A7M266-D, dual athlon 1800+ kernel-smp APIC boot problem workaround

A7M266-D, dual athlon 1800+ kernel-smp APIC boot problem workaround

[Sean DETTRICK]

I am writing to advertise a linux-smp boot-problem workaround.

We have an Asus A7M266-D motherboard with dual Athlon MP 1800+.
We found the linux kernel-smp would seize shortly after or during booting
the second CPU, at around the time it was testing APIC.

We found that booting with the "noapic" option in grub or lilo was
sufficient to solve the problem.
e.g. in /etc/grub.conf the kernel line is:
	kernel /vmlinuz-2.4.17-0.18smp ro root=/dev/hda3 noapic
In lilo.conf I guess one would add the line:
	append="noapic"

That's it.  If you have interest in such things read on.

There was no boot problem with non-SMP linux kernels, but
we tried:
kernel-smp-2.4.9-21.athlon.rpm
kernel-smp-2.4.17-0.18.athlon.rpm
and the default RedHat 7.2 smp kernel.  They all seized at
about the same phase of the boot process, and they would all
boot successfully when the noapic option was included.

AMD and ASUS tech support had never heard of this problem.
ASUS suggested it "might" be the BIOS.

BTW the Athlons, clearly marked in the boxes as MP, identified
themselves as Athlon XP 1800+'s.   We thought this might be the
problem at first but now we guess not.

Note we have a different result to a related earlier posting:
http://www.van-dijk.net/linuxkernel/200204/0054.html

I am not a developer so have not subscribed to the list.
I would appreciate a CC or direct reply (my apologies for this
breach of protocol) if anyone has clarifying comments
e.g. what the heck is APIC anyway, will noapic effect performance,
and is there a more elegant solution.

Regards,

Sean Dettrick,
University of California, Irvine.

Re: A7M266-D, dual athlon 1800+ kernel-smp APIC boot problem workaround

[Alan Cox]

> We have an Asus A7M266-D motherboard with dual Athlon MP 1800+.
> We found the linux kernel-smp would seize shortly after or during booting
> the second CPU, at around the time it was testing APIC.

Yes - the MP 1.4 table seems to have funnies with the second IDE controller

> We found that booting with the "noapic" option in grub or lilo was
> sufficient to solve the problem.

Better yet set MP 1.1 in the BIOS - the MP1.1 table seems to work

> AMD and ASUS tech support had never heard of this problem.
> ASUS suggested it "might" be the BIOS.

AMD have heard of the problem. I've got it filed with them along with
	Won't boot with a broadcom card in
	Seems to misconfigure PCI compliance sometimes

and some other oddments. Support I suspect are not too used to it.

> BTW the Athlons, clearly marked in the boxes as MP, identified
> themselves as Athlon XP 1800+'s.   We thought this might be the
> problem at first but now we guess not.

The BIOS forgets to load the MP name string, like a load of other
problems it has. 

Alan

Re: A7M266-D, dual athlon 1800+ kernel-smp APIC boot problem workaround

[Dave Jones]

On Wed, Feb 27, 2002 at 09:17:15PM +0000, Alan Cox wrote:

 > > BTW the Athlons, clearly marked in the boxes as MP, identified
 > > themselves as Athlon XP 1800+'s.   We thought this might be the
 > > problem at first but now we guess not.
 > 
 > The BIOS forgets to load the MP name string, like a load of other
 > problems it has. 

 x86info has proper descrimination of MP/XP using a readonly
 cpu-capability if theres any doubt here.
 
-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs