Very Computer

OK, I know this has been hashed out here a bunch of times, but I'm
stuck.

I have a P5/166 with a 6 GB Quantum Bigfoot drive.  I tried to install
RH 6.0 on it.  Now I get LI and the system hangs.

The system will boot fine off a floppy.  I have checked all the obvious
stuff:  

check lilo.conf and rerun lilo once I get the system booted from a
floppy - no errors, no problems...

the entire boot partition is under the 1024 cyl. limit.

no obvious hardware problems.  This disk has been in use for a while and
it worked fine in this machine before.

I reinstalled using RH's default workstation just to make sure I did not
do somethign funky with the partition table - no dice.

Any other suggestions?

--Yan

On Sun, 05 Dec 1999 06:19:36 -0500,

Quote:>OK, I know this has been hashed out here a bunch of times, but I'm
>stuck.

>I have a P5/166 with a 6 GB Quantum Bigfoot drive.  I tried to install
>RH 6.0 on it.  Now I get LI and the system hangs.

[snip...]

Quote:>I reinstalled using RH's default workstation just to make sure I did not
>do somethign funky with the partition table - no dice.

>Any other suggestions?

Yes, I had such a problem once. This was after a reinstall (or was it an
upgrade? not sure), and I reinstalled lilo several times to no avail.

It turns out that there was an old lilo-installation in the MBR, which was
now disfunctional, while I was trying to install it in the boot sector of my
Linux file system partition. Thus, the disfunctional incarnation got control
and my current lilo-installation was never started. Now, there are two
solutions to this problem:

1) Boot DOS, and run FDISK /MBR. Then, set your linux partition as the
primary/boot partition, with fdisk. Probably safest to do this through the
Linux fdisk. Now, run lilo to install it in the boot sector of that partition.

2) Change lilo.conf so that it installs it in the MBR, and re-run lilo. Now,
your new lilo configuration will be there instead of the one not working.

By the way, the syntax for installing lilo in MBR is:

boot=/dev/hda
(or /dev/sda if you're running SCSI). This line should be in the section
before listing the various OS/boot alternatives.

For installing it in the boot sector of the Linux partition, the same line
should read:
boot=/dev/hda1
(or whatever is your linux partition)

This was my suggestion to what might be wrong. There might be something else
wrong.

- Vegard

On Sun, 05 Dec 1999 06:19:36 -0500,

Quote:> OK, I know this has been hashed out here a bunch of times, but I'm
> stuck.

> I have a P5/166 with a 6 GB Quantum Bigfoot drive.  I tried to install
> RH 6.0 on it.  Now I get LI and the system hangs.

> The system will boot fine off a floppy.  I have checked all the obvious
> stuff:  

> check lilo.conf and rerun lilo once I get the system booted from a
> floppy - no errors, no problems...

> the entire boot partition is under the 1024 cyl. limit.

> no obvious hardware problems.  This disk has been in use for a while and
> it worked fine in this machine before.

> I reinstalled using RH's default workstation just to make sure I did not
> do somethign funky with the partition table - no dice.

> Any other suggestions?

PC hardware sucks.  A lot.

One of the sucky things about is the games that the some BIOSes play
with the notion of Cylinder/Head/Sector mapping.  In order to deal with
"large" (ie, > 500M) drives, they play silly games with logical
remapping of the drive geometry.  The drive will tell Linux that it has
a certain geomtery and therefore the kernel will be starting at a certain
block determined by cyl/head/sector, but the BIOS will have an entirely
different view of things and in its view, it will be at an entirely
different "virtual" cyl/head/sector.

You can usually get around this by using the 'linear' option in
/etc/lilo.conf.  This will use an alternative interface to the BIOS and
instead of asking for a given cylinder/head/sector, will ask for a
certain block number and let the BIOS figure out where that is.  This is
effectively sneaking around any sort of remapping the BIOS does
internally.

--
Brian Moore                       | Of course vi is God's editor.
      Sysadmin, C/Perl Hacker     | If He used Emacs, He'd still be waiting
      Usenet Vandal               |  for it to load on the seventh day.
      Netscum, Bane of Elves.

This could be caused by having something else in the MBR.  It happened to me
because I've got an overlay program in it.  I got around it by installing
LILO to the first section of the partition (an option in the installation of
RH6.1.  I don't know how it would be done on a different distribution,
though).  Then I would just use a boot disk when I wanted to boot linux, or
just boot up normally if I wanted to boot windows.  Not very convient, but
it works.

Good luck.

Ken

Anyone suggest adding a single line with the single word 'linear' in
your /etc/lilo.conf ???

Sounds crazy, but works on my old screwball Compaq HDs (otherwise I get
'LI' also)!

Jeff
--
 ( >-    Jeff Rose - Everyone's Linux User Group (ELUG)     -< )

|  \)          *** * Linux/IT Writer ***            (/  |
 |_|_                 eFAX: +1.630.604.4130                 _|_|

Quote:

>You can usually get around this by using the 'linear' option in
>/etc/lilo.conf.  This will use an alternative interface to the BIOS and
>instead of asking for a given cylinder/head/sector, will ask for a
>certain block number and let the BIOS figure out where that is.  This is
>effectively sneaking around any sort of remapping the BIOS does
>internally.

I guess that is wishful thinking.  What happen with the linear option is
that lilo makes a bios call to get the disk geometry, and with that
information it convert all linear addresses to CHS format before calling
the bios disk read function.

That means cyl1024 is still very much an issue.

But check this out from comp.os.linux.development.system:

::
:: I have submitted a new version of LILO (beyond v. 21) to its author,
:: Werner Almsberger, with the hopes that it will be released as version
:: 22.  It adds a new directive which allows booting beyond 1024
:: cylinders, as long as your BIOS supports the Enhanced Disk Drive BIOS
:: calls.
::
:: I use it on a daily basis to dual boot RedHat 6.1 and Windows98se, on
:: a system with 27GB and 17GB hard drives.  It has been tested on two
:: older systems to make sure that all previous functionality is still
:: working.
::
:: Additional beta testing, comments, and other feedback is sought.  The
:: source code for egcc 2.8 (RH 6.0) is available at:
::
::    ftp://sd.dynhost.com/pub/linux/lilo
::
:: Please e-mail me if you download this package.  The new documentation
:: is in the CHANGES.txt and man pages.
::
:: --John Coffman


::

--
Villy

On 6 Dec 1999 09:09:28 GMT,

No, it doesn't.  It asks the drive itself what the geometry is, and this can
differ from what the BIOS presents.  Hence the problem.

You are deceived if you believe that the BIOS has anything to do with the
operation of Linux once it has booted.

Quote:> That means cyl1024 is still very much an issue.

Which of course has nothing to do with what I said, or what the users
problem was.

LILO will notify you upon installation if the Cyl is >1024 and will refuse
to install it.  In this case, though, the user said that LILO ran just fine,
installing itself into the boot record, and as per the subject of this
thread, upon booting it said "LI"

This situation is quite well documented in the LILO documentation:
|   LI   The first stage boot loader was able to load the second stage boot
|     loader, but has failed to execute it. This can either be caused by a
|     geometry mismatch or by moving /boot/boot.b without running the map
|     installer.

Again, the 'linear' option is to address this issue and it is your own
wishful thinking and misreading the articles that makes you believe that
this has anything to do with 1024 cylinders.

Specifically, again from the LILO documentation, note the description
of the 'linear' option and why you would use it:

|  LINEAR  Generate linear sector addresses instead of sector/head/cylinder
|    addresses. Linear addresses are translated at run time and do not
|    depend on disk geometry. Note that boot disks may not be portable if
|    LINEAR is used, because the BIOS service to determine the disk geometry
|    does not work reliably for floppy disks. When using LINEAR with large
|    disks, /sbin/lilo may generate references to inaccessible disk areas
|    (see section "BIOS restrictions"), because 3D sector addresses are not
|    known before boot time. LINEAR may conflict with COMPACT, see section
|    "Other problems".

'LI' as the boot message is typically a geometry mismatch, and 'linear' use
a method that doesn't depend on geometry.... hrrm, one would think that
could be related, no?

Please read the articles before you post your misdiagnosis.

--
Brian Moore                       | Of course vi is God's editor.
      Sysadmin, C/Perl Hacker     | If He used Emacs, He'd still be waiting
      Usenet Vandal               |  for it to load on the seventh day.
      Netscum, Bane of Elves.