NForce2 pseudoscience stability testing (2.6.0-test11)

From: ross . alexander
Date: Fri Nov 28 2003 - 10:15:08 EST


Brendan et al,

I have been test various kernel parameter combinations to test stability.

Basic scenario is default IDE driver, AIC7xxx PCI SCSI connected to HP
externel DVD.
Using grip (CDDA ripper) to test if system locks up. Unstable systems
will normally lock up
around track two or three. Stable systems are those which haven't locked
up after half
a dozen different CDs have been ripped.

MB: ASUS A7N8X
CPU: Athlon XP 2700+
Memory: 1.5GB (3 x 512MB DIMMs)
Disk: Internal 80GB IDE

COMPILE SMP ACPI PCI LAPIC APIC RESULT NOTE
SMP,PREM ON F
SMP,PREM ON NOACPI F
SMP,PREM OFF F
SMP,PREM OFF NO NO S
SMP,PREM OFF NO F
SMP,PREM OFF NO S
SMP,PREM ON NO NO F 1
SMP ON NO F 2
SMP F 3
APIC,LAPIC S
PREM,APIC,LAPIC S

* SMP = On (if compiled it) unless nosmp set. Using nosmp with smp kernel
causes very odd results.
* ACPI = Compiled by default. Set off using kernel paramter acpi=off.
* PCI = Kernel parameter. Only used to turn APCI routing off.
* LAPIC = Kernel paramter to turn it off (nolapic).
* APIC = Kernel paramter to turn it off (noapic).

1. Using APCI PCI routing and nolapic gives very odd results. As soon as
the network tries
to configure itself it simple hangs (but C-Alt-Del will reboot it).

2. Using the nolapic kernel parameter without disabling ACPI does nothing.

3. Using kernel parameter nosmp on and SMP kernel causes all the lower 16
IRQs to work in XT-PIC
mode and the PCI network card to use IRQ 21 with IO-APIC-level. Trying to
modprobe aic7xxx hung
modprobe but not system.

The conclusion to this is the problem is in Local APIC with SMP. I'm not
saying this is actually true
only that is what the data suggests. If anybody wants me to try some
other stuff feel free to suggest
ideas.

Cheers,

Ross

---------------------------------------------------------------------------------
Ross Alexander "We demand clearly defined
MIS - NEC Europe Limited boundaries of uncertainty and
Work ph: +44 20 8752 3394 doubt."
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/