On 16/06/2017 15:43, Adalbert Lazar wrote:
> + while (!list_empty(&kvm->access_list)) {
> + struct kvmi_mem_access *m =
> + list_first_entry(&kvm->access_list, struct kvmi_mem_access,
> + link);
> +
> + list_del(&m->link);
> + INIT_LIST_HEAD(&m->link);
> +
> + kvmi_apply_mem_access(vcpu, m->gfn, m->access);
> + }
How does this work when multiple VCPUs are running with different MMU
roles? One VCPU is emptying the access_list for all, but
kvm_mmu_set_spte is using for_each_shadow_entry per-VCPU.
I'm really afraid of introducing subtle bugs, with possible security
effects. I'm not really able to provide a suggestion yet, since I
haven't grasped the protocol entirely.
Paolo
