Re: [PATCH v9 1/6] x86/tdx: Add TDX Guest attestation interface driver

[Sathyanarayanan Kuppuswamy <sathyanarayanan.kuppuswamy@linux.intel.com>]

Hi,

On 8/22/22 2:44 PM, Dave Hansen wrote:
> On 8/22/22 14:36, Borislav Petkov wrote:
>> Which makes my initial suggestion of calling this whole guest
>> functionality a "tdx" driver not such a bad idea... Depends on
>> whether there will be a split at all or it'll continue gaining more
>> functionality.
> 
> Yep, let's get the crystal ball out.
> 
> TDX folks:
> 
> What other ioctl()s are in the pipeline for the guest side?

In addition to the GetReport support currently implemented, we have
following two attestation related IOCTLs in the pipeline.

1. GetQuote     - Adds support to get signed Quote for the given TDREPORT. This
                  is currently in the review process.
2. VerifyReport - Verifies whether given Reportdata is generated in
                  the current platform. It is only enabled in v1.5 TDX
                  Module specification.

In addition to above, I think there is a possibility to add IOCTL to get
storage keys from the ACPI SVKL table. Storage Volume Key Table (SVKL) is
used by the VBIOS to share keys required to access encrypted drives. Although
we don't have a clear requirement, I suspect that we might add IOCTL for it.

Kirill/Isaku/Kai, If I missed any other IOCTL requirements, please add it.
              
> 
> What ioctl()s are in the pipeline for the host side?  Are they all part
> of /dev/kvm, or are there any TDX-specific "drivers" for the host?
> 
> We want to avoid both:
> 
>  1. A driver called /dev/tdx (or "tdx-guest) which is only and will only
>     ever do TDX guest attestation.

>  2. A driver called /dev/tdx-guest-attest which shares a ton of
>     functionality with some future TDX guest feature like
>     /dev/tdx-guest-snazzy-feature-foo.  Then, a new driver every time
>     a new snazzy TDX feature shows up.

-- 
Sathyanarayanan Kuppuswamy
Linux Kernel Developer