On Mon, Sep 22, 2025 at 01:22:13PM -0400, Ben Boeckel wrote: > On Mon, Sep 22, 2025 at 19:43:16 +0300, Jarkko Sakkinen wrote: > > From: Jarkko Sakkinen <[email protected]> > > > > Open code tpm_buf_append_hmac_session_opt() because it adds unnecessary > > disperancy to the call sites (and reduces the amount of code). > ^^^^^^^^^^ > > "discrepancy" as in "difference"? But that doesn't feel like the right > usage either. Perhaps "unnecessary abstraction"? Also, open coding it > reduces the amount of code, so some clarification to not read as > something else that "it" (`tpm_buf_append_hmac_session_opt`) does would > be clearer.
Fair points. I'll re-edit the commit message and try to address the issues you reported. Intend of these changes is to essentially uncover the code paths so that we know how to wrap it up better than it is wrapped up right now. Also, they help to reveal possible regression paths. So while not functional per se, they do serve a purpose. Once these fixes have been applied I'll start to look up the call patterns and try to find a model where essentially we can transform a TPM command to HMAC wrapped TPM command i.e., from tpm_buf to tpm_buf operation where both sides of the function are TPM commands. That way we can better selectively use the feature and it is easier to fixup up e.g., a persistent parent key because key generation is a huge bottleneck. > > Thanks, > > --Ben
