# Project Executive Intent: SMI Firmware

Status: active
Type: decision
Role: project-executive-intent-v1
Updated: 2026-09-24

## Current release-gate reconciliation — 2026-09-24

SMI 00-05-01 remains NOT PRODUCTION RELEASED. This capsule and the linked current release-gate review supersede earlier candidate, policy, pending-test and next-action statements. Dated observations below remain history; do not replay their completed tests or use them to override newer evidence.

The corrected IMX SHA-256 is `4E4321E5BDA3942BF70ED9304B935DF33DC40E5FA1E9B8A4AEF60DCA6ED3B97A`, complete main flash `1ACC79159FCD3F97BD994AA086106348BEC6B418A87A8C409B707156978BFC2D`. Local IMX hashing reconfirmed it on September 24. The second CID capture now occurs only after format/reopen. A installed this image, passed an ordinary retained-FAT32 downgrade to exact 00-04-11 and an ordinary return to the corrected image, then restored same-CID HPT2. B installed the same exact image and restored its own same-CID HPT2. Both raw stores were open and updates idle at the last September 23 captures. No new bench action or fresh live-state claim is made in this review.

The owner-directed ordinary updater and service 151 share admission, picker, confirmation and worker. Valid older SMI can use the ordinary path with target-driven FAT32 preparation; the former blanket ordinary personality floor/service-only downgrade is superseded. Missing/unreadable/busy required media and invalid identities fail closed. Verified FAT32 is retained; another readable layout is formatted only after explicit consent. DisplayTest 01-00 remains excluded; SMI 00-05-00 remains blocked/abandoned. No CVO/DVO release or firmware work is implied.

The retained-FAT32 901 cause and its bounded correction are separate from the earlier legacy 00-04-11 Success-with-unchanged-flash. That legacy observation remains unisolated despite the later exact return pass. Final retained-FAT32 warning wording and legacy post-downgrade CID were not captured; service 467 is unavailable in 00-04-11. Preserve these limits for explicit disposition, not invented evidence or automatic retest. DVO's intermittent legacy 901 investigation stays owner-closed with the alternate-verified-USB-stick workaround. A post-install power cycle is accepted behavior, not standing authorization for a new power action.

The corrected source is still dirty/uncommitted. The immutable candidate pointer still selects historical `00.05.01-1bf8afe6aeba-6476100e5ae5`; no active release cohort/register exists for the corrected image. Freeze, exact-candidate evidence applicability, warning/cancel and applicable refusal/fault-recovery coverage, remaining B transition/logging/export/retention, evidence-limit disposition and a separate release decision remain open. A's earlier fixture/return/logging/export/retention passes are credited at their recorded hashes; transfer requires a change-impact rationale, not relabeling them as corrected-image tests.

The smallest open bench action is a current-A ordinary warning/cancel-only check, after a fresh serial-bound matched-firmware/CID/idle baseline, verified target media and a reviewed safe cancel point. Stop before install confirmation; verify unchanged flash, CID, HPT2 and idle requests after cancel. It cannot recreate the missing retained-FAT32 prompt while the card is HPT2. No programming, formatting, power operation, SD swap, staging, commit or publication is authorized here. DisplayTest stays read-only and unrelated abandoned work stays abandoned.

Canonical gate matrix and prerequisites: [current release-gate review](docs/firmware-update-validation-and-recovery.md#current-release-gate-reconciliation--2026-09-24). Primary evidence: `work/evidence/smi-retained-fat32-correction-candidate-20260923.json` and `work/evidence/smi-displaytest-handoff-reconciliation-20260923.md`. MAP-0009 and its owning IDEA-0007 outcome remain open; no lifecycle completion is inferred.

## A normal forward and accepted ordinary return completed — 2026-09-23

Controller A completed exact DisplayTest 01-01 to accepted SMI 00-05-01 normal installation, explicit HPT2 initialization, fresh logging/history, six-row healthy customer USB export with independent CSV hash/content check, and post-restart HPT2 retention. Its ordinary SMI-to-DisplayTest return selected exact `7520708_DispTest_FW_01-01.imx`, accepted the conditional FAT32 warning, and yielded a complete four-MiB readback matching the controlled DisplayTest payload SHA-256 `ACA37CEBA6621701CAF3909D8712EB672872940ECC0E835DE690D6AD829D6D81`. A is FAT32-ready with the unchanged complete CID `05000C61736466672200001C8901A34F`; focused SD read/write passed. The first DisplayTest NVRAM screen showed Read Error / Not Tested. After an owner-operated restart, both displayed statuses and read-only target flags were OK/OK, with CID/FAT32 unchanged. The owner accepts a power cycle after new firmware installation and requests no mitigation. No Erase EEPROM service or firmware change was performed. The owner also confirmed post-return About 01-01, responsive Home touch and no FAT32 warning. Evidence: `work/evidence/smi-a-hpt2-post-restart-retention-20260923.json`, `work/evidence/smi-a-customer-export-20260923.json`, `work/evidence/smi-a-ordinary-displaytest-return-nvram-hold-20260923.json`, `work/evidence/smi-a-displaytest-nvram-restart-resolution-20260923.json`, and the SMI handoff ledger.

The A forward and ordinary return path passes with the accepted post-install power-cycle behavior. A below-floor refusal, representative service rollback and fault/recovery tests, plus B forward/return and remaining release evidence, are open. The DisplayTest SD Qualification Full Run was not performed or claimed. Earlier dated NVRAM-hold sections below record the resolved first-boot observation.

## A ordinary DisplayTest return has an NVRAM hold — 2026-09-23

The owner selected exact `7520708_DispTest_FW_01-01.imx` through the ordinary SMI Settings updater, confirmed the conditional FAT32 data-loss warning, and reported Update Success. A read-only full-flash capture matches the controlled DisplayTest 01-01 payload SHA-256 `ACA37CEBA6621701CAF3909D8712EB672872940ECC0E835DE690D6AD829D6D81`. A is FAT32-ready with valid unchanged CID `05000C61736466672200001C8901A34F`, no storage error, and a focused SD read/write PASS. The owner then supplied a photo of the DisplayTest NVRAM screen: “NVRAM Status: Read Error” and “Test Status: Not Tested.” Read-only target flags confirm invalid NVRAM and no completed power test. Evidence: `work/evidence/smi-a-ordinary-displaytest-return-nvram-hold-20260923.json` and the handoff ledger.

The matched sources show SMI writes 1,028-byte SPI NVRAM records, whereas DisplayTest 01-01 reads/writes 1,024-byte records with a different CRC position. A first-boot format mismatch is plausible, but SPI/hardware failure remains possible. DisplayTest only updates the validity flag on boot and periodically writes its own format. The owner proposed Service > Erase EEPROM; that service zeros configuration data flash, not SPI NVRAM, so it has not been run. A separately authorized DisplayTest restart is the focused discriminator. The ordinary A return image and FAT32 transition passed, but end-to-end return qualification is held pending NVRAM resolution and operator-facing checks. A refusal/rollback/recovery and B remain open. Earlier return-pending sections below are history.

## A HPT2 post-restart retention verified — 2026-09-23

The owner explicitly approved an A power-off/restart with the SD card installed and USB absent. A matched-SMI 10:33 baseline showed HPT2 ready, same complete CID `05000C61736466672200001C8901A34F`, retained oldest normal record sequence 2, live chart data and no pending update/export. The owner reported A booted 00-05-01 with pre-restart history visible. Fresh 10:35 read-only state found the same valid CID and representative-sector read, HPT2 ready, completed chart cache, retained oldest record sequence 2 / HPT timestamp 306930000, and 16 selected records in the postboot chart rebuild. New queued logging resumed with two completions and no drops since boot. Preboot cumulative counters reset on boot and are not compared directly. Evidence: `work/evidence/smi-a-hpt2-post-restart-retention-20260923.json` and the handoff ledger.

The A normal forward installation, HPT2 initialization, fresh logging/history, customer USB export with host CSV validation, and focused post-restart retention gates have passed. The USB CSV copy is preserved under SMI evidence; the drive is unmounted at the Pi. Next is the controlled ordinary A return to DisplayTest 01-01 with explicit FAT32 media preparation and fresh CID/status checks. A refusal/rollback/fault matrix and B remain open. No SD Qualification Full Run pass is claimed. Earlier dated retention-pending sections below are history.

## A customer HPT2 export and independent CSV check passed — 2026-09-23

On A, ordinary History/Graph showed recent records. The first Settings > Export Log attempt was rejected as “Range too short” before an export request. The owner retried the 10:20–10:26 controller-clock range and reported Success. Matched-SMI read-only status shows request 1 published, complete 1, USB status 0, final USB phase complete, six healthy records selected with no gaps, and no active export. The `IMX2` USB drive was mounted read-only on `srv-rpi00.local`: the 975-byte CSV contains six 10:21–10:26 minute rows and a COMPLETE footer matching target status. Its local evidence copy has the same SHA-256 `28C0D6E2005F868B76BC2EB77219DE67CC894E57B42354996D605E85FA8A041A`; the drive was unmounted cleanly. The A baseline and post-format result link the unchanged complete CID `05000C61736466672200001C8901A34F`. Evidence: `work/evidence/smi-a-customer-export-20260923.json`, `work/evidence/smi-a-customer-export-20260923.csv`, and `work/evidence/smi-displaytest-handoff-reconciliation-20260923.md`.

This closes A's normal fresh logging, history view and customer export/content gate. The next A test is post-restart HPT2 retention, which requires separate owner authority for a controller power operation. Then A return/refusal/rollback/fault recovery and B transition gates remain. No SD Qualification Full Run pass is claimed. Earlier dated export-pending sections below are history.

## A HPT2 ready and fresh logging verified — 2026-09-23

The owner reported SMI Format 2 **Success / log records erased**. A read-only matched-SMI capture found initialization complete, HPT2 ready, valid representative-sector read, and the unchanged complete CID `05000C61736466672200001C8901A34F`. Subsequent A captures showed acknowledged records rising from the one commissioning record to four, valid retained bounds, completed initial chart load, and live chart real data. The owner confirmed recent records are visible in the normal History/Graph view. The cumulative dropped-log count was five at 10:24 PDT; its timing is not established and remains an observation for follow-up. Evidence: `work/evidence/smi-a-post-forward-hpt2-initialization-20260923.json` and `work/evidence/smi-displaytest-handoff-reconciliation-20260923.md`.

This closes A's explicit HPT2 initialization and fresh logging/history view after the verified normal forward installation. Customer USB export/content validation, post-restart retention, return/refusal/rollback, fault/recovery and B remain open. The SD Qualification Full Run has not been performed or claimed. Earlier dated format-pending sections below are history.

## A normal DisplayTest-to-SMI forward transition verified — 2026-09-23

The owner completed the native DisplayTest FAT32 format and SD read/write, confirmed responsive Home touch after the affected sequence, and entered Prep for Ship/Lockout. Read-only A state confirmed FAT32 handoff PASS, unchanged valid CID `05000C61736466672200001C8901A34F` and USB ready. `IMX2` on-drive SMI 00-05-01 and DisplayTest 01-01 files were independently hash-verified on a Pi read-only mount; the drive returned to A.

The owner selected `7520553_SMI_00_05_01.imx` through Lockout > Install Software and reported Success. A read-only complete four-MiB internal-flash capture has SHA-256 `40560E9D046293BE345676BC97CD09E88E29D543DBEB02AD3B764EBF029D4E87`, exactly matching the accepted SMI Release candidate and controlled IMX main payload. Matched SMI RAM state has A assembly 1751136, same valid CID, representative-sector read valid, FAT32 needs Format 2, no initialization request and no update request. The owner confirmed 00-05-01 Home/About and the expected Format 2 warning. This closes the A forward installation gate, not the separate HPT2, return, refusal, rollback, fault, retention or B gates. Structured evidence: `work/evidence/smi-a-normal-fixture-install-20260923.json` and handoff ledger `work/evidence/smi-displaytest-handoff-reconciliation-20260923.md`.

The owner declared the card history expendable and continued the reviewed sequence. The operator has been instructed to use SMI's explicit two-stage Format 2 warning/service-471 flow and report its terminal result. No initialization pass is inferred before that result and a fresh CID/status capture. Older dated pause/missing-image/format-pending paragraphs below are historical.

## A transition execution checkpoint — 2026-09-23

The owner declared the retained HPT2 history expendable, reviewed the concrete sequence and said to continue. A FAT32 preparation and the normal SMI installation are authorized within that sequence, subject to their verification stops. The owner reports `IMX2` at A and can use A's touchscreen. SMI can read the J-Link state but has no approved touchscreen control path; the operator was given the native `System Test → SD Card Test → Format SD as FAT32?` step with its left confirmation.

Read-only A baseline at 09:48 PDT reconfirmed exact DisplayTest 01-01, assembly 1751136, serial 69664082, unchanged valid CID `05000C61736466672200001C8901A34F`, ready HPT2, cached valid sector approval, and idle update/format/SDQ request flags. Another read-only capture at 09:55 PDT showed card state 2 (HPT2), format request 0 and format result 0. Thus no format completion can be claimed. The SD read/write result field of 2 is pre-existing/stale; no such test was requested in this SMI execution. No install or power operation occurred. Raw captures: `work/evidence/generated/smi-displaytest-takeover-20260923-094826/` and `...-095550/`.

On reported format completion, capture fresh CID, protocol/card state, CID validity, representative-sector evidence and format status. Stop on CID change or failure. Then run only the affected FAT32 SD read/write–Home touch sequence, verify fresh FAT32 Lockout handoff, and check exact SMI IMX bytes on `IMX2` before the normal installation. The A/B release matrix remains open. `work/evidence/smi-displaytest-handoff-reconciliation-20260923.md` is the SMI evidence ledger; prior handoff and paused statements below remain history.

## Current handoff reconciliation — 2026-09-23 09:27 PDT release

Coordination is resumed from SMI; live takeover is verified, with card-data disposition and explicit hardware-action authority still pending. DisplayTest released A with no pending operation, running 01-01 on assembly 1751136 / probe 69664082 with retained HPT2 CID `05000C61736466672200001C8901A34F`. Its controller-generated IMX, source `da955dc92a1652eba247f07b32d154b4b7001e02`, is now controlled at SHA-256 `E11CEB483BA478F6A3D98F75DBF303EA09800165463EFE17880D5CDB2ACD78AB`; local hashing matched it and the accepted SMI IMX. The missing-image prerequisite is complete.

Credit existing SMI candidate/UI/export and resolved updater evidence, plus DisplayTest HPT2 startup, export, scoped navigation and Card Identity; do not replay them. Full SD Qualification was not run and is not required merely for this transition. The affected SD read/write–then-Home touch sequence and verified FAT32 production handoff remain open.

Live read-only takeover succeeded at 09:40–09:41 PDT after the earlier DNS failure. Serial-bound A / assembly 1751136 / probe 69664082 has exact DisplayTest 01-01 main-flash SHA-256 `ACA37CEBA6621701CAF3909D8712EB672872940ECC0E835DE690D6AD829D6D81`. HPT2 is ready, driver CID valid, raw CID unchanged, cached sector-zero approval valid, and captured update/format/SD-test/SDQ request flags idle; FAT32 handoff remains NONE. See `work/evidence/smi-displaytest-live-takeover-20260923.json`. No halt/go/reset/program/power/format command was issued; J-Link initialization reported a low-power wake. The smallest open SMI-owned test is A's ordinary DisplayTest 01-01 → accepted SMI 00-05-01 installation through verified FAT32 fixture handoff. The retained HPT2 card first needs a recorded preservation/archive or expendable-data decision and explicit erase consent; an IMX export does not preserve SD history. After authorization, minimal FAT32 preparation and the focused affected touch check can supply that prerequisite without Full Run. Programming, power operations and formatting are not authorized by this handoff or reconciliation.

Remaining exact-candidate A/B forward/return, refusal/renamed-file, representative rollback, fault/recovery, initialization, logging/history/export and retention gates remain open. Earlier dated pause/missing-image statements below are history superseded by this capsule; prior owner-abandoned commitments remain abandoned. Evidence and the concrete next-step sequence: `work/evidence/smi-displaytest-handoff-reconciliation-20260923.md`.

## North Star

Deliver dependable SMI oven firmware that keeps control, monitoring, and alarms operating through storage faults, and releases only behavior proven on both controller assemblies.

## Current Completion Horizon

Complete corrected SMI 00-05-01 under the shared ordinary/service updater policy, with controlled DisplayTest 01-01 return/FAT32 handoff and explicit HPT2 initialization. Done requires exact corrected immutable candidate identity, applicable both-assembly transition/refusal/fault/logging/export/retention evidence, disposition of known limits, and a separate release decision. A passing build or bounded A/B install is insufficient.

## Strategic Context

The field SD card is not practically replaceable, so admitted legacy downgrades must prepare it safely in place. The owner rejected a service-only downgrade on September 23; ordinary and service updates now share policy and UI. Current evidence qualifies the retained-FAT32 correction in a bounded matrix, while the earlier legacy Success mismatch remains unisolated. Full HPT2, SD-family, supply and unrelated abandoned work are not revived by this release review.

## Current Priorities

- Preserve corrected IMX `4E4321E5…` and its exact main flash `1ACC7915…`; the old immutable bundle is historical.
- Credit completed A/B work and target only the current warning/cancel, applicable negative/fault and unevidenced B portions.
- Obtain explicit authority for corrected source freeze and later hardware mutations; preserve unrelated dirty work.
- Resolve release evidence limits and the legacy Success observation before a separate production decision.
- Keep DisplayTest read-only and CVO/DVO delivery outside this lane.

## Deliberate Non-Goals

- On 2026-09-23 the owner abandoned the temperature-control recovery and RTD-interlock successor campaigns, standalone SD Qualifier and independent physical-media validator delivery, SD specimen/family approval and supply-change closeout, and Tool Shed defect tickets TKT-0035/TKT-0036. These are no longer local execution commitments. Existing evidence, accepted limitations, and unresolved technical risks remain recorded in `docs/project-work-disposition-20260923.md`; abandonment is not qualification or defect resolution.
- No unconfirmed SD-card conversion or erase, and no field SD-card swap as an update or rollback workflow.
- No claim that the update Title authenticates its publisher.
- No automatic CVO/DVO release, SD-family approval, ECO, or customer rollout from the SMI transition result.

## Decisions Needed

- Explicit disposition or focused investigation of legacy 00-04-11 Success-with-unchanged-flash.
- Evidence applicability/disposition for earlier candidate results, missing retained-FAT32 warning and unavailable legacy CID.
- Separate authority for scoped corrected-source freeze and concrete remaining hardware tests; no release approval is inferred.

Resume only the warning/cancel-only A check after a fresh exact-firmware/CID/idle baseline and operator coordination; separately authorize corrected-candidate freeze and remaining target qualification. No install, format, power, stage or commit is authorized by this checkpoint.

## Review Triggers

- A DisplayTest 01-01 or later candidate is qualified for safe HPT2 startup and FAT32 handoff.
- Controller A visual boot or storage evidence contradicts the read-only target capture.
- Either assembly produces contradictory transition evidence.
- The in-place FAT32 preparation or service-471 recovery path produces contradictory evidence.
- An exact candidate reaches release qualification.
- The SD qualification or supply-change scope changes.

## 2026-09-22 requirements and evidence update

Controller A accepted and staged the prior SMI 00-05-01 IMX, with exact QSPI main SHA-256 `7F49E615FF971D1C63E44CF94B8DCD9BF919DA7846E052408189997A58B78B8E`, but internal flash remained at SHA-256 `CFC9ECC6ABCCDDA5633FFF6AB8A6595ECAC21F6AC951ED35F0D51407BB8CE407`. The owner added a target-driven FAT32 rule and disabled-chart rule. The revised source passes focused checks and canonical Debug/Release builds; Release S-record SHA-256 is `4E7DFB2F7333202C71A736DF6692512030DB43508A1CCD0C556E7E8C1EB6B326`. It is not frozen or target-qualified. No controller or SD mutation occurred during this requirements detour.

## 2026-09-23 reconciliation

The 2026-09-22 internal-flash failure was diagnosed and corrected by watchdog, FCU command-lock, and ROM-cache changes. A controlled self update completed with diagnostic error `0` and exact four-MiB flash readback. Accepted candidate `00.05.01-1bf8afe6aeba-6476100e5ae5` has Release S-record SHA-256 `6476100E5AE5B5ED5EA38F5BBFDFC8AE6BC44513200678406F2711B0B91C1D9D`; Controller A exact readback SHA-256 is `40560E9D046293BE345676BC97CD09E88E29D543DBEB02AD3B764EBF029D4E87`. The owner accepted the Format 2 startup and disabled-chart UI behavior. Service 150 then generated the final SMI IMX with SHA-256 `AF8132D13837F6030CD73689E77833AA563C0C878C19126448E869DFEB337332`, valid CRCs, and an exact main-payload match. The remaining current prerequisites are the controlled DisplayTest 01-01 IMX and the paused A/B transition, refusal, rollback, and fault matrix. No release decision has been made.
