Kernel panic Virtual Machine ParaVirtualization

SUSE Linux Enterprise Server 10 Service Pack 2

Symptoms:

128 Giga bytes memory Xen Hypervisor
Depending of the setting on Guest VM configuration file, it will produce a kernel panic.

(eg) Xen VM configuration file

name="sles10-sp2"
uuid="85d7e42e-ae5c-6a28-04ec-1cd7628be215"
memory=512
maxmem=1536
vcpus=1
on_poweroff="destroy"
on_reboot="restart"
on_crash="destroy"
localtime=0
builder="linux"
bootloader="/usr/lib/xen/boot/domUloader.py"
bootargs="--entry=xvda2:/boot/vmlinuz-xen,/boot/initrd-xen"
extra=""
disk=[ 'file:/home/geeko/xen/sles10-sp2/disk0,xvda,w', ]
vif=[ 'mac=00:16:3e:65:f4:0d', ]
vfb=['type=vnc,vncunused=1']
 

Setting that provoke a kernel panic: - memory=2560 maxmem=130371.

Memory = 2560 means that the virtual machine will start with 2.5 giga bytes of memory .

Maxmem=130371 correspond to the maximum memory that can be hot-plug allocated to  a  virtual machine. With this setting about 127 giga bytes of memory.

It is working if we set maxmem to a lower value or if we increase memory (in the previous example, it works with memory=5120).

Kernel panic :
...
Setting mem allocation to 2621440 kiB
IA32 emulation $Id: sys_ia32.c,v 1.32 2002/03/24 13:02:28 ak Exp $
audit: initializing netlink socket (disabled)
audit(1251968031.958:1): initialized
divide error: 0000 [1] SMP
last sysfs file:
CPU 0
Modules linked in:
Pid: 6, comm: kthread Not tainted 2.6.16.60-0.37_f594963d-xen #1
RIP: e030:[<ffffffff80163060>]<ffffffff80163060>{shrink_zone+471}RSP:
e02b:ffff8800010e5980 EFLAGS: 00010246

We recommended to not use the entire hypervisor  memory as setting for max-mem because the guest has to allocate management structures for all this not-present memory.

A reasonable range should be of 10-20 times the initial memory amount.

(e.g.) max-mem<=  20*mem guest

Latest SLES 10 SP3 kernel since version 2.6.16.60-0.59.1 do not have the issue. Please apply latest SLES10 kernel.