# Additional USB drive compatibility sampling — 2026-08-01

Status: complete
Type: evidence
Updated: 2026-08-01
Next Action: none; preserve the T00_8GB incident for standards-led architecture review
Parent: work/checklists/checklist-additional-usb-drive-compatibility-sampling.md
Canonical Truth: docs/firmware-update-validation-and-recovery.md

## Baseline

- Source baseline: `ca6f573a`; Release S-record SHA-256
  `6907F88FB91D5DC58EBF15A918C3993FDE3A076AFAE9DAEFB099899F55932581`.
- Controller B: J-Link `69660123`; SD CID
  `035344535330384780976F0085018507`, CID-valid and representative-sector-read
  valid at the campaign baseline and every controller capture.
- Raw controller captures are under
  `work/evidence/hpt-log-store-circular-mirror/controller-b/additional-usb-sampling-20260801/`.

## T01_16GB — PASS

- Identity on return to the Pi: Silicon Power 16G, serial
  `14007109030E600436A9C4EDB92`; FAT32 label `T01_16GB`; UUID `3A27-46DC`;
  unmounted before inspection.
- First insertion on Controller B mounted at high speed with acceptance mask
  `15`, both prepared WAV files visible, no enumeration recovery, and zero
  current system/application/USBX errors.
- WAV load completed with zero current errors.
- Image publication reached result `1`: 8,193 FileX writes, 8,388,864 bytes,
  controller progress `1000`, and successful final rename. The single raw
  FileX status `4` is the accepted first-publication `FX_NOT_FOUND` old-image
  probe; it did not increment controller application or USBX errors.
- Unmounted Pi `fsck.fat -n -v` was clean: final-sector access succeeded, both
  FATs were consistent, and 4 files used 1,073 clusters. The final image
  `7520553_SMI_00_05_00.imx` was 8,388,864 bytes with SHA-256
  `ed193752464f814f2b417e229dfb1188352836ee23a5b3bf5dfd83ceee8649bc`.
  Prepared `W00.wav` and `W01.wav` retained their baseline hashes.

## T02_8GB — PASS

- First insertion on Controller B mounted at high speed with acceptance mask
  `15`, both prepared WAV files visible, no enumeration recovery, and zero
  current system/application/USBX errors.
- WAV load completed with zero current errors.
- Image publication reached result `1`: 8,193 FileX writes, 8,388,864 bytes,
  controller progress `1000`, successful close/flush/rename, and zero current
  controller application or USBX errors. The recorded raw FileX status `4` is
  the accepted first-publication `FX_NOT_FOUND` old-image probe.
- Identity on return to the Pi: PNY USB 2.0 FD, serial `0410080000012019`;
  FAT32 label `T02_8GB`; UUID `3A36-AD64`; unmounted before inspection.
  `fsck.fat -n -v` was clean: final-sector access succeeded, both FATs were
  consistent, and 4 files used 2,143 clusters. The final image measured
  8,388,864 bytes with SHA-256
  `ed193752464f814f2b417e229dfb1188352836ee23a5b3bf5dfd83ceee8649bc`.
  Prepared `W00.wav` and `W01.wav` retained their baseline hashes.

## T03_32GB — PASS

- First insertion on Controller B mounted at high speed with acceptance mask
  `15`, both prepared WAV files visible, no enumeration recovery, and zero
  current system/application/USBX errors.
- WAV load completed with zero current errors.
- Image publication reached result `1`: 8,193 FileX writes, 8,388,864 bytes,
  controller progress `1000`, successful flush/rename, and zero current
  controller application or USBX errors. The recorded raw FileX status `4` is
  the accepted first-publication `FX_NOT_FOUND` old-image probe.
- Identity on return to the Pi: PNY USB 3.0 FD, serial `AA61142100013798`;
  FAT32 label `T03_32GB`; UUID `3A6A-B799`; unmounted before inspection.
  `fsck.fat -n -v` was clean: final-sector access succeeded, both FATs were
  consistent, and 4 files used 538 clusters. The final image measured
  8,388,864 bytes with SHA-256
  `ed193752464f814f2b417e229dfb1188352836ee23a5b3bf5dfd83ceee8649bc`.
  Prepared `W00.wav` and `W01.wav` retained their baseline hashes.

## Disposition

Three distinct specimens passed end to end (`T01_16GB`, `T02_8GB`, and
`T03_32GB`); `T00_8GB` remains a single preserved fail-closed timeout. The
sample confirms that the current normal interaction works across these three
additional USB Mass Storage specimens, but does not establish universal device
compatibility or justify any device-specific production behavior. No firmware
was changed during this campaign.

## T00_8GB — FAIL, preserved

See work/incidents/incident-t00-usb-image-save-transfer-timeout.md. The
controller failed closed on raw USBX `0x5C` during the ordinary 1 KiB write
stream, and the returned filesystem remained clean with no published image.
