Subject: Linux-Misc Digest #820
From: Digestifier <Linux-Misc-Request@senator-bedfellow.MIT.EDU>
To: Linux-Misc@senator-bedfellow.MIT.EDU
Reply-To: Linux-Misc@senator-bedfellow.MIT.EDU
Date:     Wed, 16 Mar 94 06:13:13 EST

Linux-Misc Digest #820, Volume #1                Wed, 16 Mar 94 06:13:13 EST

Contents:
  Re: Installing Linux (Bruce Wielinga)
  Re: pronunciation of linux (Kevin Lentin)
  Re: DOOM for X (Alfred Longyear)
  Re: Linux and Hayes 28,800's - anyone have them working? (Nathan Stratton)
  SEEKING: CD-ROM info (Rick Emerson)
  Re: Prepare for DOOM (parody) (Whitstler)
  Re: Upgradability? (Anders Eriksson)
  Re: Notebook - PCMCIA (Terry Lambert)
  Re: Why Linux has problems on IBM Valuepoint and PS/1 and how to fix it (I think) (Another time's forgotten space... (mickey))
  Re: Opinions wanted about SCO-unix (vs AIX/Linux). (Nathan Bardsley)
  Re: Linux and Tapestreamer (Sanchez Rodriguez Jorge)
  Re: libm.so.4.5.21 in /usr/lib ?? (J.S. van Oosten)
  Re: An idea to cut down noise  (J.S. van Oosten)
  Hangman (was Re: Cyrix 486DLC/40 Compatibility with Linux) (J.S. van Oosten)
  Re: Multiple screen feature in Linux? (J.S. van Oosten)
  Re: 486SX vs 486DX - difference in performance using Linux (John Lockwood)
  Why Linux has problems on IBM Valuepoint and PS/1 and how to fix it (I think) (Dimitri Vulis, CUNY GC Math)

----------------------------------------------------------------------------

From: wielinga@dingo.cc.uq.oz.au (Bruce Wielinga)
Subject: Re: Installing Linux
Date: 15 Mar 1994 00:55:19 GMT

Steve Havelka,,, (shavelk@agora.rain.com) wrote:
> Is it possible to install Linux on a machine with only 2 meg of memory?
> I've got a 386sx-16 with 2 meg of ram and a 40 meg HD, and would like to
> use Linux.  Is it possible to do this without upgrading?

Yes, I'm running on a laptop with 2megs 386sx and a 60meg hardrive(which might turn out to be the real problem). It's possible but really squeezy. So I think
upgrading should be looked into.

> -- 
> Steve Havelka
> shavelk@agora.rain.com

Bruce

------------------------------

From: kevinl@bruce.cs.monash.edu.au (Kevin Lentin)
Subject: Re: pronunciation of linux
Date: 15 Mar 1994 01:00:08 GMT

On 14 Mar 1994 23:15:31 GMT, Tom Fox wrote:
> Linux rhymes with clinics

Why? All the posts from Linus et al seem to indicate that the u is
pronounced as in uncle but yes the i is pronouced as in clinic. (Even
though this sounds quite strange to me as an English speaker. I prefer the
i to be pronounced as in the pronoun I).

In his comment, Linus refers to 'like minix' and I am pretty sure he is
referring to the first i. 

-- 
[==================================================================]
[ Kevin Lentin                   |___/~\__/~\___/~~~~\__/~\__/~\_| ]
[ kevinl@bruce.cs.monash.edu.au  |___/~\/~\_____/~\______/~\/~\__| ]
[ Macintrash: 'Just say NO!'     |___/~\__/~\___/~~~~\____/~~\___| ]
[==================================================================]

------------------------------

Crossposted-To: comp.os.386bsd.apps
From: longyear@netcom.com (Alfred Longyear)
Subject: Re: DOOM for X
Date: Tue, 15 Mar 1994 02:50:44 GMT

iiitac@uk.ac.swan.pyr (Alan Cox) writes:

>I'd much rather have Linux joystick support to be honest, that way I can
>read news with one hand and play doom with the other.

Ask and ye shall receive . . . 

sunsite.unc.edu:/pub/Linux/kernel/misc-patches/joystick-0.6.tar.gz



------------------------------

From: nstn@netcom.com (Nathan Stratton)
Subject: Re: Linux and Hayes 28,800's - anyone have them working?
Date: Tue, 15 Mar 1994 02:59:45 GMT

In article <Mar11.220605.33112@acs.ucalgary.ca> jcsokolo@acs.ucalgary.ca (Jason Conrad Sokolosky) writes:
>Brian Kramer (bjkramer@remus.rutgers.edu) wrote:
>: Does anyone have linux working with hayes 28.8's?  I may try
>: it for a slip connection if possible.
>
>I didn't know they had 28.8's out yet.  Who sells them besides Hayes????
AOOM sells V.FAST 28.8 internal Faxmodems. MicoeCenter sells them for 
$249.95.

-- 
 +---------------------------------------------------------------------------+
 | Nathan Stratton                                                           | 
 +---------------------------------------------------------------------------+
 |              Internet:       nstn@netcom.com                              |
 |              Phone:          (703)534-9755                                |
 |              Beeper:         (202)956-0523                                |
 +---------------------------------------------------------------------------+

------------------------------

Subject: SEEKING: CD-ROM info
From: rick.emerson@dscmail.com (Rick Emerson)
Date: Tue, 15 Mar 94 14:30:00 -0640

I've got a CD-ROM drive, pl15j loads it up, but I need a source for
something to get it to at least run as a simple CD drive.  Also, I'm
having no luck mounting the Prime Time Freeware CD-ROM's which *may*
be in High Sierra format.  Suggestions?  FAQ?

Rick
  
...
 * ATP/Linux 1.42 * Wear old clothes when you fight for truth and liberty.

------------------------------

From: slouken@pennington.cs.ucdavis.edu (Whitstler)
Subject: Re: Prepare for DOOM (parody)
Date: Wed, 16 Mar 1994 00:23:12 GMT

Jayme (jaymecox@coyote.rain.org) wrote:
:       Well, it looks like the Doom port to linux (and quite a few
: other Unix's also) is well on it's way. So I thought I'd just get you ready
: for it by saying to prepare for at least 100+ articles with such exciting
: titles as:

:       How to run DOOM over Term?!?
:       Doom won't run on SLS1.0
:       Doom is k00l!!! Play or Die!!


Don't forget:

        Help!  I can't stop playing DOOM!
and
        Can I get rid of all those big files in /lib so
                I can fit more WADS on my hard-drive?

: and many more!  :^)


*grin*  Just my two cents worth. :)

        -Sam


------------------------------

From: anders@lulea.trab.se (Anders Eriksson)
Subject: Re: Upgradability?
Date: 15 Mar 94 08:05:23 GMT

Andreas Helke (andreas@orion.mgen.uni-heidelberg.de) wrote:
: RYAN HUNG (rhung@physics.ubc.ca) wrote:
: : I'm waiting for the release 1.0 of the Linux kernel before I install
: : Linux.  However, I've been wondering: considering the rapid rate of change
: : in Linux, how easy is upgrading?  Say, for example, there were a kernel
: : upgrade.  Do distributions usually provide upgrade packages that can run
: : quick and easy?  Or is a lot of hacking necessary?

: Upgrading is easy, just unpack a tar or cpio archive. But you never know
: in advance whether you still have a working system after the upgrade. The
: debian linux package is trying to solve this problem by providing a formalized
: upgrading mechanism. But this release is still in beta development and there
: can be unexpected big changes in it's layout during the beta phase.

: Exchanging the kernel is in general trivial, except that in some cases you
: have to upgrade gcc and the librarys first to compile the new kernel.

: BTW the official Linux kernel version 1.0 is out since sunday. Does anyone
: one know what has changed between pre-1 and 1 ?

: Andreas
: --

: Andreas Helke

: Institut fuer molekulare Genetik, Universitaet Heidelberg
: Im Neuenheimer Feld 230 
: 69122 Heidelberg

Hi folks,

I too have waited for the 1.0 release and now the question arise: How do I
know when the distributions are updated with the new kernel?
--
      ____________________________________________________________________
     /                                                                  / \
    / Anders Eriksson               Fax:        +46 920 75490          /___\
    | Phone: +46 920 75403          Speed mail: anders@lulea.trab.se   |
    | Snail mail: Telia Research AB, Aurorum 6, S-977 75 Lulea, Sweden |
 _ /_________________________________________________________________  /
 \___________________________________________________________________\/ 
 

------------------------------

From: terry@cs.weber.edu (Terry Lambert)
Crossposted-To: comp.unix.bsd
Subject: Re: Notebook - PCMCIA
Date: 15 Mar 1994 01:14:54 GMT

In article <2m1bfj$ptg@rzsun02.rrz.uni-hamburg.de> andrae@rzdspc2.informatik.uni-hamburg.de (Caroline Andrae) writes:
>|> >Have you tried to put your notebook on a network with a PCMCIA ethernet
>|> >card?

ME> Works great; all you need is a PCMCIA enabler shim, which is very few lines
ME> of code to write, especially if you have doc's for PCMCIA.  Most of the
ME> network cards are NE2000 compatible.  Should take you half an hour.
ME> 
ME> It also works with anything else PCMCIA if you code things right (or one
ME> network card exactly if you do it wrong), including FAX modems and SCSI
ME> interfaces for things like tape drives.  8-).

>Could you please explain, or post the thing you wrote?

I can't post code for well known reasons, or we'd have had Sun-style
shared libraries in June of 93 (...er; without C++ support) and have
Streams and run Xenix and UNIX binaries today, etc.

I can tell you what I did to get the information needed, though.

The first thing I did was notice that Kurt Mahan wrote something like
it for UnixWare, which meant it was possible.  8-).

Then I bought a copy of "Undocumented DOS, Second edition" and started
disassembling the PCMCIA BIOS calls documented there to see what they
did to do PCMCIA'ish things (this is the same book that talks about
the EISA bus signature word for identifying EISA bus machines to turn
off bounce-buffers, which are not applicable to EISA).  This will get
you enough to attach cards on startup, although there's little else you
can do in line with the standard.

Most recently, I've called up Intel and asked for documentation for the
chipset.  Each chip can handle two cards, and most machines have only
one chip.  The chip lets you map an interrupt and up to four address
windows onto what it expects to be an ISA bus on the other side.  You
get to pick where they go.

An attach is a determination of device type, a lookup of the driver, and
a mapping matching the driver settings for the device.  The attach should
be done before the probes go.  This is sufficient to get most ethernet
cards for PCMCIA running, since they mostly all look like NE2000's.

I'll hack more on the thing when I get the Intel docs (thanks to Kurt
who pointed me at them) and can borrow the laptop again with a couple
of different types of cards.  I suspect getting a SCSI interface up will
be quite a chore for the attach requirements.


                                        Terry Lambert
                                        terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.

------------------------------

From: louray@seas.gwu.edu (Another time's forgotten space... (mickey))
Crossposted-To: comp.os.linux.help
Subject: Re: Why Linux has problems on IBM Valuepoint and PS/1 and how to fix it (I think)
Date: 15 Mar 1994 23:33:37 GMT


Hey 'gain..
This seemed to be almost what I wanted...but not quite..

I'm trying to install linux on my home PC (only about 50-60 megs for
linux, though :(..anyway..)..I currently only have the flopy made with
rawrite3 (I've gotten the MCC distribution on DOS floppies right now..

well..neither read-only , nor read-write works..but anyway..when I go
to fdisk, it says that it can't read /dev/hda...

I tried holding down the 'alt' key and the 'shift' key, but lilo never
stopped to ask me 'bout my disk..for the second option described in
the original post, I don't know how I can recopmile the kernel, as I
don't have linux install'd...

anyone got any ideas?

(Oh, and one more thing..(I'm new at this..can't you tell? :)...is the
number of sectors in a disk the actual number (in my case 133263) or
is it sectors/treack? (here 17) ??)

thanks, i.a.,
-- 
pe-|| ||  "it feels about like running a red light:
ace|| ||   there ain't no point in looking behind us now...."
|||| \/|                         louray@seas.gwu.edu
\\\\   |  MICHAEL PANAYIOTAKIS   ...!uunet!seas.gwu.edu!louray

------------------------------

From: nathan@ftp.health.org (Nathan Bardsley)
Crossposted-To: comp.unix.advocacy,biz.sco.general
Subject: Re: Opinions wanted about SCO-unix (vs AIX/Linux).
Date: 15 Mar 1994 22:04:23 -0000

In article <d0bpl.763777041@dtek.chalmers.se>,
Patrik Larsson <d0bpl@dtek.chalmers.se> wrote:

>   What are the pros and cons for SCO-unix in general, and
>compared to AIX (and maybe Linux) in particular?

SCO is a nice system if all you're going to do is to run their
software.  If you want to compile your own applications or develope
some inhouse networking apps, good luck.  The SCO developement
package is, in a word, crippled.
-- 
Nathan Bardsley -- nathan@health.org

------------------------------

Subject: Re: Linux and Tapestreamer
From: zz91f027@trono.etsiig.uniovi.es (Sanchez Rodriguez Jorge)
Date: 15 Mar 94 16:48:12 +0100

Michael Will (michaelw@desaster.sunflower.sub.org) wrote:
> If you can afford I would recommend SCSI-tapes rather than 
> floppy-streamer-tapes - it is faster, reliable (and that is
> what is important on backups, mind you :-) ) and exchangable.

A question: are floppy-streamer-tapes supported in standard SLS kernel
or need additional drivers / patches ? . I'm planning to buy one of
these tape drives. Thanks in advance.

Jorch.


------------------------------

Crossposted-To: comp.os.linux.admin
From: jvoosten@compiler.tdcnet.nl (J.S. van Oosten)
Subject: Re: libm.so.4.5.21 in /usr/lib ??
Date: Tue, 15 Mar 1994 01:05:08 GMT

David E. Filip (dfilip@colornet.com) wrote:
: I am comming in at the end of this thread ...
The end of the thread of your wits ? :-)

: I am trying to compile some X sample code (specifically, the sspkg stuff
: under xview from the Slackware distribution) and I am told that ld cannot
: find libm (specifically, ld: no such file or directory libm). Now, does
: that mean that for some reason I am running ld and not ld.so, which is
: not looking for shared libraries ? Or, is it for some reason not looking
: in /lib (which is where libm.so.4 lives) and I need to move it to /usr/lib ?

*grin* While compiling, you're always using 'ld'. 'ld.so' is used by your
compiled binary to dynamically load in shared libraries "on the fly", so you
don't need all that bulky code code in your program, and save on diskspace
and RAM. Two completely seperate programs and functions.

: According to the man page for ld.so, it does implictly look in both /lib and
: /usr/lib, again, unless for some reason it thinks that it needs a static
: lib (and I do NOT have a libm.a).
                           ^^^^^^

You just answered your own question! While linking, you always need these
libXXXXX.[s]a files from /usr/lib. They contain some necassarey routines and
some code to load from /lib/libXXXX.so. If you want shared code, use
the '-N -s' flags on the final compilation (which is actually linking). This
will create significantly smaller binaries. The libXXXX.sa libraries are
just the non-static variants of the normal ones.

As mentioned before, in this stage 'ld.so' is not relevant, you need 'ld' or
better, the man page of 'gcc'.

: Any help would be appreciated. Yes, I am a newbie here, and will need to build
: some rather large X-based clients in the VERY near future (like next week) and
: would appreciate any help I can get.

_Excuse_ me_ , but a Newbie on Linux, needing to build some large X
apps quickly ? Sounds like a contradiction to me... (for any larger X
program, anyway....).

J. v. O.

--
I always cry when I have to reboot Linux to return to DOS... *sniff*
-- 
My PGP public key (you know when, why and how...) :

=====BEGIN PGP PUBLIC KEY BLOCK=====
Version: 2.3
mQCNAi1lYqsAAAEEAMCgUKS7DxyGF8D7QIGYXxRuh2n9Q2+5gIrrb1n9iOl4Xlgo
cO8Y3DE71J5K6WhlpEGDqXZIwY/Xx8mxq80ZHJ3n0pHOUxOQGdxxMT1mrKotjE4Y
wmGqnQhMhpcCKgT/5+5xhuMEluyGQqjyud3PCDogJCC/Sia7eO9+56e/13btAAUR
tC1KLlMuIHZhbiBPb3N0ZW4gPGp2b29zdGVuQGNvbXBpbGVyLnRkY25ldC5ubD4=
=3brb
=====END PGP PUBLIC KEY BLOCK=====

------------------------------

From: jvoosten@compiler.tdcnet.nl (J.S. van Oosten)
Subject: Re: An idea to cut down noise 
Date: Tue, 15 Mar 1994 01:44:40 GMT

Juha Laiho (jlaiho@ichaos.nullnet.fi) wrote:
: ron@draconia.hacktic.nl (Ron Smits) said:
: >I do see another problem. People who cannot connect to the server because
: >they are a uucp site. But maybe the mail-server-program-guru's know how
: >to handle a problem like that.

: UUCP point of view: Why ask from some server that may not have the
: answer, when it takes just as much time to post the question to news and
: wait for the answer?

Because that will prevent millions of people seeing your question, thus
saving a lot of bandwidth and money and reducing noise ? A single connection
or mail to a server won't disturb anyone. If you're an UUCP site, mailing or
posting will make little difference in time. If you're on Internet, posting
will actually take longer ! 

BTW, posting on Usenet won't give any guarantees for an answer either.

: Also, there's a problem with all new users: how could they find
: information about the server before reading the groups and FAQs and
: before posting the first question?

Ah. That's what the FAQs are for ! They could point out that a lot of
knowledge is stored in databases which can be accessed easily. Of course we
wouldn't want comp.os.linux.* to disappear...

I'm afraid this thread doesn't cut down on noise in any way.... :-(

: Testing your 8-bit cleanness/ISO-8859-1 charset: 9 2 3 - can you read me?
                                                   ^^^^^
                                                   Yeah. I see a 9, a 2, a 3
and a dash in reverse. Why ?            

J. v. O.

--
Don't utter the word 'vi' while I'm near. I might just die of a heartattack.
-- 
My PGP public key (you know when, why and how...) :

=====BEGIN PGP PUBLIC KEY BLOCK=====
Version: 2.3
mQCNAi1lYqsAAAEEAMCgUKS7DxyGF8D7QIGYXxRuh2n9Q2+5gIrrb1n9iOl4Xlgo
cO8Y3DE71J5K6WhlpEGDqXZIwY/Xx8mxq80ZHJ3n0pHOUxOQGdxxMT1mrKotjE4Y
wmGqnQhMhpcCKgT/5+5xhuMEluyGQqjyud3PCDogJCC/Sia7eO9+56e/13btAAUR
tC1KLlMuIHZhbiBPb3N0ZW4gPGp2b29zdGVuQGNvbXBpbGVyLnRkY25ldC5ubD4=
=3brb
=====END PGP PUBLIC KEY BLOCK=====

------------------------------

From: jvoosten@compiler.tdcnet.nl (J.S. van Oosten)
Subject: Hangman (was Re: Cyrix 486DLC/40 Compatibility with Linux)
Date: Tue, 15 Mar 1994 01:54:58 GMT

: >     I have been running Linux on a Cyrix 486 for 3 months now with 
: >out much problem.  But I have patch level 14 ( I believe.)  What ever 
: >was latest in Dec. 93.  Does this mean that newer versions of Linux 
: >-----------------------------------------------------------------------
: >Patrick W. Brewer           CATT Alumni  
: >noble@catt.ncsu.edu

:       Yes, it does. But you must

*grin* Is this supposed to be some sort of 'hangman' game ? 

QM: "Patrick, you turn. Try a word."
P:  "hmm, 'compile' ?"
    *BZZZZZZZZZZZZZT*
QM: "That word is not used, unfortunately. I'm afraid you're only missing
     your left leg now. Next candidate please."

*giggle*

J. v. O.

--
I need coffee.
-- 
My PGP public key (you know when, why and how...) :

=====BEGIN PGP PUBLIC KEY BLOCK=====
Version: 2.3
mQCNAi1lYqsAAAEEAMCgUKS7DxyGF8D7QIGYXxRuh2n9Q2+5gIrrb1n9iOl4Xlgo
cO8Y3DE71J5K6WhlpEGDqXZIwY/Xx8mxq80ZHJ3n0pHOUxOQGdxxMT1mrKotjE4Y
wmGqnQhMhpcCKgT/5+5xhuMEluyGQqjyud3PCDogJCC/Sia7eO9+56e/13btAAUR
tC1KLlMuIHZhbiBPb3N0ZW4gPGp2b29zdGVuQGNvbXBpbGVyLnRkY25ldC5ubD4=
=3brb
=====END PGP PUBLIC KEY BLOCK=====

------------------------------

From: jvoosten@compiler.tdcnet.nl (J.S. van Oosten)
Subject: Re: Multiple screen feature in Linux?
Date: Tue, 15 Mar 1994 02:04:05 GMT

Ron Smits (ron@draconia.hacktic.nl) wrote:
: It is exactly the same in Linux, You must start up gettys on the different
: virtual consoles who are named tty[1-8]. You can then switch between
: them with ALT-F[1-8].

: Note of warning, If you run X you must leave one virtual console free for
: X. So you could have seven VC's with a getty and keep the 8th for X.

Another note: to leave X temporarely you need to press Ctrl-Alt-Fx to get
back to normal text mode, then you can use the normal Alt-Fx keycombinations
to switch the VCs', and must use the Alt-F8 or whatever VC was free to
switch back to X... Con-fu-sing... (that's not a Chinese.)

: The kernel has default 8 VC's.

if you want more, change NR_CONSOLES constant in /linux/include/linux/tty.h
and rebuild your kernel. I have 12, handy with debugging multiuser thingies
and keeping an eye on dialout/dialins etc. in the mean time.

If that's not enough, run screens :-).

J. v. O.

--
Sometimes I wonder if this whole universe isn't anything more than a
simulation on a BIG computer. And they just might press Ctrl-C any time
now...
-- 
My PGP public key (you know when, why and how...) :

=====BEGIN PGP PUBLIC KEY BLOCK=====
Version: 2.3
mQCNAi1lYqsAAAEEAMCgUKS7DxyGF8D7QIGYXxRuh2n9Q2+5gIrrb1n9iOl4Xlgo
cO8Y3DE71J5K6WhlpEGDqXZIwY/Xx8mxq80ZHJ3n0pHOUxOQGdxxMT1mrKotjE4Y
wmGqnQhMhpcCKgT/5+5xhuMEluyGQqjyud3PCDogJCC/Sia7eO9+56e/13btAAUR
tC1KLlMuIHZhbiBPb3N0ZW4gPGp2b29zdGVuQGNvbXBpbGVyLnRkY25ldC5ubD4=
=3brb
=====END PGP PUBLIC KEY BLOCK=====

------------------------------

From: lockwood@pan.vlsi.uiuc.edu (John Lockwood)
Crossposted-To: comp.sys.intel
Subject: Re: 486SX vs 486DX - difference in performance using Linux
Date: 15 Mar 94 04:13:17 GMT

doolitt@cebaf4.cebaf.gov (Larry Doolittle) writes:

>Non-obvious programs that make demands on the FPU are GhostScript
>and the InterViews toolkit.  Anybody care to add to this list?

Try drawing arcs with Xfig w/o co-pro.  It is quite a bit slower.
But, then again, how many arcs do you plan to draw with your new machine :)


------------------------------

Crossposted-To: comp.os.linux.help
From: dlv@CUNYVMS1.GC.CUNY.EDU (Dimitri Vulis, CUNY GC Math)
Subject: Why Linux has problems on IBM Valuepoint and PS/1 and how to fix it (I think)
Reply-To: dlv@CUNYVMS1.GC.CUNY.EDU
Date: Tue, 15 Mar 1994 03:14:00 GMT

Folks, a few weeks ago I asked for help installing Linux on an IBM ValuePoint.

The problem was that the kernel was unable to determine the hard geometry
(i.e., the number of cylinders, heads, and sectors) on the ValuePoint (and some
other machines, such as PS/1). I also wanted to know WHY Linux fails where SCO
and OS/2 work fine.

I've received suggestions from various kind people whom I thank again (I hope
I've thanked everyone). The suggestions fell into 2 categories:

1. It turns out that if one holds down Alt (or is it Shift) while Lilo is
booting, it will pause and ask for a boot device, at which point you can enter
device followed by hd=cyl,heads,sectors. Well, my lilo did pause, but rudely
ignored any disk geometry I tried to enter this way. Perhaps I have an old
lilo. Anyway, this is not what I wanted...

2. One can recompile the kernel, hardwiring the disk geometry. This involves
changing, e.g., /usr/src/linux/include/linux/config.h to hardwire the disk
geometry, and writing the custom kernel to a floppy. The procedure is described
by gg10@prism.gatech.EDU (Gregory Lyle Galloway) in the howto file. One defines
#define HD_TYPE { 16,38,683,0,683,8 } in the above file, where the numbers are
your geometry interpreted thus in the file /usr/src/linux/kernel/blk_dev/hd.c:

struct hd_i_struct {
        unsigned int head,sect,cyl,wpcom,lzone,ctl;
        };
#ifdef HD_TYPE
struct hd_i_struct hd_info[] = { HD_TYPE };
static int NR_HD = ((sizeof (hd_info))/(sizeof (struct hd_i_struct)));
#else
struct hd_i_struct hd_info[] = { {0,0,0,0,0,0},{0,0,0,0,0,0} };
static int NR_HD = 0;
#endif

That is, if HD_TYPE is defined, we use it and don't try to determine the
geometry; and if it not defined, then in the function hd_geninit we
copy the above numbers from a certain place in memory:

                hd_info[drive].cyl = *(unsigned short *) BIOS;
                hd_info[drive].head = *(unsigned char *) (2+BIOS);
                hd_info[drive].wpcom = *(unsigned short *) (5+BIOS);
                hd_info[drive].ctl = *(unsigned char *) (8+BIOS);
                hd_info[drive].lzone = *(unsigned short *) (12+BIOS);
                hd_info[drive].sect = *(unsigned char *) (14+BIOS);

Clearly, the C code assumes that (char *)BIOS points to this data:

 number of cylinders (2 bytes)
 number of heads (1 byte)
 garbage (2 bytes) (actually, starting reduced writing current cylinder)
 starting write precompensation cylinder (2 bytes)
 garbage (2 bytes) (actually, maximum ecc burst length)
 control byte (1 byte) (80=disable retries, 40=disable ecc retries, etc)
 garbage (3 bytes) (actually, standard timeout and check drive timeout)
 landing zone (2 bytes)
 number of sectors (1 byte)
 gargabe (4 bytes) (reserved)

Thus, my question remains, where do these numbers come from? It turns out that
Linux's kernel determines the hard disk geometry at boot time in file
/usr/source/linux/boot/setup.s using the following simple code:

! Get hd0 data

        xor     ax,ax           ! clear ax
        mov     ds,ax
        lds     si,[4*0x41]
        mov     ax,#INITSEG
        mov     es,ax
        mov     di,#0x0080
        mov     cx,#0x10
        cld
        rep
        movsb

and likewise for /dev/hd1 with int 0x46 and destination 0090. This code,
executing in real mode, gets DS:SI from 0:4*41 and copies 16 bytes into
INITSEG:80. Later these 16 bytes are used by protected mode code to determine
the drive geometry.

This makes the assumption that the drive geometry data for drive /dev/hd0 can
be found in memory (usually ROM) pointed to by segment:offset stored at the
address 0:4*41, and likewise for /dev/hd1 at 0:4*46. This assumption is valid
on IBM AT, Phoenix BIOS, AMI BIOS, and some other clone BIOSes (even Compaq);
but NOT valid for ValuePoint, PS/1, most PS/2's I've checked this for. On these
machines, the 41 and 46 vectors point to EXECUTABLE code in ROM. Interpreting
it as drive geometry gives rise to strange and meaningless numbers. However
on every machine I tried the code starts with the instructions:

          PUSH AX
          PUSH BX

It seems to me that it would be prudent for the above setup.s code to check for
ds:[si] being the above 2 bytes after the lds instruction, and if it is, to
resort to the "official" way to get the drive geometry endorsed by IBM, which
is to issue INT 13 with AH=8. It is permissible to issue INT to get BIOS data,
as is done elsewhere in setup.s. After the following instructions:

          mov ah,8
          mov dl,#0x80
          int 0x13

(and one would use dl=81 to get /dev/hd1), the following information is
on registers:

ah = 0 if success, 7 (or other non-zero) if failure (e.g., no such drive)

total number of drives        - dl

max head number               - dh (i.e., the number linux wants is obtained
                              by saying inc dh)


max cylinder number           - split between cl and ch, so the
                              number linux wants can be obtained by saying:
                              and cl,#0xc0
                              rol cl,1
                              rol cl,1
                              xchg cl,ch
                              inc cx

max sector number           - in cl, number of sectors is obtained by saying
                              and cx,#0x003f
                              inc cx

Again, on many clones (but NOT on valuepoints, ps/2's, etc) es:di points to the
drive parameters after the int 13, but on a vp we must compose this record in
memory. When the control byte is unavaileble, hd.c obtains it by saying:

hd_info[...].ctl = (hd_info[...].head > 8 ? 8 : 0);

I think that adding the above code to setup.s would be a very good idea.

Any other suggestions?
Dimitri Vulis
CUNY GC Math
DLV@CUNYVMS1.BITNET DLV@CUNYVMS1.GC.CUNY.EDU

Disclaimer: my Usenet postings don't necessarily represent anyone's views,
especially my own and/or CUNY's.

------------------------------


** FOR YOUR REFERENCE **

The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:

    Internet: Linux-Misc-Request@NEWS-DIGESTS.MIT.EDU

You can send mail to the entire list (and comp.os.linux.misc) via:

    Internet: Linux-Misc@NEWS-DIGESTS.MIT.EDU

Linux may be obtained via one of these FTP sites:
    nic.funet.fi				pub/OS/Linux
    tsx-11.mit.edu				pub/linux
    sunsite.unc.edu				pub/Linux

End of Linux-Misc Digest
******************************
