From: Nick Leverton on
In article <slrngigcop.j9k.devnull(a)feynman.home.woodall.me.uk>,
Tim Woodall <news003(a)woodall.me.uk> wrote:
>Since upgrading to lenny I've got one (old pentium 60) that will not
>boot with kernel 2.6.26. It's booting fine with 2.6.18 from Etch.
>
>It gets as far as Parsing ELF... Done and then it reboots. I don't think
>it's even printing "now booting the kernel".
>
>Anyone got any ideas?

Are you using lilo or grub ? If lilo, check whether some recent
update has "helpfully" added the large-memory option to lilo.conf, and
remove it if so.

I've just been bitten by this on my AMD K7 system. I updated LILO a
week ago, but didn't reboot at the time as there was no kernel update.
I vaguely remember that debconf might have asked me abuot making this
change to lilo.conf - but since my BIOS is dated 2003 and it referred to
the risks of the change as being limited to pre-2001 systems I thought
I was safe.

Today I did another update and it brought in a new 2.6.26 kernel - which
kept rebooting immediately after loading the initial kernel and initrd,
similarly to what you are experiencing.

The changelog in /usr/sbin/update-lilo says "Use large-memory by default,
hoping that BIOSes now days aren't so broken." But as mentioned my BIOS
is dated 2003 not 2001 yet I still experience this problem. Read more
background in /usr/share/doc/lilo/README.Debian

If you're using grub, I don't know whether it has a similar option but
it might be worth looking into.

Nick
--
Serendipity: http://www.leverton.org/blosxom (last update 19th September 2008)
"The Internet, a sort of ersatz counterfeit of real life"
-- Janet Street-Porter, BBC2, 19th March 1996
From: Mark Hobley on
Robert Harris <robert.f.harris(a)blueyonder.co.uk> wrote:

> Try installing and booting from the package:
>
> linux-image-486

It wont make any difference. There is a bug in the current kernel as
being build by the distributers. The following distributions are
affected:

Debian (Lenny)
Ubuntu
Gentoo
T2
Rock Linux

and probably others.

The error is not limited to Intel Pentium processors. It affects the Cyrix 686
and AMD K5, K6, and K7 series computers.

I found that by compiling my own vanilla kernel from a Gentoo based
system running on a true i586 machine, the bug does not appear.

There are a couple of patches that provide cpuinfo falsification on
newer machines, so this may be useful to system builders wanting to build i586
compatible binaries.

The patches are:

Install a uname_hack kernel module from:

http://ftp.cross-lfs.org/pub/clfs/conglomeration/uname_hack/

uname -m
i686 <--- Damn! That is no good!
insmod uname_hack.ko
uname -m
i486 <-- Excellent. That will work just fine with a Pentium.

Falsify the /proc/cpuinfo file using a bindmount for lowest common
denominator compatibility:

Create a file /etc/cpuinfo as follows:

processor : 0
vendor_id : GenuineIntel
cpu family : 4
model : 2
model name : 486
stepping : 6
cpu MHz : 120.274
cache size : 0 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr mce cx8
bogomips : 241.80
clflush size : 32
power management:

Now mount the file using a bindmount:

mount --bind /etc/cpuinfo /proc/cpuinfo

I haven't tried to build the Debian supplied kernel from source using
these patches yet, but I only discovered these this morning.

Regards,

Mark.

--
Mark Hobley
Linux User: #370818 http://markhobley.yi.org/

From: Nick Leverton on
In article <h37ov5-0kd.ln1(a)neptune.markhobley.yi.org>,
Mark Hobley <markhobley(a)hotpop.donottypethisbit.com> wrote:

>It wont make any difference. There is a bug in the current kernel as
>being build by the distributers. The following distributions are
>affected:
>
>Debian (Lenny)
>Ubuntu
>Gentoo
>T2
>Rock Linux
>
>and probably others.
>
>The error is not limited to Intel Pentium processors. It affects the Cyrix 686
>and AMD K5, K6, and K7 series computers.

Can you give any more details of the problem ? I'm running Lenny
(2.6.16-1-686 version 2.6.26-10) on an AMD K7 and it seems to be working
OK, though I did have to remove the lilo.conf change I referred to in
a previous posting.

Nick
--
Serendipity: http://www.leverton.org/blosxom (last update 19th September 2008)
"The Internet, a sort of ersatz counterfeit of real life"
-- Janet Street-Porter, BBC2, 19th March 1996
From: Mark Hobley on
Nick Leverton <nick(a)leverton.org> wrote:

> Can you give any more details of the problem ?

I don't know specifically where the problem lies, but I believe that the
problem is due to invalid instructions appearing in the compiled kernel.
A "Bug Int 6" error occurs at boot up and the system halts before the
init process starts.

I don't know how to disassemble the distribution supplied compressed
kernel image into an assembly language listing, so I haven't been able
to examine the code at the point of occurrence.

I guess, that as the kernel is being built at distribution, the i686
architecture of the distribution build machine is being detected at
compile time. This is causing invalid instructions on recipient machines
that are fitted with older generation processors.

> I'm running Lenny (2.6.16-1-686 version 2.6.26-10) on an AMD K7 and it
seems to be working OK

Yeah. It seems to be working on my AMD K7 too, but some of my other
older machines are screwed. (I use mainly Intel Pentium 120 machines,
but some computers have AMD K5, K6, and Cyrix 686 processors fitted).

Unfortunately, the faulty kernel is on the distribution boot disks, so I
am unable to boot into a rescue console. What I had to do was remove the
hard disks containing the problematic kernel and clone the working hand
rolled kernel and base system from a working test machine.

(I haven't yet learned how to make a bootable rescue disk images.)

Regards,

Mark.

--
Mark Hobley
Linux User: #370818 http://markhobley.yi.org/

From: Tim Woodall on
On Sun, 23 Nov 2008 15:08:12 GMT,
Mark Hobley <markhobley(a)hotpop.donottypethisbit.com> wrote:
> Tim Woodall <devnull(a)woodall.me.uk> wrote:
>> Since upgrading to lenny I've got one (old pentium 60) that will not
>> boot with kernel 2.6.26.
>
> Tim. If you are in the Midlands or Staffordshire area, I have got a few
> Pentium 90 and Pentium 100 motherboards behind me. You can have them, if
> you want to upgrade that machine.
>
> I travel a regular orbit (Quinton, Aldridge, Streetly, Four Oaks,
> Tamworth, Hammerwich, Burntwood, Cannock, Walsall), so any of those
> areas are good for me. Just let me know if you want them.
>

Thanks, but it's not necessary. (I'm from that area - I still own a
house in Willenhall that I let out - but I'm working in central London
now) It's my plan to replace this machine with a low power fanless
machine. It is just my firewall and doesn't do anything else at all.

I started having a second problem yesterday - "UPS data stale, UPS
connected, UPS data stale, UPS connected, Battery Low, Starting shutdown
now" so I've upgraded it to a Pentium III (Coppermine) with 256MB ram
that I had lying around. That's been working fine.

Thanks for your other comments as well. If I have time I'll do some
experiments.

Tim.

p.s. I have some scripts that build me a rescue and install disk as a
bootable CD (particularly useful to mail to people so they can boot
their machine from the CD and then I can login remotely)

My "installer" consists of an ext3 dump of the minimum packages possible
in Debian (without doing major surgery).

If you're interested, I've just uploaded my scripts to
http://www.woodall.me.uk/installer/

You will definitely have to make changes for it to work for you - in a
number of places I've got hard coded references to my local debian
mirror. There are probably other things that will need fixing as well.

The script setup.sh is the one I run when booted from the CD image that
the other scripts build and it will "blast away everything on hda and
install my system instead" so beware running it.



--
God said, "div D = rho, div B = 0, curl E = - @B/@t, curl H = J + @D/@t,"
and there was light.

http://www.woodall.me.uk/