Diagnose limited capacity state initialization

[Deadbytes101]

Alpha.4 physical first boot on the Dell test node reports limited_capacity after the boot USB and firstboot UI are reached. Investigate this in a state-only branch without changing the Alpha.5 config engine commits.\n\nRequired evidence\n\n- capture the redacted state-status message and bounded state service logs\n- identify the exact failed state stage\n- preserve exact boot USB and layout verification\n- verify state reaches ready or grown across reboot\n- keep internal disks out of scope\n\nDependency\n\nAlpha.5 must treat limited_capacity as a negative state gate with zero config or service mutation. Positive Alpha.5 physical config validation remains blocked until this issue passes and the state fix is integrated into the stacked base.

[Deadbytes101]

Physical Alpha.5 diagnostics identified the exact limited_capacity trigger.

Observed on hardware

• rigos-state.service exited successfully but emitted limited_capacity
• the bounded command failure is lsblk rejecting the requested PARTN output column on Debian 12 util-linux 2.38.1
• the verified boot USB is /dev/sdb, transport usb, size 28.7 GB
• the expected four partitions and labels are present with PARTUUIDs 5249474f-01 through 5249474f-04
• ROOT_A is mounted read-only as the live medium
• the internal SATA disk is /dev/sda and was not selected as the boot USB or state target
• /run/rigos/state-status.json was not present when inspected after login, so status persistence also needs verification

Root cause

rigos-state-init requests PARTN from lsblk, but that column is unavailable on the deployed util-linux version. Discovery fails before layout validation and state growth.

Fix direction

Remove the PARTN dependency from the lsblk query and derive partition numbers from the kernel sysfs partition attribute using each device's MAJ:MIN. Add a regression test for Debian 12-compatible discovery and preserve exact USB, partition-table, and internal-disk safety checks.

[Deadbytes101]

Implemented the Debian 12 lsblk fix on head 013cd913af96087c10f0df1f1edb5aec4a990fec. The state service now uses an isolated compatibility executable that removes unsupported PARTN from the real lsblk request and restores partition numbers from /sys/dev/block/<major:minor>/partition. It does not parse device names. A regression test locks the service wiring. Verify, rebuild, reflash, and physically confirm state reaches grown before closing this issue.

[Deadbytes101]

Alpha.6 fresh physical boot regression reproduced on commit 8fc0a4b.

Observed boot truth

• /run/rigos/boot-device.json was missing
• verified PARTUUID symlink 5249474f-04 -> /dev/sdb4 existed
• /dev/sdb4 was ext4 with label RIGOS_STATE
• recovery access worked and configuration remained absent
• rigos-state-ready.service correctly blocked downstream services
• first retry of rigos-state.service failed at filesystem growth with resize2fs requiring e2fsck -f /dev/sdb4
• state status remained limited_capacity
• /var/lib/rigos was not mounted
• rigos-profile-apply, firstboot configuration, huge pages, and miner remained blocked
• internal SATA disk remained untouched

Failure chain

1. Initial state initialization partially mutated the state partition
2. The boot attestation file was not preserved at /run/rigos/boot-device.json
3. A later exact-device retry located /dev/sdb4 successfully
4. resize2fs refused to proceed until a forced filesystem check was run

Required fix

• preserve or regenerate authoritative boot-device attestation before state initialization
• when state partition growth is required, run a bounded offline e2fsck -pf or equivalent safe preflight before resize2fs
• if repair requires manual intervention, emit a distinct visible status rather than generic limited_capacity
• never mount or grow an unverified partition
• keep downstream state-ready blocking behavior
• add a physical/integration test for an ext4 seed filesystem that requires fsck before resize

Alpha.6 physical gate remains failed. Do not start Native Hardware Telemetry yet.