Re: [PATCH v16 24/45] KVM: arm64: Expose support for private memory

[Suzuki K Poulose]

On 05/08/2026 17:02, Ackerley Tng wrote:
> Steven Price <steven.price@xxxxxxx> writes:
>
> > Select KVM_GENERIC_MEMORY_ATTRIBUTES and provide the necessary support
> > functions.
> >
> > Reviewed-by: Suzuki K Poulose <suzuki.poulose@xxxxxxx>
> > Signed-off-by: Steven Price <steven.price@xxxxxxx>
> > ---
> > Changes since v14:
> >   * Switch to guest_memfd in-place memory conversion and drop support
> >     for KVM_VM_MEMORY_ATTRIBUTES.
> >
> > [...snip...]
> >
> > diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> > index 15d42d1993bb..146c52f867c0 100644
> > --- a/arch/arm64/include/asm/kvm_host.h
> > +++ b/arch/arm64/include/asm/kvm_host.h
> > @@ -1510,6 +1510,8 @@ struct kvm *kvm_arch_alloc_vm(void);
> >
> >   #define vcpu_is_protected(vcpu)		kvm_vm_is_protected((vcpu)->kvm)
> >
> > +#define kvm_arch_has_private_mem(kvm) ((kvm)->arch.is_realm)
> > +
>
> Perhaps related to Fuad's question about pKVM and CCA state: would
> conditioning this on is_realm mess up pKVM?

Not really. The comment is about unifiying the conditions which
both pKVM and Realm selects. Right now, only CCA uses guest_memfd.
When pkvm comes along to use guest_memfd, this could be something
like:

kvm->arch.needs_private_mem  (set by both pKVM and Realm) ?

Cheers
Suzuki


> >   int kvm_arm_vcpu_finalize(struct kvm_vcpu *vcpu, int feature);
> >   bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
> >
> > --
> > 2.43.0