Soft lockup detected on CPU0# - 2.6.16.18

[goo]

Hey guys,

Got a system currently running 2.6.8-2-686-smp with VMware server. Everything is running perfectly, except the ram is capped at 4GB (system physically has 6GB)

# free -m
total used free shared buffers cached
Mem: 3978 124 3853 0 33 32
-/+ buffers/cache: 59 3919
Swap: 2255 0 2255
#

I downloaded, configured and compiled 2.6.16.18

using configuration:
http://goo.sgl.org.au/config-2.6.8-2-686-smp.txt

and it halts booting and says

soft lockup detected on CPU#0!
Pid: 1, comm: init

The rest i will have to write down if it's essential.

I can still send keys like reboot and it will switch RL ok, but it just doesn't proceed past this error

Configuration:

Debian 3.1r2
adaptec aic7899 controller - 3x SCSI disks (using LVM)
Dual Xeon system
6GB ECC

relevent entry in /etc/lilo.conf :
iimage=/vmlinuz
label=LinuxHM
read-only
# restricted
# alias=1
append="noapic"
initrd=/initrd.img

I used these commands to compile the kernel :

make menuconfig
make-kpkg clean
fakeroot make-kpkg --revision=custom.1.0 kernel_image
cd ..
dpgk^C
dpkg -i kernel-image-2.6.16.18_custom.1.0_i386.deb
(it gave me an error out about missing the initrd file)
cd /boot/
mkinitrd -o /boot/initrd.img-2.6.16.18 2.6.16.18
cd /
mv initrd.img initrd.img.old
ln -s /boot/initrd.img-2.6.16.18 initrd.img
lilo -v

[Roxoff]

Hmmm, I'm no expert at either Debian or Xeon systems, but I have an observation/question. Why do you have 'append=noapic' in the lilo config? My understanding is that the APIC marshalls interrupts between cpus on a multi-cpu system, it's better than having both processors try to respond to the same interrupt. If it is turned it off (as it is in your system), you might find that one of the processors could lock up waiting to process an interrupt that has already been handled by another cpu...