On Fri, 6 May 2016 15:44:10 +0530 Bharata B Rao <bhar...@linux.vnet.ibm.com> wrote:
> On Fri, May 06, 2016 at 10:57:21AM +0200, Igor Mammedov wrote: > > On Thu, 31 Mar 2016 14:09:19 +0530 > > Bharata B Rao <bhar...@linux.vnet.ibm.com> wrote: > > > > > Set up device tree entries for the hotplugged CPU core and use the > > > exising RTAS event logging infrastructure to send CPU hotplug notification > > > to the guest. > > > > perhaps you could reuse: > > > > [RFC,20/42] machine: add cpu-hotplug machine option > > http://patchwork.ozlabs.org/patch/617516/ > > > > to enable cpu hotplug explicitly, > > and by default machine would be hotplug less saving resources. > > We explicitly enable hotplug for specific machine type versions by > setting sPAPRMachineClass.dr_cpu_enabled. So we check it to ensure > that there are no DR connector objects (the resources needed for hotplug) > created for a machine that doesn't support CPU hotplug. Similar is > the case for memory hotplug. > > So you are suggesting that we switch from sAPPR specific check > (sPAPRMachineClass.dr_cpu_enabled) to the above MachineState.cpu_hotplug ? yep, i.e. make user explicitly say that he want cpu hotplug enabled. instead of allocating resources for it when user doesn't care about it. It saves some memory/io ports for x86, not sure what it would save for sPAPR, but at least one doesn't have to guess if hotplug is needed by looking at smp_cpus & max_cpus. > > Regards, > Bharata. > >
