M1 — Linux 6.18 First Boot

Historical milestone record. For current platform status and work ordering, use CURRENT_PLATFORM_STATE.md and the Platform v1 roadmap.

Canonical milestone 2. Core objective achieved on the physical Y2: Linux 6.18 + initramfs + native PID1. The M1 milestone is closed for this core objective. Physical UART research #16 remains open outside the milestone; remaining diagnostic uncertainties are recorded separately rather than treated as failures of the core boot objective.

The latest owner report confirms 6.18.0-y2-m1, increasing BEAT and uptime, working sleep, proc/sysfs mounts, framebuffer diagnostics, watchdog stopped, CPU0 online and static D08 RAM visible. [Hardware result and remaining diagnostic uncertainties](../knowledge/m1-runtime-hardware-result.md). This supersedes the historical sleep -38 and hardware-pending status below. It does not certify unreported counters, run duration, expanded RAM or unrelated peripherals.

The historical foundation queue was planning only. No new code/build/flash or M2 execution occurred in this session. M1 core success does not require another build or retrospective full-suite run: today's change records hardware evidence, not a new release-candidate certification. The sections below preserve the earlier implementation/research chronology.

The owner explicitly changed execution for M1: the lead researches and implements directly, without delegation. Tiny sequential issues preserve reviewable checkpoints; none is state:luna-ready. All commits use the owner's configured identity and issues use the authenticated owner account. Platform work remains solely in Y2Linux.

Result, exact versions/hashes/map and reproduction commands are authoritative for this artifact. D08 limits are unchanged. Configuration resolution exposed a hidden upstream option; bounded research produced one Kconfig prompt-visibility patch (D09) rather than accepting a policy violation.

The hardware launch-gate review identifies the next smallest missing proofs. No speculative peripheral/rootfs/SMP backlog or next implementation wave is created. The current output is an unsigned offline candidate. Closing the eight implementation checkpoints does not close M1, pass M0, authorize a DA, raw acquisition, flash, boot or restore, or prove the physical console. The next wave must start from launch-gate evidence and a concrete separately authorized experiment.

Observation gate follow-up

Y2E-140 / #16 is open: installed ttyMT0/UART0 evidence prompted D10 and two fully validated reproducible builds, plus receive-only host tooling. The first-artifact hashes above now resolve to the current UART0 result; historical UART3 reports/bundle remain retained. Physical TX/GND/level and actual stock LK/kernel serial capture are still missing. No further kernel feature or device experiment was added.

Y2E-145 / #17 records the subsequent NO-GO decision for the unchanged candidate, hardware-dependent authentication, denied current raw acquisition and adverse normal-path watchdog evidence. Decision. M1 remains open; no first-boot operation issue exists.

Y2E-150 / #18 now supplies D11–D13 and the risk-accepted diagnostic result. Research 3aeeb03; implementation 1234f85; two final builds match and all 11 tests pass. The previous NO-GO hard gates are superseded by owner-accepted risk; watchdog and guarded visual diagnostics are implemented. Y2B-240 prepares one BOOTIMG-only experiment requiring separate hardware authorization. M1 remains open until actual boot evidence is reviewed.

Execution checkpoint: Y2B-240 / #19, awaiting separate owner hardware authorization.

Y2B-245 offline implementation is complete at source d83fccf. Two release builds match across 12 outputs and all 14 tests pass. [Final artifact and validation](../build/y2b245-result.md). New text-candidate hardware validation remains pending separate authorization; Linux-plus-PID1 boot is already achieved by Y2B-240 solid green.

Current hardware boundary: Y2B-250

The owner reports the real Y2B-245 renderer, PID1 and successful proc/sysfs mounts, followed by BEAT: 0 / STOP: SLEEP ERROR / -38. Evidence and root cause. Y2B-250 #21 enables the missing ARM time32 ABI. No increasing heartbeat has been established; no new flash or M2 is authorized.

Localized iterations now use one clean build, directly affected tests, D08/layout/BOOTIMG safety validation and final image size/hash. Trust the unchanged audited source and locked toolchain; do not repeat full provenance/recovery audits or two reproducibility builds. Reserve exhaustive verification for architecture/memory/packaging changes, new subsystems, milestone completion and release candidates.

Y2B-250 is built and validated offline: candidate and focused results. The real timer/heartbeat trial remains pending; no milestone-completion claim is added.

Post-core roadmap audit, 2026-09-09

Full coverage and remaining phase gates. M1 core remains achieved. Broader exact-artifact boot repeatability, timing accuracy and long-run qualification remain under Y2H-300 #28 rather than being silently inferred from the successful screen. Five later-phase epics are deferred; no new implementation or hardware test was performed.

Historical foundation research queue

The September 8–10 queue followed the initial M1 runtime result. Its order and bounded research goals are retained here after removing the superseded rolling "next five" task list. These are historical scopes; the current implementation, qualification gaps and next work are in the Platform v1 roadmap. The original detailed specifications remain in issues/.

The ordering put RAM headroom before shared power, local controls, display and removable storage. Removable SD was the smaller initial storage surface than internal eMMC; inherited pixels did not establish a display-driver contract. The later DEV-02 qualification records the narrow achieved scope. Source availability and external hardware success did not authorize RAM expansion or substitute for this unit's evidence.