L1 guide: watchdog and health

Watchdog - kernel/core/core_watchdog.h (18 declarations; numbered 5)

watchdog_register(name, timeout_ms, policy)  -> slot
watchdog_start(slot)
watchdog_feed(slot)
watchdog_stop(slot)
watchdog_set_timeout(slot, ms)

A registered slot is a policy watchdog: the task must feed it; timeout kills the task (reason logged) while the kernel lives. Hardware watchdogs (Intel TCO, ACPI WDAT) are layered underneath and reset the whole machine - the E2E suite proves both with real resets and second boots.

Health - kernel/core/core_health.h (8 declarations; numbered 5)

health_get_cpu(out)
health_get_mem(out)
health_get_tasks(out)
health_get_io(out)
health_export(path)          // valid JSON to /var/log/health.json

Thresholds raise alarms into the kernel log (CPU-LOCKED on a pinned core...); health, health watch, health export are the commands; the E2E asserts the alarm lines and JSON validity.

klog - kernel/lib/lib_klog.h (29 declarations; numbered 6)

klog_write(level, fmt, ...)
klog_set_level(lvl)
klog_set_level_enabled(lvl, on)
klog_flush()
klog_rotate()
klog_filter_set(min_level)

Levels: DEBUG/INFO/WARN/ERROR/FATAL; rotation (kernel.log -> kernel.log.1, threshold 4096) and filtering are real, E2E-verified behaviours.