Handoff for 2026-10-02 — documentation and repository catch-up

Where things stand

The device runs storage ceiling candidate 2:

• Kernel 6.18.0-y2linux-storage-ceiling-02, root 2025.02.18-platform-v1.12.
• Linux 4db6d6de, Reborn b92d312c.
• Package out/y2linux-storage-ceiling-02-candidate/: BOOTIMG 5c17f744…, Y2ROOT dc611011….

Quick device check on candidate 2, 2026-10-01, read-only 128 MiB direct reads:

Candidate 1 booted earlier the same day. eMMC was fine, but SD never initialised: the PWRAP gate refused the VMC/VMCH writes. Commit 61a353b fixed it.

1. Documentation to update

The candidate 2 section of Y2-STORAGE-CEILING.md still says SD at the ceiling is "not yet observed". Update:

• docs/validation/Y2-STORAGE-CEILING.md: add the candidate 2 device

  result above. SDR104 negotiates but is not stable on this card/board under load. The fallback ladder worked as designed. Restate the per-host answers: - eMMC HS200: observed working. - SD 1.8 V UHS: observed working. - SD SDR104: negotiates, but CRC errors under load. - DDR50: observed stable in a short read.

• docs/CURRENT_PLATFORM_STATE.md:
  • Candidate 2 is now the installed pair, not just "the candidate to flash".
  • Add the storage observations.
  • Decide whether this update also refreshes the "Latest physically observed

    identity" table, which still shows Fix02. A full physical qualification was not run, so record this as a quick storage check only.

• docs/knowledge/candidate-index.md: mark candidate 2 as

  owner-flashed and booted, with the storage quick-check result.

• docs/validation/Y2-HARDWARE-FINAL-CAPABILITIES.md (capability

  ledger): eMMC/SD rows still describe 50 MHz. Add the new ceilings and the observed state.

• tools/platform/capabilities.json: decide whether SD ceiling text

  should say SDR104 is unstable on the tested card. Qualification flags stay false.

• docs/architecture/platform-storage-v1.md: mention the

  y2_storage status record, the y2.emmc_mode/y2.sd_mode boot controls and the fallback ladder.

• Reborn docs:
  • Diagnostics → Storage now shows mode, card, tuning and fault lines (Reborn

    b92d312).

  • Check ../Y2Reborn/docs/CURRENT_REBORN_STATE.md and the platform boundary

    doc for the new storage.controllers[].mode field.

• Run python3 tools/development/check-docs.py (both repos) and update

  catalog counts for any new page.

2. Repository catch-up

Nothing has been pushed. Both main branches are ahead of origin/main:

• Review git log origin/main..main in both repos, then push once the

  docs above are committed. Pushing is the owner's decision; nothing was pushed automatically.

• Optional cleanup: many build worktrees under out/*-worktrees/ and

  prunable ones under /tmp/y2linux-* (git worktree prune removes the prunable ones). Keep out/storage-ceiling-worktrees/ until candidate 2 is accepted.

3. Open technical decision (not documentation)

SDR104 hits CRC errors under sustained reads. Every boot starts at SDR104 and falls back on the first load. Options:

• Try the stock 1.8 V pad drive. LK leaves the SD pads at 1/1/1; the

  driver's 1.8 V table is 4/4/4. Check whether the drive write took effect, or test stronger values.

• Extend tuning (data-path delay), then re-test.
• Or set the SD ceiling to DDR50/SDR50 (y2.sd_mode= or the default ladder)

  so it never starts with errors. Throughput is about the same, ~40–50 MB/s.

Known unrelated notes

• tests/test_baseline.py fails locally on a clean tree too. It predates the

  connectivity PMIC fields and is not in the production suite.

• test_hardware_final_contract, test_dev_*, test_bootimg need

  pyelftools and only run in the build container (tools/production/tests.sh).

• The Y2Linux repo forbids model attribution in commits. Reborn has no local

  git identity configured; commit with git -c user.name=Luca -c [email protected].