On 12/14/20 2:58 PM, David Woodhouse wrote:
> On Mon, 2020-12-14 at 13:29 +0000, Joao Martins wrote:
>> We might be missing the case where only shared_info is registered. Something like:
>>
>> if (vcpu->xen.shinfo_set && !vcpu->xen.vcpu_info_set) {
>> offset = offsetof(struct compat_vcpu_info, time);
>> offset += offsetof(struct shared_info, vcpu_info);
>> offset += (v - kvm_get_vcpu_by_id(0)) * sizeof(struct compat_vcpu_info);
>>
>> kvm_setup_pvclock_page(v, &vcpu->xen.shinfo_cache, offset);
>> }
>>
>> Part of the reason I had a kvm_xen_setup_pvclock_page() was to handle this all these
>> combinations i.e. 1) shared_info && !vcpu_info 2) vcpu_info and unilaterially updating
>> secondary time info.
>>
>> But maybe introducing this xen_vcpu_info() helper to accommodate all this.
>
> There was complexity.
>
> I don't like complexity.
>
> I made it go away.
>
Considering what you said earlier, yes, it would be unnecessary complexity.
Joao
