Re: [PATCH 13/17] tpm: Add verification of kernel signing key provenance

[Matthew Garrett <mjg59@srcf.ucam.org>]

On Sat, Oct 10, 2026 at 11:42:58AM +0200, James Bottomley wrote:

> This does strike me as very elaborate, especially the ephemeral key
> bit; would you countenance a simplification?  Since all you want is
> proof that the hibernate image hash matches the one the creating kernel
> made, you could simply create your kernel only NV index as a PCR and
> measure the hash directly to that.  Since user space is barred from
> writing, all you need is an audit of the measurement going into the NV
> register (indeed, it could be an ordinary NV index you simply write the
> hash value to), coupled with a read of PCR 5.  Then your audit log
> directly verifies the hibernate image hash and you don't have to bother
> with the ephemeral keys.

Just to be clear, "ephemeral keys" here refers to the ephemeral signing 
keys and not the generated AK, right? I /think/ that would work, but let 
me ponder that some more.