This is more of an interesting project but I have find of hit a dead end.

I am playing with an old version of SuSE (9.3) to make a cut down
firewall/router from an AST Bravo LC 5133 machine. This is one of those
old BIOS 1024 cylinder no LBA things with the 72 pin EDO RAM in it..! It
doesnt have a bootable CD and I am using a disk larger than what the PC
BIOS supports.

I first tried the 9.3 install disks (floppies) but after loading the lot
the screen just went into fast scroll. I figured it was a memory
limitation (32MB) even though swap was already available. I then started
playing with building the disk on a PIII/500 but when I transfer it to
the AST box it fails about where the point that (maybe) the IDE chipset
is probed or it cant find the root partition. The first error that comes
straight after "Freeing unused kernel memory" is [1] Illegal instruction
mkdir..... After that it gets a pile of mknod and insmod failures

To prove a point I put 32MB (only) into the PIII and it booted up
completely.

I did rebuild the kernel for fun (made it smaller) and for all tests did
a mkinitrd to provide both the piix and sis5311 drivers. I also added
the IDE generic driver. That didnt help. I note that in one location the
AST Bravo is spec'd (online) as having a 440TX chipset but physically
the box I have has the SIS chipset! I think i covered all bases!

Am trying to establish whether it is IDE driver related, a disk geometry
issue or something else. I didnt want to weigh down the NG with all the
detail in case I went off on the wrong tangent.

I had a short play with grub and bios geometry with no success.

Any thoughts appreciated (that dont involve using a later PC, updating
the BIOS, or using another purpose distro)

Bob

On Sun, 24 Aug 2008 10:44:16 -0500
Bob Bob wrote:

>I had a short play with grub and bios geometry with no success.

Since you didn't specify *what* you did as far as BIOS geometry, I will
take a blind stab at it.

Partition the hard drive with an initial partition of about 128 MB
(plenty of room for extra kernels) as the '/boot', so the BIOS
limitation is neutralized. Make the second partition the swap partition
(whatever size you like, I won't get into the swap size argument here),
and the rest of the disk partitioned for '/'. Make sure you specify
those two partitions for /boot and for /, in case you partitioned it
outside of the install process. Then, during the install, install GRUB
to the MBR. See if that makes any difference . . .


--
Kevin Nathan (Arizona, USA)
Linux Potpourri and a.o.l.s. FAQ -- (temporarily offline)

Open standards. Open source. Open minds.
The command line is the front line.
Linux 2.6.25.11-0.1-pae
9:38am up 5 days 13:27, 19 users, load average: 0.51, 0.39, 0.36

Hi Kevin

Tnxs your comment. If I make the post too long it wont get read! I am
not so much after a specific solution but a bank of ideas to develop on.
ie What have I missed that I should have checked?

The native disk geometry is 4092/16/63. The partitions are arranged in
order boot/swap/root. Both boot and swap are well within the first 1024
cylinders. I tried configuring the AST BIOS for auto (1024/63/XX) and
whatever gave me 16 heads. It is of course booting LBA on the PIII box.
I did make an abortive attempted to use "large" BIOS settings on the
PIII and couldnt get it to boot despite starting with the install CD and
rewriting the MBR etc. I cant say however that I have gone down this
path enough to prove it isnt the cause of the problem.

It has about 30MB of boot and maybe 256 of swap. (I fell into a low
memory problem once before with depmod needing 220 or so to run.) Total
space is about 2.1GB.

I am getting the impression its a 2.6 problem with support for SIS 5311
chipsets. I have seen a few web comments on that, but nothing overly
helpful. One could argue that if it boots the kernel then the controller
has beens seen, but it probably does that all with BIOS calls.

I see two possible causes;
- A 2.6/5513 controller support problem. I never see a succesful probe
in the dmesg like output. I am aware that you can make a manual probe at
the kernel line but how to force its use?
- A geometry issue such that when root gets mounted the kernel cant find
it. I did however do a grub setup/install on the AST box and it quite
happily found the stage files. ie it recognizes all entrees in the
partition table and can certainly mount and read /boot.

Any further ideas appreciated.

Bob


Kevin Nathan wrote:
> On Sun, 24 Aug 2008 10:44:16 -0500
> Bob Bob wrote:
>
>> I had a short play with grub and bios geometry with no success.
>
> Since you didn't specify *what* you did as far as BIOS geometry, I will
> take a blind stab at it.

On Sun, 24 Aug 2008 14:47:21 -0500
Bob Bob wrote:

>The native disk geometry is 4092/16/63. The partitions are arranged in
>order boot/swap/root. Both boot and swap are well within the first
>1024 cylinders.

/boot is the only one that you need to have with the first 1024
cylinders. Once GRUB has loaded the kernel, the kernel takes over the
drive detection.


>It has about 30MB of boot and maybe 256 of swap. (I fell into a low
>memory problem once before with depmod needing 220 or so to run.)
>Total space is about 2.1GB.
>

2.1GB is a bit small, but I am assuming you are only installing a non-X
system, minimal install. Make sure you start with the absolute minimum
of software to be installed -- some packages can pull in a lot of
dependencies. With 9.3, you should be able to do about a 500MB install,
IIRC, without too much 'trimming'.


>I am getting the impression its a 2.6 problem with support for SIS
>5311 chipsets. I have seen a few web comments on that, but nothing
>overly helpful. One could argue that if it boots the kernel then the
>controller has beens seen, but it probably does that all with BIOS
>calls.
>

In general, the kernel does not trust BIOS settings. To see what kinds
of things over which you have control at boot time, see:

/usr/src/linux/Documentation/kernel-parameters.txt


>I see two possible causes;
>- A 2.6/5513 controller support problem. I never see a succesful probe
>in the dmesg like output. I am aware that you can make a manual probe
>at the kernel line but how to force its use?

Of that I am not sure -- there might be some help in the kernel
documentation directory (specified above). If you don't have it on
another box, make sure you have the kernel source installed.


>- A geometry issue such that when root gets mounted the kernel cant
>find it.

I would tend to think that is not the problem. Rather than a geometry
issue, it could be a disk error problem. Re-reading your first post, I
realized I missed this section:

>The first error that comes straight after "Freeing unused kernel
>memory" is [1] Illegal instruction mkdir..... After that it gets a
>pile of mknod and insmod failures
>

which seems to indicate that you have a problem with the disk. It could
be a physical error or, for some reason, the disk is mounted read-only.
I've seen the read-only problem a few times at work on some client
Kubuntu boxes (could not create files and dirs in /dev and/or /var
directories at boot) and an fsck fixed it most of the time. When it
didn't, it was a physical error on hard disk -- when we had time to
check.

If you can, put the drive in another box and force an fsck on each
partition. If there is no problem there, completely wipe the partition
table and rebuild it during the install on the target machine.


--
Kevin Nathan (Arizona, USA)
Linux Potpourri and a.o.l.s. FAQ -- (temporarily offline)

Open standards. Open source. Open minds.
The command line is the front line.
Linux 2.6.25.11-0.1-pae
9:46pm up 6 days 1:36, 19 users, load average: 0.20, 0.45, 0.59

Hi Kevin

The 2GB is small if I want X. It is however only running the CLI. I did
indeed start with the standard minimal install and aded a few packages
from there. Biggest problem was recompiling a kernel. During module
install they exist in two places. I had to prune a few things before the
install but afterwards a make clean got me back a lot of space.

I had a further look at the 2.6 issue. Quite a few Ubuntu users seem to
have had trouble with it. Even to the point of "I upgraded my kernel to
2.6.x and it no longer boots". The 5513 kernel driver not only covers
their older products but the current stream as well. Apparently the
5513 as it was back 10 years ago is still used in one of the bridge
chips even to the point of having then same PCI ID. I had a brief look
at the 5513 kernel source code but found no obvious "we have disabled"
comments.

I tried a kernel 2.4 install and it worked fine. This partly infers it
may be a kernel problem.

Whatever the case I have now given up on it and put DSL on it for sale
at a local non profit store. Have to expose Linux to as mamy people as
possible!

All your other comments noted, thanks. I shd point out however that the
disk works fine in the PIII/500 box with piix. When plugged into the
older box it deosnt succsufully probe the 5513 like the 2.4 kernel
does.. More proof it is kernel related.. I cant actually install the 9.3
software on the older box as it has no boot CD capability and the 9.3
floppies send it into screen rolls...

Thanks for your input

Cheers Bob (Australian in East Texas)

Kevin Nathan wrote:
> /boot is the only one that you need to have with the first 1024
> cylinders. Once GRUB has loaded the kernel, the kernel takes over the
> drive detection.

On Mon, 25 Aug 2008 18:56:09 -0500
Bob Bob wrote:

>I tried a kernel 2.4 install and it worked fine. This partly infers it
>may be a kernel problem.
>

That would be my guess, as well. Thanks for the feedback.

As for spreading Linux, I'm slowly getting my relatives down here on
it. I refuse to work on their Windows machines but will help them if
they want to switch to Linux. ;-)


--
Kevin Nathan (Arizona, USA)
Linux Potpourri and a.o.l.s. FAQ -- (temporarily offline)

Open standards. Open source. Open minds.
The command line is the front line.
Linux 2.6.25.11-0.1-pae
8:30pm up 7:19, 16 users, load average: 1.19, 1.42, 1.01

Kevin Nathan wrote:
> As for spreading Linux, I'm slowly getting my relatives down here on
> it. I refuse to work on their Windows machines but will help them if
> they want to switch to Linux. ;-)

I do the same and now have no friends anymore. :-D
Seriously though, allmost all my friends either don't HAVE a pc or run
Linux and just one I would call a 'geek'

One even has no idea how a PC works, except for turning it on and off.

houghi
--
The whole principle [of censorship] is wrong. It's like demanding that
grown men live on skim milk because the baby can't have steak.
-- Robert A. Heinlein in "The Man Who Sold the Moon"

On Tue, 26 Aug 2008 09:34:42 +0200
houghi wrote:

>I do the same and now have no friends anymore. :-D

Ah, but you see, I don't have any friends to begin with, so I have to
torture the relatives! ;-)


>Seriously though, allmost all my friends either don't HAVE a pc or run
>Linux and just one I would call a 'geek'
>

No one I know down here (except for my boss) knows anything about
computers -- which is why it's so easy to get them on Linux. I only
have two cousins left still on Windows. Their loss . . .


>One even has no idea how a PC works, except for turning it on and off.
>

I'd pretty much have to say that applies to all my local relatives, too.


--
Kevin Nathan (Arizona, USA)
Linux Potpourri and a.o.l.s. FAQ -- (temporarily offline)

Open standards. Open source. Open minds.
The command line is the front line.
Linux 2.6.25.11-0.1-pae
6:52pm up 1 day 5:41, 16 users, load average: 0.09, 0.14, 0.35