Re: [PATCH v6 05/10] x86/tdx: Handle port I/O

[Sean Christopherson]

On Thu, Sep 23, 2021, Tom Lendacky wrote:
> On 9/23/21 12:24 PM, Kuppuswamy, Sathyanarayanan wrote:
> > Any reason for using sev_enable_key over CC attribute? IMO, CC attribute
> > exist to generalize the common feature code. My impression is SEV is
> > specific to AMD code.

Unless CC attributes have static_<whatever> support, that would add a CMP+Jcc to
every I/O instruction in the kernel.

> When the SEV series was initially submitted, it originally did an
> sev_active() check. For various reasons a static key and the
> sev_key_active() call was requested.
> 
> My suggestion was to change the name to something that doesn't have SEV/sev
> in it that can be used by both SEV and TDX. The sev_enable_key can be moved
> to a common file (maybe cc_platform.c) and renamed. Then
> arch/x86/include/asm/io.h can change the #ifdef from CONFIG_AMD_MEM_ENCRYPT
> to CONFIG_ARCH_HAS_CC_PLATFORM.
> 
> Not sure if anyone else feels the same, though, so just my suggestion.

+1 to a static key to gate high volume and/or performance critical things that
are common to SEV and TDX.