Reboot fails with virtio disks

Bug #292587 reported by Dean Holland
22
This bug affects 3 people
Affects Status Importance Assigned to Milestone
kvm (Debian)
Fix Released
Unknown
kvm (Ubuntu)
Fix Released
Undecided
Unassigned
Intrepid
Won't Fix
Undecided
Unassigned

Bug Description

Binary package hint: kvm

When rebooting a guest with virtio disks either by sending it Ctrl-Alt-Del in virt-manager or issuing a reboot from within the guest itself, the guest fails to find the disk on next boot. The error in the guest's console is:

Booting from Hard Disk... Boot from Hard Disk failed: could not read boot disk ...
FATAL: No bootable device

Destroying the guest and starting it again restores the guest to operation. If the type of disk is changed from "virtio" to "ide" then the reboot works without issue.

This thread from the linux-kvm list seems to explain the problem also: http://kerneltrap.org/mailarchive/linux-kvm/2008/8/29/3131254/thread

speedster@whimper:~$ apt-cache policy kvm
kvm:
  Installed: 1:72+dfsg-1ubuntu6
  Candidate: 1:72+dfsg-1ubuntu6
  Version table:
 *** 1:72+dfsg-1ubuntu6 0
        500 http://mirror.3fl.net.au intrepid/main Packages
        100 /var/lib/dpkg/status

Steps to reproduce:
1. Boot a guest that has been configured with virtio model disks
2. Reboot the guest
3. Change to ide model disks and boot/reboot. This should work fine.

Revision history for this message
Dustin Kirkland  (kirkland) wrote :

Hello-

This bug has been fixed in upstream KVM. See git commit 77c125369426a519fb9ea92dc159fa5ce392f354.

I've confirmed on a Jaunty host that I can boot a VM with:
 $ kvm -drive file=intrepid-server.img,if=virtio,boot=on

And reboot it without a problem.

Thanks for the report!
:-Dustin

Changed in kvm:
status: New → Fix Released
Revision history for this message
Philipp Kern (pkern) wrote :

This should be fixed also in Intrepid. 1:72+dfsg-1ubuntu7~ppa0 out of my PPA fixes the issue by taking the patch from Debian Lenny. I only found this bug report later than I did the upload, so the changelog does not reference this bug. I would be happy if someone could sponsor this to -proposed.

Revision history for this message
Philipp Kern (pkern) wrote :
Revision history for this message
Martin Pitt (pitti) wrote :

Is this a regression from hardy? If not, this doesn't look like meeting the SRU criteria. Also, what's the potential regression impact from this ROM reset? The patch isn't very focused on this particular bug, it rather seems to change behaviour on a general level.

Revision history for this message
Bryan McLellan (btm) wrote :

The patch cited wasn't upstream until kvm-75. Hardy had kvm-62, which should [1] have virtio support. I didn't switch to virtio until intrepid myself, I can't comment on if it's reproducible there. As important as being able to soft reboot your guests is, you can workaround this with a forced power off. Meeting the SRU for "obviously safe" sounds simple with how minimal the patch is, but I'd side that the workarounds are safer than risking a kvm patch, with how important KVM is.

Jaunty ships shortly, and this bug is fixed there.

[1] http://www.linux-kvm.org/page/Virtio

Revision history for this message
Martin Pitt (pitti) wrote :

Intrepid is not popular enough any more to ship non-critical bug fixes.

Changed in kvm (Ubuntu Intrepid):
status: New → Won't Fix
Changed in kvm (Debian):
status: Unknown → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.