Xen_ComputerSystem RequestStateChange
When testing the Xen_ComputerSystem RequestStateChange, I noticed that the shutdown state wasn't working but the disabled state was. Was there a difference for these(disabled (3) and shutdown (4))? Is there a reason why the shutdown isn't working but disabled is? Thanks, kwoodson
Kenny Woodson wrote:
When testing the Xen_ComputerSystem RequestStateChange, I noticed that the shutdown state wasn't working but the disabled state was. Was there a difference for these(disabled (3) and shutdown (4))? Is there a reason why the shutdown isn't working but disabled is?
Right now, shutdown isn't supported. The DMTF Virtual System Profile lists an extra step for shutdown - the provider needs to tell the software running on the guest to shutdown. Once the software has terminated its tasks and shut itself down, the provider then treats the shutdown action in the same manner it treats disable. The steps to indicate to the guest that the OS needs to shutdown are involved. This is probably out of the project's scope, for awhile at least. It would be possible to have shutdown and disable behave the same way, but that wouldn't conform to the DMTF profile. Dan - thoughts? -- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com
KR> Right now, shutdown isn't supported. The DMTF Virtual System Profile KR> lists an extra step for shutdown - the provider needs to tell the KR> software running on the guest to shutdown. Once the software has KR> terminated its tasks and shut itself down, the provider then treats KR> the shutdown action in the same manner it treats disable. KR> The steps to indicate to the guest that the OS needs to shutdown are KR> involved. This is probably out of the project's scope, for awhile at KR> least. KR> It would be possible to have shutdown and disable behave the same way, KR> but that wouldn't conform to the DMTF profile. KR> Dan - thoughts? For paravirt Xen, issuing the shutdown command actually triggers an action in the guest kernel to start a shutdown (which becomes the equivalent of "shutdown -h now"). For fullvirt and KVM guests, a shutdown action should be simulated via ACPI (the equivalent of pushing the power button on an ACPI-enabled physical machine). I think I recall there being some plumbing issues to make this really work for KVM guests, but I think we should assume that it works. From what I can tell, requesting the shutdown state is equivalent to requesting the disabled state from our perspective. It would be trivial to make both of these target states perform the same action and have the same result. Doing so seems reasonable to me... anybody have a comment about this? -- Dan Smith IBM Linux Technology Center Open Hypervisor Team email: danms@us.ibm.com
Dan Smith wrote:
KR> Right now, shutdown isn't supported. The DMTF Virtual System Profile KR> lists an extra step for shutdown - the provider needs to tell the KR> software running on the guest to shutdown. Once the software has KR> terminated its tasks and shut itself down, the provider then treats KR> the shutdown action in the same manner it treats disable.
KR> The steps to indicate to the guest that the OS needs to shutdown are KR> involved. This is probably out of the project's scope, for awhile at KR> least.
KR> It would be possible to have shutdown and disable behave the same way, KR> but that wouldn't conform to the DMTF profile.
KR> Dan - thoughts?
For paravirt Xen, issuing the shutdown command actually triggers an action in the guest kernel to start a shutdown (which becomes the equivalent of "shutdown -h now"). For fullvirt and KVM guests, a shutdown action should be simulated via ACPI (the equivalent of pushing the power button on an ACPI-enabled physical machine). I think I recall there being some plumbing issues to make this really work for KVM guests, but I think we should assume that it works.
IIRC, someone in the Xen community is doing similar work to inject ACPI events in HVM guests. Additionally, many existing Xen HVM guests using PV drivers support the "shutdown -h now" semantics.
From what I can tell, requesting the shutdown state is equivalent to requesting the disabled state from our perspective.
Agreed. Regards, Jim
participants (4)
-
Dan Smith -
Jim Fehlig -
Kaitlin Rupert -
Kenny Woodson