KVM does not boot Ubuntu i386 10.04 minimal server images

Thanks for taking the time to report this bug. Have you tried running one
of the misbehaving images on maverick on x86-64? If so, did that succeed
or fail?

I have no x86-64 setup on any pc and I am out of time to experiment right now. I should notice here that apart from the old one misbehaving image, to reproduce it is as easy as making a fresh 10.04.1 i386 MINIMAL installation through libvirt or by directly using kvm with command line. After the installation has finished succesfully, it will never boot.

Quoting SqUe (<email address hidden>):
> I have no x86-64 setup on any pc and I am out of time to experiment
> right now. I should notice here that apart from the old one misbehaving
> image, to reproduce it is as easy as making a fresh 10.04.1 i386 MINIMAL
> installation through libvirt or by directly using kvm with command line.
> After the installation has finished succesfully, it will never boot.

Yup, again thanks for your time in reporting it. Unfortunately I don't
have a kvm-capable 32-bit machine to test on, but I'll look for one and
attempt to reproduce.

This seems much like my problem. If I boot the previous kernel 2.6.32-24 from everything works OK. This makes me suspect that the problems is within the kernel, did Maverick drop i386 or i686 support or something?

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/663191

I had a boot problem with maveric, too. My solution was to add the apparmor
sevurity rules in virt-manager. Now it works again.

Am 28.10.2010 11:41 schrieb "Harald Hannelius" <email address hidden>:

This seems much like my problem. If I boot the previous kernel 2.6.32-24
from everything works OK. This makes me suspect that the problems is
within the kernel, did Maverick drop i386 or i686 support or something?

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/663191

--
KVM does not boot Ubuntu i386 10.04 minimal sever images
https://bugs.launchpad.net/bugs/666180
You received this bug notification because you are subscribed to qemu-
kvm in ubuntu.

Apparmor doesn't seem to be the culprit for me. virt-manager show me the dynamic allocation of an UUID when starting the guest and I have no errors in syslog regarding apparmor.

The guest just stops just after loading the first stage of the bootloader. This affects my ubuntu 10.10 guest, as well as my Windows Xp SP3 guest. The XP-guest does boot into safe mode, and safe mode with networking though.

Also, changing the disk type from IDE to RH's Virtio SCSI in WInXP didn't change anything either. The guest just sits there chewing up CPU, and the blue-bar-animation runs from side to side.

I have exactly the same symptoms as in comment #7. Also, using `qemu` `qemu -enable-kvm` instead of `kvm` allows the guest OS to fully boot.

This bug seems to be triggered by the fact that /usr/bin/kvm is a symbolic link that points to qemu-system-x86_64. The shell (i.e. libvirt) that is invoking the virtual machine should add a "-cpu qemu32" to the command line if the guest OS is 32-bit. This seems to be a change between the Lucid and Maverick qemu-kvm packages, however I am no longer sure this is really a bug with qemu-kvm.

@Philippe Gauthier

I tried getting rid of libvirt and invoking qemu directly but that did not help.

All of these fail the same way:

qemu -M pc -hda /img/lenny-vol.img -m 1024 -cdrom /iso/debian-507-i386-netinst.iso -net nic,vlan=0 -net user,vlan=0,hostname=emu -monitor pty -boot d

qemu -M pc -hda /img/lenny-vol.img -m 1024 -cdrom /iso/debian-507-i386-netinst.iso -cpu qemu32 -net nic,vlan=0 -net user,vlan=0,hostname=emu -monitor pty -boot d

qemu -M pc -hda /img/lenny-vol.img -m 1024 -cdrom /iso/debian-507-i386-netinst.iso -enable-kvm -net nic,vlan=0 -net user,vlan=0,hostname=emu -monitor pty -boot d

qemu -M pc -hda /img/lenny-vol.img -m 1024 -cdrom /iso/debian-507-i386-netinst.iso -enable-kvm -cpu qemu32 -net nic,vlan=0 -net user,vlan=0,hostname=emu -monitor pty -boot d

Attempting to boot Debian 5 32-bit guest gets as far as
"Decompressing Linux... Parsing ELF... done."
"Booting the kernel"
Then nothing. The kvm process spins the CPU 100% on the host.

Details:

$ uname -a
Linux lakshmi 2.6.35-24-generic-pae #42-Ubuntu SMP Thu Dec 2 03:21:31 UTC 2010 i686 GNU/Linux
$ lsb_release -a
LSB Version: core-2.0-ia32:core-2.0-noarch:core-3.0-ia32:core-3.0-noarch:core-3.1-ia32:core-3.1-noarch:core-3.2-ia32:core-3.2-noarch:core-4.0-ia32:core-4.0-noarch
Distributor ID: Ubuntu
Description: Ubuntu 10.10
Release: 10.10
Codename: maverick
$ dpkg -S /usr/bin/kvm
qemu-kvm: /usr/bin/kvm
$ dpkg -l qemu-kvm | cat
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name Version Description
+++-=====================================-=================================================-==========================================================================
ii qemu-kvm 0.12.5+noroms-0ubuntu7.1 Full virtualization on i386 and amd64 hardware
$ ls -l /usr/bin/kvm
lrwxrwxrwx 1 root root 18 2011-01-09 09:54 /usr/bin/kvm -> qemu-system-x86_64
$

@nutznboltz : I tried and failed to reproduce the bug you are describing. Also this bug report is about when an system update causes virtualization to fail. If this does not apply to your situation, I encourage you to open a new bug report. because there is a high probability that the cause of the problem is different. Thanks!

@Philippe Gauthier : Is it for sure a bug after upgrade? As there isnt anyone here saying that creating a fresh (host) installation solves the problem.

You are right. I was implying that libvirt in 10.10 would create proper configuration files and not have this bug, but I didn't verify that. The only thing that I could verify is that some images (32-bit guests) that used to work with /usr/bin/kvm in 10.04 stopped working in 10.10. In my case, replacing /usr/bin/kvm (which points to /usr/bin/qemu-system-x86_64) by /usr/bin/qemu fixed the problem.

@Phiippe - I had the same problem, and your workaround has me working again. Cheers!

10.10, Maverick
After upgrading my system (MB replacement, added memory up to 4Gb), and after switching to PAE kernel all my virtual machines has stopped booting linux kernel (only grub). Symthoms are similar - CPU usage on host=50% (1 of 2 cores I suppose)

@Aaz, if you run the non-PAE 32-bit kernel does the issue go away? I know that you won't be able to use all 4GB RAM without PAE, I'm just interested in testing PAE vs. non-PAE.

I wonder if upstream kernel commit id f87f928882d080eaec8b0d76aecff003d664697d (from https://bugzilla.kernel.org/show_bug.cgi?id=25262) would fix this for you.

I can't test; my only AMD box is now converted to 64-bit in reaction to this issue.