======================================
impact: qemu-kvm cannot be installed if modprobe of kvm will fail
test case: create a precise container; rm /bin/running-in-container; ln -s /bin/false /bin/running-in-container; apt-get install qemu-kvm
regression potential: if the modprobe failures were due to real errors, kvm may fail later on
======================================
Iso 172, 180, 183
Ubuntu (guest) 12.04.04 qemu-kvm fails to install because upstart wont pass.
root@node-2:~# apt-get install -f
Reading package lists... Done
Building dependency tree
Reading state information... Done
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
3 not fully installed or removed.
After this operation, 0 B of additional disk space will be used.
Setting up qemu-kvm (1.2.0+noroms-0ubuntu7.12.10) ...
start: Job failed to start
invoke-rc.d: initscript qemu-kvm, action "start" failed.
dpkg: error processing qemu-kvm (--configure):
subprocess installed post-installation script returned error exit status 1
expected result: qemu-kvm should install and "start" anyways even if acceleration can't be loaded.
grep "^flags.* svm" /proc/cpuinfo
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext fxsr_opt pdpe1gb lm 3dnowext 3dnow rep_good nopl extd_apicid pni cx16 popcnt hypervisor cmp_legacy svm cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw
root@node-1:/etc/init# modprobe kvm
root@node-1:/etc/init# modprobe kvm_amd
FATAL: Error inserting kvm_amd (/lib/modules/3.11.0-18-generic/kernel/arch/x86/kvm/kvm-amd.ko): Operation not supported
root@node-1:/etc/init# ls -l /dev/kvm
ls: cannot access /dev/kvm: No such file or directory
dmesg:
[10925.939670] has_svm: can't execute cpuid_8000000a
[10925.939677] kvm: no hardware support
init fails due to
cat /etc/init/qemu-kvm.conf
...
pre-start script
# Silently exit if the package isn't installed anymore
if [ ! -e /usr/bin/kvm ]; then
exit 0
fi
[ -r /etc/default/qemu-kvm ] && . /etc/default/qemu-kvm
# Load the appropriate module, respecting blacklists
if grep -qs "^flags.* vmx" /proc/cpuinfo; then
modprobe -b kvm_intel "$KVM_NESTED" || true
chown root:kvm /dev/kvm
chmod 0660 /dev/kvm
elif grep -qs "^flags.* svm" /proc/cpuinfo; then
modprobe -b kvm_amd || true
chown root:kvm /dev/kvm
chmod 0660 /dev/kvm
fi
...
It appears that guest might have been passed bad CPU flags and qemu-kvm in host might not be allowing kvm acceleration to load. It appears that similar results have been noted in Debian upstream, where they are testing for -e /dev/kvm before attempting to load more acceleration
in Debian SID
case "$1" in
start|restart|force-reload)
[ ! -e /dev/kvm ] || exit 0 # nothing to do
Returning 0 if /dev/kvm does not exist after the modprobe would be ok, so long as those users do nto want to have hugetlbfs mounted. Otherwise we can simply guard the chmod with || true as we do the modprobe.