Re: boot-time vtpm helper (was "Re: <void>")

[Jarkko Sakkinen]

On Thu Aug 3, 2023 at 3:46 PM EEST, Stefan Berger wrote:
>
>
> On 8/3/23 05:06, Jarkko Sakkinen wrote:
> > On Thu Aug 3, 2023 at 11:25 AM EEST, Jarkko Sakkinen wrote:
> >> Hi,
> >>
> >> I have a working PoC for boot-time initialization of vtpm inside
> >> tpm_vtpm_proxy. ATM, it uses the Linux firmware interface to load a ELF
> >> binary for the vtpm, and delivers a communication end for the helper
> >> process.
> >>
> >> It is a great feature with the current narrow scope for continuous
> >> integration. Obviously the scope could be later on extended to e.g.
>
> Since VMs with vTPMs exist, which CI/CD environment would one use this in?
>
> Where does the binary for the vtpm live when it's loaded with the firmware interface?

Obviously vtpm is identified by a filename. The rest of the details are
in the Linux firmware documentation.

> >> from unencrypted plain text to a vTPM living inside SGX enclave.
>
> I would run swtpm inside an SGX enclave using Gramine.

Having a zero configuration and also zero dependency mechanism for
trivial kernel testing is an objective benefit.

> Where does the binary for the vtpm live when it's loaded with the firmware interface?

Either in initrd or with the help of CONFIG_EXTRA_FIRMWARE_* in vmlinux.

> >> from unencrypted plain text to a vTPM living inside SGX enclave.
>
> I would run swtpm inside an SGX enclave using Gramine.

Why would do you want to enforce the use of swtpm?

BR, Jarkko