On Tue, 2014-04-29 at 11:01 +0100, Daniel P. Berrange wrote:
On Tue, Apr 29, 2014 at 10:53:00AM +0100, Ian Campbell wrote:
> On Tue, 2014-04-29 at 10:06 +0100, Daniel P. Berrange wrote:
> > IIUC libxl_domain_shutdown() will use the Xen paravirt channel
> > to trigger a controlled shutdown in the guest. Currently we have
> > the following flags defined
> >
> > VIR_DOMAIN_SHUTDOWN_DEFAULT = 0, /* hypervisor choice */
> > VIR_DOMAIN_SHUTDOWN_ACPI_POWER_BTN = (1 << 0), /* Send ACPI event */
> > VIR_DOMAIN_SHUTDOWN_GUEST_AGENT = (1 << 1), /* Use guest agent */
> > VIR_DOMAIN_SHUTDOWN_INITCTL = (1 << 2), /* Use initctl */
> > VIR_DOMAIN_SHUTDOWN_SIGNAL = (1 << 3), /* Send a signal */
> >
> > None of those really map to the Xen paravirt shutdown
>
> It's not a million miles away from VIR_DOMAIN_SHUTDOWN_GUEST_AGENT if
> you consider the bit of the kernel which responds to the paravirt
> shutdown request as an agent of sorts. With Windows PV drivers I think
> this is even literally a userspace component.
>
> Not that having a more explicit type is a bad idea, but I thought I'd
> mention it.
I guess my rationale for suggesting a new flag is that it is conceivable
that Xen could support the virtio-console devices, in which case it would
be possible for someone to configure a Xen guest with the QEMU (or another)
guest agent provided shutdown facility. In such a scenario, it'd thus be
desirable for the API to be able to distinguish Xen paravirt control from
the guest agent.
That makes sense.
Ian.