On Thu, 29 Aug 2019, Thomas Gleixner wrote: > On Thu, 29 Aug 2019, Borislav Petkov wrote: > > > On Wed, Jul 10, 2019 at 08:13:11PM +0000, Singh, Brijesh wrote: > > > @@ -2060,6 +2067,14 @@ static int __set_memory_enc_dec(unsigned long addr, int numpages, bool enc) > > > */ > > > cpa_flush(&cpa, 0); > > > > > > + /* > > > + * When SEV is active, notify hypervisor that a given memory range is mapped > > > + * encrypted or decrypted. Hypervisor will use this information during > > > + * the VM migration. > > > + */ > > > + if (sev_active()) > > > + set_memory_enc_dec_hypercall(addr, numpages << PAGE_SHIFT, enc); > > > > Btw, tglx has a another valid design concern here: why isn't this a > > pv_ops thing? So that it is active only when the hypervisor is actually > > present? > > > > I know, I know, this will run on SEV guests only because it is all > > (hopefully) behind "if (sev_active())" checks but the clean and accepted > > design is a paravirt call, I'd say. > > No. sev_active() has nothing to do with guest mode. It tells whether SEV is > active or not. So yes, this calls into this function on both guest and > host. The latter is beyond pointless. Oops. sme != sev. But yes, can we please hide that a bit better....
