APIC problems in linux

[Crusty]

APIC error on CPU0: 40(40)


I'm using the 2.6.8 smp kernel from the debian packages without any modifications to it.

This is on a dual p3 733 and ASUS cuv4x-d

What does that error mean? Any other info you might need to help?

 

[Nothinman]

AFAIK they're nothing, I get them once in a while on my dual Athlon box and they have never caused any problems.

From arch/i386/kernel/apic.c :

1189 /* Here is what the APIC error bits mean:
1190 0: Send CS error
1191 1: Receive CS error
1192 2: Send accept error
1193 3: Receive accept error
1194 4: Reserved
1195 5: Send illegal vector
1196 6: Received illegal vector
1197 7: Illegal register address
1198 */
1199 printk (KERN_DEBUG "APIC error on CPU%d: %02lx(%02lx)\n",
1200 smp_processor_id(), v , v1);

 

[Crusty]

Ah, well im trying to track my system lockups and as of now that was the only problem i could find in my log files.

It's very weird, it only locks up when i'm not using it...i thought it was a problem with xscreensaver but i disabled it and still the same problem. The system completely locks up, I can't even ssh into it from the network

 

[Nothinman]

It's a bit of a pain, but have you tried setting up a serial console so you can see if there's any oops output?

 

[Crusty]

No I haven't, and I don't know how to.

Can you point me in the direction for learning how to?

 

[Nothinman]

Documentation/serial-console.txt

The general idea is to get a serial cable to run between two machines, on the host you're trying to debug you boot with 'console=ttyS1,9600 console=tty0' and run something like minicom (or Hyperterminal if it's a Windows box) on the other machine, now when you boot the Linux box you should see all the normal boot messages from the kernel on both boxes, this way if an oops happens you see it on the other box so you can copy/paste it. If you want you can setup a getty on the Linux box to start on the serial console so that you can login via there too.

 

[drag]

when I think of "machine locking up while I am away" I am thinking of power saving managment issues.

Try disabling all power managment stuff in your BIOS. Try passing (I am thinking that these are the correct parameters) acpi=off and apm=off to your kernel at bootup time. Also check to make sure that you have only one of apcid or apmd running at any time.

See if that helps it go away. maybe the BIOS is trying to shut off your vid card or cause your harddrive to spin down and the hardware doesn't like it.


For the serial cable stuff you need to make sure that you use the right type of cable, too. They are called 'null modem cables' or 'laplink' cables (after a namebrand). They are used, i beleive, in external modems, so if you have one of those you could probably use it's cable.

If you use the wrong type of cable you can blow out some circuits on your motherboard. (i had a old boss blow out a 5000 dollar RIP thingy for a very expensive printer once doing that, warned her but she didn't listen to me. Live and learn)

Very usefull for servers. Say you have a dozen servers on a rack, you could setup one monitoring station with a multi-port serial card hooked up to all of them. Lilo, and maybe Grub, supports redirecting your main console output directly to the serial port. Then you have a bunch of minicom sessions going on and you can see the output for all those machines. Very usefull. Especially for security. You could have all root access denied to the servers except thru those serial ports for instance.

I like to use a old laptop with battery issues as a monitor for servers.. you have 2 serial ports usually. Lot nicer then having a bulky old monitor and keyboard with a few broken buttons that nobody wants attached to each and every server.