Hi Ilias, On Mon, 5 Oct 2026 at 12:08, Ilias Apalodimas <[email protected]> wrote: > Yeah, I'm not particularly glued on "let's make it on probe" either. I > am fine with the event as well. I just need a proper justification for > it, because initializing it during probe seems much simpler.
I looked at what the probe variant would need: 1. Nothing probes the TPM on a plain booti/bootm, so post_probe() alone does not help. The uclass would also need DM_FLAG_PROBE_AFTER_BIND. 2. dm_autoprobe() runs in SPL and before and after relocation. With a bootph-* tagged TPM node (e.g. chromebook_coral) the self-test would run in each phase unless extra conditions are added, whereas the event runs exactly once. 3. The event is purely additive. With the option disabled nothing changes, and the uclass, the existing callers (tpm command, efi_tcg2, tpm_tcg2) and the tests stay as they are. Having looked at both, I would personally prefer to keep the event version. Regards, Michael
