KVM crashes with "exception 13 (33)" after an interrupted installation

[Henning Sprang]

Binary package hint: kvm

Steps to reproduce the bug:

* create an empty disk image with qemu-img myimage.img 5G
* start debian lenny netinstall d-i and run installation with "kvm -hda myimage.img -cdrom lenny-installer.iso
* in the middle of the install(while package installation, partitioning is succeeded), close the kvm window, that way killing the whole process.
* try to start the installer again, with the same command as above.

Result:

in the consoile window I get:

henning@xnote:/hier/produktion/run/vm-images$ kvm -hda
libvirt-test_broken-image.img -cdrom
/hier/produktion/tmp/test-isos/debian-LennyBeta2-i386-netinst.iso
exception 13 (33)
rax 00000000000082ff rbx 0000000000007e06 rcx 0000000000007ed4 rdx
00000000000005fd
rsi 00000000ffff093e rdi 00000000000807ff rsp 000000000000e149 rbp
000000000000b14e
r8 0000000000000000 r9 0000000000000000 r10 0000000000000000 r11
0000000000000000
r12 0000000000000000 r13 0000000000000000 r14 0000000000000000 r15
0000000000000000
rip 0000000000000003 rflags 00033602
cs 0000 (00000000/0000ffff p 1 dpl 3 db 0 s 1 type 3 l 0 g 0 avl 0)
ds 0000 (00000000/0000ffff p 1 dpl 3 db 0 s 1 type 3 l 0 g 0 avl 0)
es 0800 (00008000/0000ffff p 1 dpl 3 db 0 s 1 type 3 l 0 g 0 avl 0)
ss 1000 (00010000/0000ffff p 1 dpl 3 db 0 s 1 type 3 l 0 g 0 avl 0)
fs 0000 (00000000/0000ffff p 1 dpl 3 db 0 s 1 type 3 l 0 g 0 avl 0)
gs 0000 (00000000/0000ffff p 1 dpl 3 db 0 s 1 type 3 l 0 g 0 avl 0)
tr 0000 (fffbd000/00002088 p 1 dpl 0 db 0 s 0 type b l 0 g 0 avl 0)
ldt 0000 (00000000/0000ffff p 1 dpl 0 db 0 s 0 type 2 l 0 g 0 avl 0)
gdt fb1f2/30
idt 0/3ff
cr0 10 cr2 0 cr3 0 cr4 0 cr8 0 efer 0
code: 53 ff 00 --> f0 53 ff 00 f0 53 ff 00 f0 53 ff 00 f0 53 ff 00 f0 53 ff 00
f0 53 ff 00 f0 53 ff 00 f0 a5 fe 00 f0 87 e9 00 f0 53 ff 00 f0 53 ff 00 f0 53
ff
Aborted

....

dmesg says:

[ 5069.009173] kvm: guest NX capability removed
[ 5069.737849] handle_exception: unexpected, vectoring info 0x80000306 intr
info 0x80000b0d
[ 5069.738077] pending exception: not handled yet
[ 5069.738645] pending exception: not handled yet

Solution:

Starting kvm with the additional argument "-boot d" runs the installation
again.

So, it's obviously some strange effect with kvm and a disk that does have
partitions but no proper bootloader.

Still, and unexperienced user (I have more than 5 years of qemu and
virtualization experience), would not easily find a way out of this
situation, with such an error message.

-- Package-specific info:

selected information from lshal(1):

/proc/cpuinfo:

processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 14
model name : Intel(R) Core(TM) Duo CPU L2400 @ 1.66GHz
stepping : 12
cpu MHz : 1667.000
cache size : 2048 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 2
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx constant_tsc arch_perfmon bts pni monitor vmx est tm2 xtpr
bogomips : 3327.94
clflush size : 64

processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 14
model name : Intel(R) Core(TM) Duo CPU L2400 @ 1.66GHz
stepping : 12
cpu MHz : 1667.000
cache size : 2048 KB
physical id : 0
siblings : 2
core id : 1
cpu cores : 2
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 10
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe nx constant_tsc arch_perfmon bts pni monitor vmx est tm2 xtpr
bogomips : 3324.41
clflush size : 64

-- System Information:
Debian Release: lenny/sid
  APT prefers hardy-security
  APT policy: (500, 'hardy-security'), (500, 'hardy')
Architecture: i386 (i686)

Kernel: Linux 2.6.24-19-rt (SMP w/2 CPU cores; PREEMPT)
Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968)
Shell: /bin/sh linked to /bin/dash

Versions of packages kvm depends on:
ii adduser 3.105ubuntu1 add and remove users and groups
ii bridge-utils 1.2-2 Utilities for configuring the Linu
ii iproute 20071016-2ubuntu1 Professional tools to control the
ii libasound2 1.0.15-3ubuntu4 ALSA library
ii libc6 2.7-10ubuntu3 GNU C Library: Shared libraries
ii libgnutls13 2.0.4-1ubuntu2.1 the GNU TLS library - runtime libr
ii libsdl1.2debian 1.2.13-1ubuntu1 Simple DirectMedia Layer
ii vgabios 0.6a-3ubuntu1 VGA BIOS software for the Bochs an
ii zlib1g 1:1.2.3.3.dfsg-7ubuntu1 compression library - runtime

Versions of packages kvm recommends:
ii qemu 0.9.1-1ubuntu1 fast processor emulator
pn vde2 <none> (no description available)

-- no debconf information

[Kai Blin (kai.blin)]

Also present in Intrepid beta2, if you create the images with virt-install. I wonder where the qemu packages went on 8.10

However, you don't even have to go through the fuss of trying to install debian and aborting. For me, trying to boot from an opensuse iso was sufficient to break the "initial boot". Once your install cd vanished in virt-manager (see bug #238692), you can't start the system via the command line either.

[Dustin Kirkland  (kirkland)]

Hi, thanks for the report.

I need to mark 'Invalid', as this isn't actually a bug.

Basically, once the install has started, data has been written to the hard disk image, but if the install doesn't complete, there's no boot loader installed in the boot sector yet.

KVM is then going to try and execute arbitrary code in that disk's boot sector, and poof, it's not going to boot ;-)

This is an issue with Intel processors, that don't handle real mode properly.

The "fix" is to recreate your vm image and start your install over.

:-Dustin

[Henning Sprang (henning)]

Hi,

On Fri, Mar 6, 2009 at 5:24 PM, Dustin Kirkland
<email address hidden> wrote:
> KVM is then going to try and execute arbitrary code in that disk's boot
> sector, and poof, it's not going to boot ;-)

Hmm.
But real hardware and some other virtualization solutions say
something like "not bootloader found" or something like that.
Why shouldn't kvm be able to say that?

> This is an issue with Intel processors, that don't handle real mode
> properly.
>
> The "fix" is to recreate your vm image and start your install over.

Yes, I know it's not too hard to fix. I already wrote a workaround in
the bugreport.

Anyway, if you think it cannot be fixed, do what you think fits.

Henning