felhom-agent/REPORT.md

REPORT — agent v0.152.0: the kernel lane (2026-10-07)

What: R-836, 09 §3 decision 172 — a new kernel boots ONCE through a flag on the ESP; a crash falls back to the old kernel by itself; a healthy boot (host health rule + the hub reached, 20 min) makes it the default; a booted-but-unhealthy one is reverted ONCE. Design: felhom.eu/documentation/architecture/11-os-updates.md §5.11.

• Wrapper configs/felhom-os-apply: layer kernel (stage, kernel-reboot, kernel-boot, kernel-good, kernel-revert, kernel-cancel, kernel-status; R20–R23). Bundle: /etc/grub.d/01_felhom_oneshot, /etc/grub.d/42_felhom_oneshot.
• Agent internal/osupdate/kernel.go: the night step (told nights only), KernelAfterBoot, KernelStepExecutor.
• Tests: KernelLane (27), KernelStepCannotLeaveTheBoxOff (3), TestKernel* (13); red-proofs felhom.eu/documentation/audits/kernel-lane-2026-10-07/A/redproof.txt.
• Released v0.152.0 (d03ab7f; binary 95ff4220…, bundle f0c2cec3…) + step bundle 0.152.0-step1 (0b71d32b…). Delivered by signed jobs to Tester 1, demo-hp, demo-felhom (binary → step bundle → bundle). Not vouched (the golden is behind; waiver to 2026-10-13).
• Proven on the Tester 1 box (felhom.eu/documentation/audits/kernel-lane-2026-10-07/E/RESULT.md): forced panic → fell_back by itself; held guest → one self-revert after 20 min; healthy → 7.0.14-22 the default.
• Open: the ring-0 night run (R-836 says where it stopped); R-898 (ring 0 stages the pending kernel, not exactly the told one); R-897 (post-reboot drive re-bind races the first backup).