Runtime tracing reaches the MMC shutdown boundary
The reviewed 3e6941d7 controller was physically run with the exact asj…
restart bundle after independently verified user reset-button recovery.
The new passive readiness gate passed. Fresh reception, /bin/true, proc
setup and uptime passed, followed by all six runtime tracing stages with
returned RC 0 / MATCH 1. The exact initcall_debug parameter’s prior N,
write and Y readback were independently returned. Runtime tracing is now
physically observed; this does not establish product acceptance.
The single /bin/reboot -ff request returned its fresh receipt and
Rebooting.. Seven device-shutdown entry messages followed, ending at:
[ 5.600764] mmcblk mmc1:59b4: shutdown
The full 180-second normal-return deadline expired without a later kernel
Restarting system, SPL or normal login. The controller exited 1. No
additional recovery input was sent. Automatic mainline restart, usable root
and deliberate touch remain UNVERIFIED; tasks 5d.4 and 5b.5 stay open.
An entry message proves that checkpoint was reached, not that its callback completed or caused the stop. The next device’s locks/PM barrier precede its own entry print. The source audit must distinguish the card shutdown path from that later boundary before choosing a bounded comparison.
python3 tools/mainline-drm-initrd-shell-trial.py --mode minimal --runtime-shutdown-trace \
--bundle /nix/store/asj7l4zj5jrjkgng4nrcx3y72p7jf7aa-k230-mainline-drm-trial-boot-files \
--manifest "$HOME/tmp/k230-mainline-restart-board/candidate-manifest.json" \
--normal-report "$HOME/tmp/k230-mainline-restart-board/normal-report.json" \
--log "$HOME/tmp/k230-mainline-restart-board/runtime-ready-trace-uart.log" \
--result "$HOME/tmp/k230-mainline-restart-board/runtime-ready-trace-result.json"
Observation is a coordinator reconstruction using the production fresh-marker parsers against the complete private log. This controller revision did not persist a result on normal-return timeout; the requested result file was absent. The record preserves that limit, exact selected bundle/system/artifact identities, raw-log hash/timestamp, protected preflight and returned stages. Excerpt redacts protocol nonces and labels its timeout curation; normal-system raw output remains private.
The separately reviewed camera still shows panel boot text,
with glare/angle limits recorded in provenance. No new finger
acceptance or native capture was obtained. Root held the board/UART/camera in
~/tmp/k230-mainline-probe-integration, branch integrate/mainline-probe-path,
bounded base 3e6941d7. No kernel rebuild, protected boot-file/profile change
or persistent boot-selection write was performed. The operator then pressed reset. Protected normal postflight
passed with a fresh boot ID, exact system/profile/kernel/init, three active
shell services and all eight hashes unchanged. The first 45-second CR-only
check observed normal boot progress but no prompt; a second bounded check
passed. No commands were sent before the fresh normal prompt. This manual
recovery does not satisfy automatic mainline restart.
The subsequent MMC source audit traces callback/next-device waits and the missing clock consumers. The reviewed clock-comparison preparation passes 96 host tests and gives the exact fresh-path operator command after protected recovery. It additionally preserves returned tracing facts in a structured normal-timeout result. The comparison is physically unperformed; it tests global clock-cleanup dependence, not a particular gate or culprit.