Re: [libvirt] [PATCH RFC 3/4] Create new virDrvDomainSuspendFlags API
On 01/26/2012 07:31 AM, Michal Privoznik wrote:
On 26.01.2012 15:17, Eric Blake wrote:
> On 01/26/2012 07:06 AM, Michal Privoznik wrote:
>> define these flags:
>>
>> VIR_DOMAIN_SUSPEND_SLEEP = 0, /* Suspend to RAM */
>> VIR_DOMAIN_SUSPEND_HIBERNATE = (1 << 0), /* Suspend to disk */
>
> Not quite right. Our policy when fixing old APIs that forgot a flags
> argument is that the new API with flags 0 must be identical to the old
> API. Which means you need:
>
> VIR_DOMAIN_SUSPEND_PAUSE = 0, /* pause CPUs */
> VIR_DOMAIN_SUSPEND_SLEEP = 1<<0, /* Suspend to RAM, power stays on */
Honestly, in case of virtual machines I can't see the difference between
these two. But okay, I'll rewrite it.
VIR_DOMAIN_SUSPEND_PAUSE is what we have now, when you click the pause
button in virt-manager - it means that qemu stops all guest CPUs, with
no input from the guest. Resuming from this state requires
virDomainResume to issue a monitor command.
VIR_DOMAIN_SUSPEND_SLEEP involves the guest. S3 might be a very fast op
with qemu, but it is not an instant op - the guest CPU continues to run,
and the guest really does flush state to RAM. Furthermore, qemu is
implementing S3 so that you can re-wake the guest via input (such as a
keypress into a serial device, or a mouse click), with no intervening
monitor command.
That said, qemu is also implementing a new monitor command to wake up a
guest that is in S3 state; so we _also_ need to add a new API
virDomainResumeFlags, to give use the flexibility to expose that new
monitor command.
--
Eric Blake eblake(a)redhat.com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
Attachments:
- signature.asc (application/pgp-signature — 620 bytes)