On 4/25/2025 8:57 PM, David Hildenbrand wrote:
> On 07.04.25 09:49, Chenyi Qiang wrote:
>> To manage the private and shared RAM states in confidential VMs,
>> introduce a new class of PrivateShareManager as a child of
>> GenericStateManager, which inherits the six interface callbacks. With a
>> different interface type, it can be distinguished from the
>> RamDiscardManager object and provide the flexibility for addressing
>> specific requirements of confidential VMs in the future.
>>
>> Signed-off-by: Chenyi Qiang <chenyi.qi...@intel.com>
>> ---
>
> See my other mail, likely this is going into the wrong direction.
>
> If we want to abstract more into a RamStateManager, then it would have
> to have two two sets of states, and allow for registering a provider for
> each of the states.
>
> It would then merge these informations.
>
> But the private vs. shared provider and the plugged vs. unplugged
> provider would not be a subclass of the RamStateManager.
>
> They would have a different interface.
>
> (e.g., RamDiscardStateProvider vs. RamPrivateStateProvider)
Got it! Before the real use case (guest_memfd + virtio-mem) comes, I
would keep the original design. Maybe seek the new framework after that
work.
>
