Bug#777735: systemd: Disk corruption on reboot initiated through systemd

[Peter Colberg]

On Thu, Feb 12, 2015 at 07:14:05AM +0100, Michael Biebl wrote:
> Thanks, I was able to reproduce the issue this way (marking the bug
> accordingly).
> It's likely that the watchog is resetting systemd/rsyslog before it has
> a chance to stop properly.
> 
> Let's see, where we can go from here.

In a vimdiff comparing the two console logs I sent earlier, the last
messages before the reboot of the VM with watchdog are

  Failed to send unit remove signal for final.target: Transport endpoint is not connected
  Hardware watchdog 'i6300ESB timer', version 0

while the VM without watchdog does unmounts after final.target

  Failed to send unit remove signal for final.target: Transport endpoint is not connected
  Sending SIGTERM to remaining processes...
  [   62.134791] systemd-journald[146]: Received SIGTERM from PID 1 (systemd-shutdow).
  Sending SIGKILL to remaining processes...
  Unmounting file systems.
  Unmounting /dev/mqueue.
  Unmounting /sys/kernel/debug.
  Unmounting /dev/hugepages.
  [   62.314461] EXT4-fs (vda3): re-mounted. Opts: (null)
  [   62.316377] EXT4-fs (vda3): re-mounted. Opts: (null)
  [   62.316999] EXT4-fs (vda3): re-mounted. Opts: (null)
  All filesystems unmounted.
  Deactivating swaps.
  All swaps deactivated.
  Detaching loop devices.
  All loop devices detached.
  Detaching DM devices.
  All DM devices detached.
  /lib/systemd/system-shutdown succeeded.
  [   62.382524] Unregister pv shared memory for cpu 2
  Rebooting.
  [   62.382524] Unregister pv shared memory for cpu 0
  [   62.382524] Unregister pv shared memory for cpu 1
  [   62.385199] Unregister pv shared memory for cpu 3
  [   62.386474] reboot: Restarting system
  [   62.387022] reboot: machine restart


It is curious that the bug only manifests itself with the qemu watchdog.
Two physical machines using the hardware iTCO_wdt watchdog reboot fine.