KVM/Livirt failing to start vm's after upgrade to linux-image-2.6.32-22-generic

[duncs]

My VM's were working fine yesterday on my up-to-date Lucid system.

Noticed a new kernel release this morning, did an upgrade and now my VM's wont boot.

This is as much as the /var/log/libvirt/qemu/dev10.log log file tells me:

LC_ALL=C PATH=/usr/local/sbin:/usr/local/bin:/usr/bin:/usr/sbin:/sbin:/bin QEMU_AUDIO_DRV=none /usr/bin/kvm -S -M pc-0.12 -cpu qemu32 -enable-kvm -m 812 -smp 1 -name dev10 -uuid 47615a48-2f06-c731-b20d-8db124f79f7e -chardev socket,id=monitor,path=/var/lib/libvirt/qemu/dev10.monitor,server,nowait -monitor chardev:monitor -boot c -drive file=/srv/vm_disks/dev10/disk0.qcow2,if=ide,index=0,boot=on -drive if=ide,media=cdrom,index=2 -net nic,macaddr=52:54:00:24:c8:5a,vlan=0,name=nic.0 -net tap,fd=34,vlan=0,name=tap.0 -chardev pty,id=serial0 -serial chardev:serial0 -parallel none -usb -vnc 127.0.0.1:0 -k en-gb -vga cirrus 
char device redirected to /dev/pts/2
pci_add_option_rom: failed to find romfile "pxe-rtl8139.bin"

The console for the VM is blank and cpu use in virt-manager is constant. If i boot back into the previous kernel I have available the vm's can start again (however, due to nvidia driver silliness I cannot use that kernel easily as my dual screen don't work).

Anyone else upgraded and found a problem with their VM's?

Duncs

[Kokopelli]

This is happening to me as well. I have not tried to debug it much, came here to see if anyone else has the same problem. A quick search implies that is is a problem with libvirt.

[mgu]

All the same here

The message log you put does not seems to be relevant since it is also displayed when VMs works fine (with 2.6.32-21)

char device redirected to /dev/pts/2
pci_add_option_rom: failed to find romfile "pxe-rtl8139.bin"

I boot back with the previous kernel too.
Best regards

[Kokopelli]

The pci_add_option_rom: failed to find romfile "pxe-rtl8139.bin" is a red herring. Adding the kvm-pxe package gets rid of that message but the VMs still do not start. I am not sure what is causing it yet. Has anyone tried launching directly from the command line using "kvm <options>"?

EDIT tried a simple "kvm -m 1024 -drive file=/dev/eli/test1" and it never even got to the launching of the BIOS.

[cajsaj]

Same problem here too. Switched to previous kernel and everything looks ok.

Log file shows:
Could not open option rom 'pxe-rtl8139.bin': No such file or directory

I tried the other network devices, and they each gave a similar message. It looks like the rom files have been removed..? May be related to this:
https://bugs.launchpad.net/ubuntu/+s...ot/+bug/472485

[duncs]

The pci_add_option_rom: failed to find romfile "pxe-rtl8139.bin" is a red herring. Adding the kvm-pxe package gets rid of that message but the VMs still do not start.

I have that installed and I still get the error, but yes it is probably a red herring.

I am not sure what is causing it yet. Has anyone tried launching directly from the command line using "kvm <options>"?

EDIT tried a simple "kvm -m 1024 -drive file=/dev/eli/test1" and it never even got to the launching of the BIOS.

Yes, I tried the command line running too and make no difference for me.

Duncs

[duncs]

All the same here

The message log you put does not seems to be relevant since it is also displayed when VMs works fine (with 2.6.32-21)

Trouble is I haven't yet found any relevant logging about the problem.

A quick strace shows

[pid  3641] rt_sigprocmask(SIG_BLOCK, [BUS ALRM IO], NULL, 8) = 0
[pid  3641] signalfd(4294967295, [BUS ALRM IO], 8) = 16
[pid  3641] fcntl(16, F_GETFD)          = 0
[pid  3641] fcntl(16, F_SETFD, FD_CLOEXEC) = 0
[pid  3641] fcntl(16, F_SETFL, O_RDONLY|O_NONBLOCK) = 0
[pid  3641] write(15, "\1\0\0\0\0\0\0\0", 8) = 8
[pid  3641] futex(0x869ae4, FUTEX_CMP_REQUEUE_PRIVATE, 1, 2147483647, 0x869a60, 2) = 1
[pid  3641] futex(0x869a60, FUTEX_WAKE_PRIVATE, 1 <unfinished ...>
[pid  3642] <... futex resumed> )       = 0
[pid  3641] <... futex resumed> )       = 0
[pid  3642] futex(0x869a60, FUTEX_WAKE_PRIVATE, 1 <unfinished ...>
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {0, 0} <unfinished ...>
[pid  3642] <... futex resumed> )       = 0
[pid  3641] <... select resumed> )      = -1 EBADF (Bad file descriptor)
[pid  3642] rt_sigtimedwait([BUS RT_6],  <unfinished ...>
[pid  3641] timer_gettime(0, {it_interval={0, 0}, it_value={0, 0}}) = 0
[pid  3641] timer_settime(0, 0, {it_interval={0, 0}, it_value={0, 250000}}, NULL) = 0
[pid  3641] write(15, "\1\0\0\0\0\0\0\0", 8) = 8
[pid  3641] timer_gettime(0, {it_interval={0, 0}, it_value={0, 182897}}) = 0
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {1, 0}) = -1 EBADF (Bad file descriptor)
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {1, 0}) = -1 EBADF (Bad file descriptor)
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {1, 0}) = -1 EBADF (Bad file descriptor)
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {1, 0}) = -1 EBADF (Bad file descriptor)
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {1, 0}) = -1 EBADF (Bad file descriptor)
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {1, 0}) = -1 EBADF (Bad file descriptor)
[pid  3641] select(35, [4 5 8 11 13 14 16 34], [], [], {1, 0}) = -1 EBADF (Bad file descriptor)
.... last line repeats continually ....

More digging required.

Duncs

[bbrand]

Sigh, same here, been trying to get something working all day. No luck so far. Did anyone log this bug yet?

[Frogs Hair]

I have found my computer is flaky when it comes to kernel updates . what works every time for me is removing the proprietary Nvidia driver prior to kernel installation and then booting with the new kernel and reinstalling the driver. This requires changing the update settings to notify only.

[Kokopelli]

It has not been logged as a bug yet as far as I can see. Since this bug directly effects kvm I am guessing it effects cloud install as well. I expect that it will get noticed fairly quickly if that is the case.