KVM: x86: move vm_destroy callback at end of kvm_arch_destroy_vm

TDX needs to free the TDR control structures last, after all paging structures
have been torn down; move the vm_destroy callback at a suitable place.
The new place is also okay for AMD; the main difference is that the
MMU has been torn down and, if anything, that is better done before
the SNP ASID is released.

Extracted from a patch by Yan Zhao.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

arch/x86/kvm/x86.c

@@ -12880,7 +12880,6 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
 	}
 	kvm_unload_vcpu_mmus(kvm);
 	kvm_destroy_vcpus(kvm);
-	kvm_x86_call(vm_destroy)(kvm);
 	kvm_free_msr_filter(srcu_dereference_check(kvm->arch.msr_filter, &kvm->srcu, 1));
 	kvm_pic_destroy(kvm);
 	kvm_ioapic_destroy(kvm);
@@ -12890,6 +12889,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
 	kvm_page_track_cleanup(kvm);
 	kvm_xen_destroy_vm(kvm);
 	kvm_hv_destroy_vm(kvm);
+	kvm_x86_call(vm_destroy)(kvm);
 }
 
 static void memslot_rmap_free(struct kvm_memory_slot *slot)