Bug 821988

Unfortunately, we can't do much about it because nothing happened from the guest OS' point of view. It has no idea it was suspended and resumed and we have no way of telling this to the guest OS. Therefore I'm closing the bug as CANTFIX.

However, it might be something that qemu guest agent could help with.

This can not be right. If there was "no way of telling" things to the guest OS, then a mere pause would desync the clock as well. However it does not happen. How come the same mechanism that resets the clock after a suspend, can not be employed to reset the clock after a managed save? It seems like a genuine bug to me.

Consider:

rabbit@Dungeon:~$ clockdiff 192.168.58.165
.....
host=192.168.58.165 rtt=237(295)ms/0ms delta=-1ms/-1ms Tue Jun 19 08:16:36 2012

rabbit@Dungeon:~$ si virsh suspend ws_lin_dev
Domain ws_lin_dev suspended

rabbit@Dungeon:~$ clockdiff 192.168.58.165
192.168.58.165 is down

rabbit@Dungeon:~$ si virsh resume ws_lin_dev
Domain ws_lin_dev resumed

rabbit@Dungeon:~$ clockdiff 192.168.58.165
..
host=192.168.58.165 rtt=562(280)ms/0ms delta=-1ms/-1ms Tue Jun 19 08:17:52 2012

rabbit@Dungeon:~$ si virsh managedsave ws_lin_dev
Domain ws_lin_dev state saved by libvirt

rabbit@Dungeon:~$ clockdiff 192.168.58.165
192.168.58.165 is down

rabbit@Dungeon:~$ si virsh start ws_lin_dev
Domain ws_lin_dev started

rabbit@Dungeon:~$ clockdiff 192.168.58.165
..
host=192.168.58.165 rtt=562(280)ms/0ms delta=-57417ms/-57417ms Tue Jun 19 08:19:16 2012
# ^^ OUCH - 57 secs

rabbit@Dungeon:~$ si virsh suspend ws_lin_dev
Domain ws_lin_dev suspended

rabbit@Dungeon:~$ sleep 20
# sleep a while while paused

rabbit@Dungeon:~$ si virsh resume ws_lin_dev
Domain ws_lin_dev resumed

rabbit@Dungeon:~$ clockdiff 192.168.58.165
.
host=192.168.58.165 rtt=750(187)ms/0ms delta=-57417ms/-57417ms Tue Jun 19 08:20:00 2012
# difference exact down to the millisecond after 20 secs of paused sleep

Investigating further I found this warnocked patch, which looks like exactly the thing in question. Please reopen this ticket so that there is a placeholder of when this fix will land properly.

http://www.mail-archive.com/kvm@vger.kernel.org/msg67881.html

(In reply to comment #5)
> Investigating further I found this warnocked patch, which looks like exactly
> the thing in question. Please reopen this ticket so that there is a
> placeholder of when this fix will land properly.
> 
> http://www.mail-archive.com/kvm@vger.kernel.org/msg67881.html

(For the record)
Looks like that patch (by Marcelo Tosatti) appeared in bugzilla #694801 and has been applied to the RHEL 6.3 GA kernel.

This is still a frequently-requested issue on upstream lists; qemu 2.0 made it possible for the guest-agent to do guest-set-time without arguments to have the guest re-read the hardware clock and adjust software time from that, and qemu 2.1 is adding rtc-reset-reinjection to tell qemu when the agent has been used to force guest time and therefore qemu no longer needs to slew clock interrupts.  Libvirt 1.2.5 added a virDomainSetTime API to trigger the guest agent command, and future libvirt versions may add a flag to that API to also trigger the rtc-reset-reinjection followup.  There is also an idea of adding an <on_resume> action to the domain XML of libvirt to allow libvirt to automatically trigger time reset on resume (it can't be on by default, because it requires guest interaction which is only safe if the guest is trusted; but could be explicitly enabled by management as needed).  Of course, as this is still under active work upstream, there is no telling how soon it can be backported downstream, or even if such a backport is feasible without a rebase.