kernel watchdog
also: watchdog timer, NMI watchdog, hardware watchdog
A kernel subsystem that monitors system health and automatically reboots the machine if it becomes unresponsive or hangs, preventing indefinite frozen states.
The kernel watchdog is a safety mechanism that detects when the Linux kernel has stopped responding to normal operations. It works by requiring the system to periodically "pet" (reset) a timer; if the timer expires without being reset, the watchdog assumes the kernel has hung and forces an immediate reboot.
This is particularly useful in production servers and embedded systems where human intervention isn't available. Without a watchdog, a hung kernel could remain frozen indefinitely, requiring physical intervention or causing service downtime.
Common watchdog implementations include software-based watchdogs (like softlockup_detector) and hardware watchdogs (dedicated circuit boards). You can manage watchdog behavior through /dev/watchdog or kernel parameters like nmi_watchdog and soft_lockup_panic.