[libvirt] [PATCH v3 0/2] API to invoke S3/S4 on a host and also resume from within libvirt

[Alon Levy]

On Mon, Nov 21, 2011 at 05:26:55PM -0700, Eric Blake wrote:
> On 11/09/2011 05:05 AM, Srivatsa S. Bhat wrote:
> > (This patch is positioned to go in after the patch that exports the host
> > power management capabilities as XML, posted in [4])
> 
> I'm now reviewing that patch along with this series; if we need another
> revision, it may be easiest to merge the three patches into a single
> series and title it v6, so that they are together in one thread.
> 
> > 
> > This patch adds a new API to put a host to a suspended state (either
> > Suspend-to-RAM or Suspend-to-Disk) and setup a timed resume to get the
> > host back online, from within libvirt.
> > This uses the RTC wakeup mechanism to set up a timer alarm before
> > suspending the host, so that in-band resume is facilitated by the firing
> > of the RTC alarm, which wakes up the host.
> 
> I'm not as familiar with S3/S4 as I would like to be - am I correct that
> RTC wakeup works for S3 (where the host is maintaining minimal power and
> can thus react to interrupts), but not S4 (where the host has flushed
> completely to disk and powers off, but resumes from the state on disk on
> next power on)?  Or am I misunderstanding these two power-saving states?

Theoretically I think you're right, but in current qemu S3 does a
hardware reset without reseting the memory, so kind of a sleep +
immediate wakeup. S4 is emulated correctly - qemu doesn't have to do
anything special.

> 
> > 
> > The decision to use the RTC Wakeup mechanism to resume the host from
> > sleep was taken in [1]. An initial API was discussed in [2].
> > Some design ideas for the asynchronous mechanism implementation was
> > discussed in [3].
> > 
> > v3:
> >    * Rebased to libvirt 0.9.7
> >    * Added a check to see if alarmTime (suspend duration) is within an
> >      acceptable range.
> > 
> > v2:
> >    * Added an init function which finds out if S3/S4 is supported by the host,
> >      upon the first request to suspend/hibernate.
> >    * Added synchronization/locking to ensure that only one suspend operation
> >      is active at a time.
> > 
> > v1: http://www.redhat.com/archives/libvir-list/2011-September/msg00830.html
> > v2: http://comments.gmane.org/gmane.comp.emulators.libvirt/46729
> > 
> > References:
> > [1]. http://www.redhat.com/archives/libvir-list/2011-August/msg00327.html
> > [2]. http://www.redhat.com/archives/libvir-list/2011-August/msg00248.html
> > [3]. http://www.redhat.com/archives/libvir-list/2011-September/msg00438.html
> > [4]. http://www.redhat.com/archives/libvir-list/2011-November/msg00378.html
> > 
> 
> -- 
> Eric Blake   eblake at redhat.com    +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
> 



> --
> libvir-list mailing list
> libvir-list at redhat.com
> https://www.redhat.com/mailman/listinfo/libvir-list