On Mon, 2020-12-14 at 12:53 +0000, Joao Martins wrote:
> Ah OK -- I didn't notice @pv_time_enabled before suggesting this.
>
> An helper would probably suffice without sacrificing footprint in
> data structures (because you will add 3 of these). The helpers would be, say:
>
> kvm_xen_shared_info_set(vcpu->kvm)
> kvm_xen_vcpu_info_set(vcpu)
> kvm_xen_vcpu_runstate_set(vcpu)
>
> And maybe backed by a:
>
> kvm_gfn_to_hva_cache_valid(&v->kvm->arch.xen.shinfo_cache)
>
> Which would check whether a cache is initialized or valid.
>
> But anyhow, I don't feel strongly about it, specially if there's an existing
> one which sort of sets the precedent.
Maybe if I end up adding a fifth. But right now they aren't even taking
any space. They're just filling what would otherwise be a hole...
/* Xen HVM per vcpu emulation context */
struct kvm_vcpu_xen {
u64 hypercall_rip; // 0x00
u32 current_runstate; // 0x08
bool vcpu_info_set; // 0x0c
bool vcpu_time_info_set; // 0x0d
bool runstate_set; // 0x0e
/* I can add one more before I care */ // 0x0f
struct gfn_to_hva_cache vcpu_info_cache; // 0x10
struct gfn_to_hva_cache vcpu_time_info_cache; // 0x38
struct gfn_to_hva_cache runstate_cache; // 0x60
u64 last_steal; // 0x88
u64 last_state_ns; // 0x90
};
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
