Re: [PATCH] tpm/tpm_crb: Access locality for non-ACPI and non-SMC start method

[Jason Gunthorpe]

On Tue, Aug 22, 2017 at 04:28:54PM -0500, Jiandi An wrote:

> I'm sorry perhaps I didn't fully understand the workaround specific to Intel
> PPT.  In previous patch thread, you mentioned the following where
> a platform could report to require start method 2 (ACPI start) which is
> sm = ACPI_TPM2_START_METHOD, and actually requires start method 8, which
> is sm = ACPI_TPM2_COMMAND_BUFFER_WITH_START_METHOD.

I'm also not sure.

To be clear, my desire to see a test that triggers only for the Intel
chips with the problem, and is written in a way that matches exactly
the ACPI data from the broken chip - so things like !CRB are not what
I want to see..

In that light the example I gave was probably not well thought out,
but I also do not understand the exact conditions needed for the Intel
work around either. Hopefully Jarkko can clarify.

Jason