On Thu, 31 Aug 2017 16:41:47 +0200 David Hildenbrand <da...@redhat.com> wrote:
> On 31.08.2017 16:35, Cornelia Huck wrote: > > On Wed, 30 Aug 2017 19:05:56 +0200 > > David Hildenbrand <da...@redhat.com> wrote: > > > >> Some time ago we discussed that using "id" as property name is not the > >> right thing to do, as it is a reserved property for other devices. > >> > >> Switch to the term "addr" instead, which matches the definition in the > >> PoP called "CPU address". There is no such thing as cpu number, so > >> rename env.cpu_num to env.cpu_addr. > >> > >> We can get rid of cpu->id now. Keep cpu->index and env->cpu_addr in sync. > >> cpu->index was already implicitly used by e.g. cpu_exists(), so keeping > >> both in sync seems to be the right thing to do. > >> > >> cpu->index will now no longer automatically get set via > >> cpu_exec_realizefn(). For now, we were lucky that both implicitly stayed > >> in sync. > >> > >> Our new cpu property "addr" can be a static property. Range checks can > >> be avoided by using the correct type and the "setting after realized" > >> check is done implicitly. > >> > >> AFAIK, s390x only supports cpu_add and not device_add for cpus. So we > >> should be able to safely rename that property (no the "id" property > >> could properly be used for device_add, which needs an artificial id for > >> identification purposes). > > > > I cannot parse the sentence in the brackets... > > Me too :) > > ...So we should be able to safely rename that property. device_add will > later need the reserved "id" property. Hotplugging a CPU would then look > like this: "device_add host-s390-cpu id=cpu2 addr=2". Yup, that's understandable :)
