On 2017-11-10 at 16:39:27 +0100, Paolo Bonzini wrote: > On 04/11/2017 17:54, Paolo Bonzini wrote: > > On 04/11/2017 01:12, Yi Zhang wrote: > >>> > >> Adding Ravi, > >> > >> Does anyone have further comments on current implementation, it is a > >> important feature in our next generation chip-set. > > > > What matters is not the feature, but the use case; without a use case, > > there is no point in including code for SPP in KVM. KVM doesn't use > > VMFUNC or #VE for example, because they are not necessary. > > > > SPP may become useful once we have the introspection interface. Or, if > > another hypervisor uses it, support for nested SPP may be useful (for > > example we support nested VMFUNC and should get nested #VE sooner or > > later, even though the features are not used on bare metal). > > > > Right now, however, supporting SPP does not seem to be particularly > > important honestly. > > Hi Yi Zhang, > > are you going to work on nested SPP? I guess that would be most useful > way to add SPP support to KVM (and you could also test it with > kvm-unit-tests). Hi Paolo, We Haven't planing on the nested support yet, so far there are many hardware assistance work on current SPP implemetation, and apply it in next generration icelake chip-set. Regards Yi. > > Thanks, > > Paolo
