc:

That's part one done. Next you have to modify the master boot record (first
sector on the hard disk) to implement the boot time partition decision.
/dev/hda is the ENTIRE hard disk (and the main way to access the master boot
record from linux). Again, read the warnings (ESPECIALLY IF YOU ONLY HAVE ONE
PARTITION) and press on if you are happy with things. For my setup I would
type

# laceup -w 2 /dev/hda

and if all goes well I will see

Boot Hd Sec Cyl Type Hd Sec Cyl Base Size
1 80 1 1 0 6 e 5c 24 28 123032
*2 0 0 41 25 80 e 9c 9 123060 96180
3 0 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0 0

What this shows is that I have 4 primary partitions (remember them?), only 2
are in use. the first one is marked bootable (Boot field of hex 80 = bootable, 0
is not). This is my Dos partition (type 6). The other one (number 2) is my Linux
partition (type hex 80). The -w 2 switch to laceup means that I want partition
2 to be my default partition (i.e. if I don't select a partition then it
should boot partition 2). It doesn't matter whether the partition you choose
is bootable or not. Shoelace will fix all 🙂

Ok, now logout, reboot (without a floppy in the floppy drive) and you should
get an unpleasant looking boot menu (with a star by the 2 in my case). Select
the number that corresponds to your Linux partition and watch it boot. Here is
a bit of mine as an example (i.e. what's still on screen when the login prompt
comes up!)

*2 0 0 41 25 80 E 9C 9 123060 96180
3 0 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0 0
/dev/fd0 not found - guessing major and minor device numbers
processor = -1
ramimagedev = 512 /dev/fd0
ramsize = 0
rootdev = 770 bootdev
scancode = 13

Component base text data bss stack symbols model
ShoeLace : 80000 843C 1AC4 FF8 3378 1D90 ComI&D
Load Parts
kernel : kernel not found

Load Image
Press to see SVGA-modes available or any other key to continue.

8 virtual consoles
4 pty's
Partition table ok.
14738/48090 free blocks
14717/16030 free inodes
1209 buffers = 1238016 bytes buffer space
Free mem: 2097152 bytes
Adding Swap: 2047 pages (8384512 bytes) swap-space

[blank lines deleted]

Welcome to Wizbit - running 0.95a (6/4/92)

Wizbit login:

==============================================================

IMPORTANT: IF YOU RECOMPILE YOUR KERNEL REMEMBER TO COPY /USR/SRC/LINUX/IMAGE
TO /VMLINUX (OR YOUR EQUIVALENT)


That's it! Before anyone asks, I will answer 3 other questions here...

1. Can I use bootany instead of shoelace?
2. How did I name my system so that I don't get (none) login:
3. How did I get 1209 buffers with only 4Mb of RAM?


Using bootany
=============

Bootany is a boot time partition selector like shoelace and, in theory should
be usable to give you the boot choice like shoelace. Unfortunately, it doesn't
seem to work. When I tried it on my machine it altered my partition table and
wouldn't boot anything. The solution to this was easy enough- boot linux from
floppy and rebuild the partition table using pfdisk.

Another warning.... if you use the w option on pfdisk be very careful
about what is. For example, don't use bootany.sys (part of bootany)
unless you want to wipe out your partition table (no Dos, no Linux....
nothing!). To get out of this you'll need a partition table editor on a
bootable floppy (remember I told you to keep your linux root and boot floppy
safe? Now you know why :-))

----
Question 2..... add a line to your /etc/rc that says

/bin/hostname Wizbit

where "Wizbit" is the name of your system. Reboot and be amazed 🙂 (if you
remembered to get hostname from the archives of course!)
p.s. And no pinching "Wizbit".... get your own!

----

Question 3: First of all be very sure that you want to change things on your
system. I run almost exclusively single user and have never run out of real
memory ley alone touched my 8Mb of swap memory. I read a posting to
alt.os.linux a while ago from someone (sorry, don't remember who) who said
that he had bumped up his buffers (on a similar setup) and had noticed a
significant speed increase. My changes are on trial to see a. if it makes a
difference to me and b. if it causes me any problems. What to do....

In /usr/src/linux/init/main.c you'll see a bit that looks like this

if (memory_end > 16*1024*1024)
memory_end = 16*1024*1024;
if (memory_end >= 12*1024*1024)
buffer_memory_end = 4*1024*1024;
else if (memory_end >= 6*1024*1024)
buffer_memory_end = 2*1024*1024;
else if (memory_end >= 4*1024*1024)
buffer_memory_end = 3*512*1024;
else
buffer_memory_end = 1*1024*1024;

If after much thought and consideration etc you want to change things then
find the memory_end entry for you (4Mb = 4*1024*1024 = my setup) and change the
buffer_memory_end bit. I changed mine to look like this

if (memory_end > 16*1024*1024)
memory_end = 16*1024*1024;
if (memory_end >= 12*1024*1024)
buffer_memory_end = 4*1024*1024;
else if (memory_end >= 6*1024*1024)
buffer_memory_end = 2*1024*1024;
else if (memory_end >= 4*1024*1024)
buffer_memory_end = 4*512*1024; /* IAR 3/4/92 */
else
buffer_memory_end = 1*1024*1024;
main_memory_start = buffer_memory_end;

Recompile the kernel and that's it.

Iain

p.s. remember to change it back if you have problems 'cos it is a non standard
hack :-). Also check that it is not the source of a problem before
posting to alt/comp.os.linux!