Re: [PATCH 3/3] x86/sev: add a SVSM vTPM platform device

["Jarkko Sakkinen" <jarkko@kernel.org>]

On Thu Jan 23, 2025 at 12:09 PM EET, Stefano Garzarella wrote:
> On Thu, Jan 23, 2025 at 11:50:40AM +0200, Jarkko Sakkinen wrote:
> >On Wed Jan 22, 2025 at 11:29 PM EET, Dionna Amalie Glaze wrote:
> >> I can appreciate this viewpoint. It even surfaced Microsoft's fTPM
> >> paper to me, which solves some interesting problems we need to solve
> >> in SVSM too. So thanks for that.
> >>
> >> Just to clarify, you're not asking for SVSM to implement the TIS-MMIO
> >> interface instead, but rather to use the fTPM stack, which could make
> >> SVSM calls a TEE device operation?
> >
> >I don't really know what I'm asking because this is barely even a
> >PoC, and I state it like this knowingly.
> >
> >You should make the argument, and the case for the solution. Then
> >it is my turn to comment on that scheme.
>
> I'll check if I can use fTPM, in the meantime I had started to simplify
> this series, avoiding the double stack and exposing some APIs from SEV
> to probe the vTPM and to send the commands. The final driver in
> drivers/char/tpm would be quite simple.
>
> But I'll try to see if reusing fTPM is a feasible way, I like the idea.
>
> >
> >That said, I would not give high odds for acceptance of a duplicate
> >TPM stack succeeding.
>
> Got it ;-)
>
> Thanks to everyone for the helpful feedbacks!
>
> I've been a bit messy these days and I'm in FOSDEM next week, so I hope
> not to take too long for the v2.

Yeah, OK one thing that I want to say.

Nail the story. What is it about what is the problem what is the
motivation to solve it etc. If you have all that properly written
up then it is easier to forgive not that well nailed code and
give reasonable arguments.

And don't rush, I have all the time in the world ;-)

> Stefano

BR, Jarkko