A Linux kernel watchdog detects prolonged processor or kernel non-responsiveness. Its failure mode provides an important first clue during low-level production incident triage. section 8.2.7
Distinguish the two principal lockup signals:
- A soft lockup occurs when kernel-mode code loops for more than 20 seconds without allowing other tasks to run. section 8.2.7
- A hard lockup occurs when the CPU loops in kernel mode for more than 10 seconds without allowing other interrupts to run. section 8.2.7
- These distinctions depend on tasks, context switching, kernel mode, and interrupts rather than ordinary user-space application behavior. section 8.2.7
Reading watchdog evidence
A message such as “BUG: soft lockup - CPU#13 stuck for 21s!” identifies a soft lockup. Related /proc settings show whether soft or hard lockups trigger a panic and which watchdog threshold is configured.
