Re: [PATCH] virt: tdx-guest: Deprecate legacy IOCTL-based interface for quote generation

[Dan Williams <dan.j.williams@intel.com>]

Haitao Huang wrote:
> On Wed, 31 Jan 2024 13:05:16 -0600, Dave Hansen <dave.hansen@intel.com>  
> wrote:
> 
> > On 1/31/24 10:18, Nikolay Borisov wrote:
> >>> There is no clear consensus from this changelog, nor the tags about  
> >>> this
> >>> approach.  It would be much appreciated if you could make this  
> >>> changelog
> >>> more freestanding and  make it clear that this approach is acceptable  
> >>> to
> >>> both the authors and users of the code you are modifying.
> >>
> >> So how about something along the lines of:
> >>
> >> IOCTL based interface was the natural choice for interacting with the
> >> quote generation machine at a time when there wasn't anything better. In
> >> its current form it's only usable for local attestation but that use
> >> case is also going to be switched to the config-tsm interface. In
> >> preparation for this introduce a Kconfig option which allows the user to
> >> choose to disable the legacy code.
> >
> > That's an interesting theory.  Do the users and authors of the TDX code
> > you're Kconfig'ing away agree with you?
> >
> 
> As DanM pointed out [1], this ioctl() is not a duplicate of the current  
> configfs-tsm ABI. And Local Attestation is a use case mentioned. Can we  
> wait until there are two platforms requiring this interface and we have a  
> clear idea what the configfs-tsm interface replacement of it is before  
> deprecating? A Kconfig to deprecate without alternative implemented also  
> does not make sense to me.

Exactly, that's why I asked for the migration path to be made clear in
the changelog for the deprecation patch.

As far as configfs-tsm is concerned it is just a blob transport, whether
the blob is a "quote" or a "report" does not rely matter, but what
matters is users, clearly identifying the blob formats that might be
returned, and limiting backwards incompatible blob format changes.