Re: git: bfe4be85a935 - main - tpm_crb: wait for idle and locality release after each command

From: Olivier_Cochard-Labbé <olivier_at_freebsd.org>
Date: Tue, 10 Nov 2026 12:02:54 UTC
On Tue, Nov 10, 2026 at 11:28 AM Michael Tuexen <tuexen@freebsd.org> wrote:

> > Can you try the following to help dump what is going on?
>
> Sure. I get:
>
> Nov 10 11:23:06 fb13 kernel: tpmcrb0: Failed to relinquish locality:
> LOC_STATE before=0x83 last=0x83 LOC_STS=0 CTRL_REQ=0 CTRL_STS=0x2
>

Thanks,

so if my understanding is correct (Kevin will confirms if not):

0x83 in binary is 1000 0011:
- bit 7 VALID = 1: the register contents are meaningful.
- bit 1 ASSIGNED = 1: a locality is assigned.
- bit 0 ESTB (tpmEstablished) = 1: a firmware flag, unrelated here.
- bits 2 to 4 = 0: the assigned locality is locality 0, FreeBSD's.

Reading each field:
- before=0x83 last=0x83: the state is the same before the release and
200 ms after it. The TPM still says "locality 0 is assigned". The
driver waits for ASSIGNED to clear, which never happens, hence the
timeout and the log line every 10 s.
- LOC_STS=0: the Granted bit is clear. From the per-locality status's
point of view, FreeBSD no longer holds the locality. This contradicts
LOC_STATE.
- CTRL_REQ=0: no request is pending. The GO_IDLE sent just before was
acknowledged.
- CTRL_STS=0x2: idle bit set, error bit clear. The TPM is healthy and idle.

The TPM works fine and handles the release. LOC_STS shows it, and the
commands succeed. Only LOC_STATE, the register the driver polls, is
never updated on this Pluton firmware. So waiting for LOC_STATE can't
work on this hardware.
The Linux driver was disabling it too, so let’s do the same on our side.

Here is a proposal of a patch:
https://ron-dev.freebsd.org/FreeBSD/src/pulls/102

Sorry for spamming your console log :-(

Olivier