Hello Folks,

I got my hands on a pretty dated Asus P2L97-DS board and two P2 cpus
and I'm having a go at smp now since this would serve me nicely as a
server for a private multi-tier project.

Unfortunately I ran into a problem getting an smp kernel to work with
that hardware.
After a little messing around with various kernels I can reproduce the
following quite weird behaviour. Every time I install a new kernel
(smp), using either a standard debian 2.4.17-686-smp kernel package or
a self compiled 2.4.17-686-smp (make-kpkg) one, the very first bootup
after installation will work just fine. "cat /proc/cpuinfo" shows two
cpu's and all's well.

However, I do actually switch that thing off now and again, and every
subsequent startup fails/crashes right at the beginning of the boot
process. After reinstalling a kernel it will boot fine once and so on
and so on...

So technically it is working but practically it is not, not good.

I am using a standard debian/sarge installation with the grub boot
loader and don't really know how to tackle this problem.

Are there any logfiles about this early stage of the boot process (no
file systems mounted yet), I think it crashes around the "calibrating
delay loop" stage?
I noticed there are config files in the /boot folder. Might they be
altered by the system, some sort of self-configuration process, for the
worse in this case?

I tried to find related information in the smp docs, faqs and the usual
forums (including usenet) but I couldn't find anything. I think I can
also rule out a problem with the actual kernel compilation since the
debian packaged kernel shows the same behaviour.

Any suggestion would be appreciated, thx in advance.

cheers

Carsten

Carsten Martens <leonid_67@hotmail.com> wrote:
> Unfortunately I ran into a problem getting an smp kernel to work with
> that hardware.
> After a little messing around with various kernels I can reproduce the
> following quite weird behaviour. Every time I install a new kernel
> (smp), using either a standard debian 2.4.17-686-smp kernel package or
> a self compiled 2.4.17-686-smp (make-kpkg) one, the very first bootup
> after installation will work just fine. "cat /proc/cpuinfo" shows two
> cpu's and all's well.

Well ... so it works.

> However, I do actually switch that thing off now and again, and every
> subsequent startup fails/crashes right at the beginning of the boot
> process. After reinstalling a kernel it will boot fine once and so on
> and so on...

Then your bios has a problem. Do a COLD boot. Or a WARM boot. Whatever
- whichever is approapriate. And use a different kernel.

> So technically it is working but practically it is not, not good.

No - something else is wrong. It sounds like the bios is peculiar and
needs some setup. Run a different kernel.

And these P2's ... are you sure they are SMP compatible? Some are not
- particularly some celeron P2s are not, but one can generally
reestablish the cut traces.

> I am using a standard debian/sarge installation with the grub boot
> loader and don't really know how to tackle this problem.

Why not? You simply use a different kernel until you find one that
works! Vary ANYTHING, systematically. "The kernel" springs to mind.

> Are there any logfiles about this early stage of the boot process (no
> file systems mounted yet), I think it crashes around the "calibrating

No. What you see is all.

> delay loop" stage?

That is right at the start. It means that the cpudoes not work. get
yourself a new one, or make sure to use a kernel that is compatible -
are you sure you have chosen a compatible kernel?

> I noticed there are config files in the /boot folder. Might they be
> altered by the system, some sort of self-configuration process, for the
> worse in this case?

Who cares! You haven't got to the stage of looking at them.

> I tried to find related information in the smp docs, faqs and the usual
> forums (including usenet) but I couldn't find anything. I think I can
> also rule out a problem with the actual kernel compilation since the
> debian packaged kernel shows the same behaviour.

Then you can rule it IN. Change your kernel until you see a change. Or
change your processor until you see a change. Or change your mobo until
you see a change. Whatever.

> Any suggestion would be appreciated, thx in advance.

Recompile with basic P2 support SMP, turn of apic, and anything else.
Check what's known about your board and linux.

Peter

In comp.os.linux.setup Carsten Martens <leonid_67@hotmail.com>:
> Hello Folks,

> I got my hands on a pretty dated Asus P2L97-DS board and two P2 cpus
> and I'm having a go at smp now since this would serve me nicely as a
> server for a private multi-tier project.

> Unfortunately I ran into a problem getting an smp kernel to work with
> that hardware.
> After a little messing around with various kernels I can reproduce the
> following quite weird behaviour. Every time I install a new kernel
> (smp), using either a standard debian 2.4.17-686-smp kernel package or
> a self compiled 2.4.17-686-smp (make-kpkg) one, the very first bootup

Hope this is a recent distro update kernel, sounds pretty old.

> after installation will work just fine. "cat /proc/cpuinfo" shows two
> cpu's and all's well.

> However, I do actually switch that thing off now and again, and every
> subsequent startup fails/crashes right at the beginning of the boot
> process. After reinstalling a kernel it will boot fine once and so on
> and so on...

Sounds like a bios problem, I'd check first the vendor homepage
for the latest version, usually discontinued products or so and
install the latest available if you aren't running this version
already.

Cleaning the system from all this dust inside, reseating
memory/slots can sometimes help in addition.

Good luck

--
Michael Heiming (X-PGP-Sig > GPG-Key ID: EDD27B94)
mail: echo zvpunry@urvzvat.qr | perl -pe 'y/a-z/n-za-m/'
#bofh excuse 298: Not enough interrupts

Carsten Martens wrote:
> Hello Folks,
>
> I got my hands on a pretty dated Asus P2L97-DS board and two P2 cpus
> and I'm having a go at smp now since this would serve me nicely as a
> server for a private multi-tier project.
>
> Unfortunately I ran into a problem getting an smp kernel to work with
> that hardware.
> After a little messing around with various kernels I can reproduce the
> following quite weird behaviour. Every time I install a new kernel
> (smp), using either a standard debian 2.4.17-686-smp kernel package or
> a self compiled 2.4.17-686-smp (make-kpkg) one, the very first bootup
> after installation will work just fine. "cat /proc/cpuinfo" shows two
> cpu's and all's well.
>
> However, I do actually switch that thing off now and again, and every
> subsequent startup fails/crashes right at the beginning of the boot
> process. After reinstalling a kernel it will boot fine once and so on
> and so on...
>
People have been suggesting kernel problems, hardware problems, and BIOS
problems. Of course they may be right. But another possibility is that you
have several kernels lying around, and that your GRUB or LILO are configured
wrongly and they are running a defective kernel by default. I suggest
checking this because if you install the kernel and it works, that tells me
the kernel is OK with your hardware, but when you reboot, you are getting
something else.

I am running Red Hat Enterprise Linux 3 with the 2.4.21-32.0.1.ELsmp kernel
(their latest) that is considerably newer than yours, though by no means
leading edge.

--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ PGP-Key: 9A2FC99A Registered Machine 241939.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 20:15:00 up 41 days, 14:08, 3 users, load average: 4.20, 4.12, 4.04

FWIW: I am having similar issues with my dual P2 running FC3. For me,
it seems to go through most of the boot process, but right before it
displays the login screen (if running X startup; else just the login
prompt), the system freezes. At that point I can only do a hard reset,
and try again. It usually takes 3 to 6 tries before it will
successfully complete the boot, and then all runs well. In the kernel
log I get the error:

WARNING: Kernel Errors Present
vesafb: probe of vesafb0 failed with error -6...: 3 Time(s)

This problem does not occur if I run the single CPU kernel.

As for hardware (CPU/motherboard) issues: this system ran several
years with MS Windows 2000, and then RH9 for a couple more years after
that with no problems. So, according to my deductive reasoning,
something changed in FC3 that causes this.

I haven't tried FC4 yet, since that seems to have broken other things,
and I don't feel comfortable with upgrading my main server just yet. I
would be very interested if anyone has any real information about this
problem though.

Just my dual cents worth.

ken AT softsteps DOT com


On 26 Jul 2005 12:29:11 -0700, "Carsten Martens"
<leonid_67@hotmail.com> wrote:
>Hello Folks,
>
>I got my hands on a pretty dated Asus P2L97-DS board and two P2 cpus
>and I'm having a go at smp now since this would serve me nicely as a
>server for a private multi-tier project.
>
>Unfortunately I ran into a problem getting an smp kernel to work with
>that hardware.
>After a little messing around with various kernels I can reproduce the
>following quite weird behaviour. Every time I install a new kernel
>(smp), using either a standard debian 2.4.17-686-smp kernel package or
>a self compiled 2.4.17-686-smp (make-kpkg) one, the very first bootup
>after installation will work just fine. "cat /proc/cpuinfo" shows two
>cpu's and all's well.
>
>However, I do actually switch that thing off now and again, and every
>subsequent startup fails/crashes right at the beginning of the boot
>process. After reinstalling a kernel it will boot fine once and so on
>and so on...
>
>So technically it is working but practically it is not, not good.
>
>I am using a standard debian/sarge installation with the grub boot
>loader and don't really know how to tackle this problem.
>
>Are there any logfiles about this early stage of the boot process (no
>file systems mounted yet), I think it crashes around the "calibrating
>delay loop" stage?
>I noticed there are config files in the /boot folder. Might they be
>altered by the system, some sort of self-configuration process, for the
>worse in this case?
>
>I tried to find related information in the smp docs, faqs and the usual
>forums (including usenet) but I couldn't find anything. I think I can
>also rule out a problem with the actual kernel compilation since the
>debian packaged kernel shows the same behaviour.
>
>Any suggestion would be appreciated, thx in advance.
>
>cheers
>
>Carsten