Add support for skipping attestation report verification

[conradgrobler]

We want to be able to verify attestation evidence as much as possible in testing, even when we are not running in an environment that supports hardware-based attestation.

This adds an option to explicitly skip verification of the attestation report signature and adds a test for using this when running without AMD SEV-SNP.

[thmsbinder]

Would the client know/care whether running on AMD or Intel hardware? Consequently the client would need to pass reference values for all platforms at once. Put differently, the lack of oneof was by intention.

[conradgrobler]

Good point. Reverted and added InsecureReferenceValues to support non TEE platforms.

[thmsbinder]

Maybe it is a Fake not a Mock ?

[conradgrobler]

Done.

[thmsbinder]

NONE may cause trouble in C++ later as it appears in the global namespace - maybe TEE_PLATFORM_NONE ? cf. yaqs/6578714921009152

[conradgrobler]

Done.

[thmsbinder]

OOC, did you make this with a hex editor (i.e. by hand)? Where does it come from?

[conradgrobler]

This was taken from the debug log of an instance when running without AMD SEV-SNP.

[conradgrobler]

Thanks

[conradgrobler]

This was taken from the debug log of an instance when running without AMD SEV-SNP.

[conradgrobler]

Good point. Reverted and added InsecureReferenceValues to support non TEE platforms.

[conradgrobler]

Done.

[conradgrobler]

Done.

[tiziano88]

Why AMD SEV-SNP? Should we send something more obviously fake / empty in that case? Just an empty value?

[conradgrobler]

This has been the behaviour of Stage 0 for a while. Since Stage 0 currently only supports AMD SEV-SNP this seemed the easiest until we have the need for very different attestation reports such as TDX.

It also keeps the size of the struct the same, which simplifies logic in a few places, and we still populate the "report_data" field so that the DICE chain is rooted in the attestation report even for fake reports.

Created #4786 to track this.