Re: [PATCH v5 06/14] x86/ioremap: Support hypervisor specified range to map as encrypted

[Sean Christopherson]

On Fri, Feb 17, 2023, Borislav Petkov wrote:
> On Fri, Feb 17, 2023 at 06:16:56AM +0000, Michael Kelley (LINUX) wrote:
> > Is that consistent with your thinking, or is the whole
> > cc_platform_has() approach problematic, including for the existing SEV
> > flavors and for TDX?
> 
> The confidential computing attributes are, yes, features. I've been
> preaching since the very beginning that vTOM *is* *also* one such
> feature. It is a feature bit in sev_features, for chrissakes. So by that
> logic, those SEV-SNP HyperV guests should return true when
> 
> 	cc_platform_has(CC_ATTR_GUEST_SEV_SNP_VTOM);
> 
> is tested.
> 
> But Sean doesn't like that.

Because vTOM is a hardware feature, whereas the IO-APIC and vTPM being accessible
via private memory are software features.  It's very possible to emulate the
IO-APIC in trusted code without vTOM.

> If the access method to the IO-APIC and vTPM are specific to the
> HyperV's vTOM implementation, then I don't mind if this were called
> 
> 	cc_platform_has(CC_ATTR_GUEST_HYPERV_VTOM);

I still think that's likely to caused problems in the future, e.g. if Hyper-V
moves more stuff into the paravisor or if Hyper-V ends up with similar functionality
for TDX.  But it's not a sticking point, the only thing I'm fiercely resistant to
is conflating hardware features with software features.