APIC errors on 2.4.4

APIC errors on 2.4.4

[Andy Arvai]

Hi,

I'm having IO-APIC errors with 2.4.4. I spent some time searching the
web to understand more about this problem and I'm still not sure if
it is a hardware problem on the motherboard or a problem with the
kernel. I will try the noapic boot option, but are there any
patches that might fix this? Here are some of the errors I was
getting:

May 15 22:47:43 rad kernel: APIC error on CPU0: 02(01)
May 15 22:48:00 rad kernel: APIC error on CPU0: 01(02)
May 15 22:48:00 rad kernel: APIC error on CPU1: 02(02)
May 15 22:49:11 rad kernel: APIC error on CPU0: 02(02)
May 15 22:49:11 rad kernel: APIC error on CPU1: 02(02)
May 15 22:49:32 rad kernel: APIC error on CPU1: 02(02)
May 15 22:49:32 rad kernel: APIC error on CPU0: 02(02)
May 15 22:50:38 rad kernel: APIC error on CPU1: 02(02)
May 15 22:50:38 rad kernel: APIC error on CPU0: 02(02)
May 15 22:50:46 rad kernel: unexpected IRQ trap at vector d9
May 15 22:51:16 rad kernel: scsi0:0:0:0: Attempting to queue an ABORT message
May 15 22:51:16 rad kernel: scsi0:0:0:0: Command already completed
May 15 22:51:16 rad kernel: aic7xxx_abort returns 8194
May 15 22:51:26 rad kernel: scsi0:0:0:0: Attempting to queue an ABORT message
May 15 22:51:26 rad kernel: scsi0:0:0:0: Command already completed
May 15 22:51:26 rad kernel: aic7xxx_abort returns 8194

Thanks for any ideas.

Andy

Re: APIC errors on 2.4.4

[Hermann Himmelbauer]

Andy Arvai wrote:
> 
> Hi,
> 
> I'm having IO-APIC errors with 2.4.4. I spent some time searching the
> web to understand more about this problem and I'm still not sure if
> it is a hardware problem on the motherboard or a problem with the
> kernel. I will try the noapic boot option, but are there any
> patches that might fix this? Here are some of the errors I was
> getting:
> 
> May 15 22:47:43 rad kernel: APIC error on CPU0: 02(01)
> May 15 22:48:00 rad kernel: APIC error on CPU0: 01(02)

This is a hardware error. I also have a buggy motherboard (586DX) and
experience the same problems. Looke at /proc/interrupts, there you can
see how many errors are detected. The problem is that double errors are
not detected - these can lead to system crashes or block network/isdn
cards.

If there are only a few APIC errors, it is very unlikely that a double
error occurs, if there are very many, the probability is high.

It seems that there are quite a lot of motherboards that have a buggy
APIC.

There is some patch by Alan Cox in the 2.4.4-ac series that does
something about this problem but I do not know what exactly.

The only thing you can do is boot with the "noapic" option, and disable
IO-Interrupts on the second CPU (I assume you have a SMP system?). This
will reduce the amount of errors but there is also a performance
decrease.

		Regards,
		Hermann

-- 
 ,_,
(O,O)     "There is more to life than increasing its speed."
(   )     -- Gandhi
-"-"--------------------------------------------------------------