16 #include <linux/cpu.h>
35  * The powerpc watchdog ensures that each CPU is able to service timers.
36  * The watchdog sets up a simple timer on each CPU to run once per timer
37  * period, and updates a per-cpu timestamp and a "pending" cpumask. This is
43  * The soft-NMI checker can detect lockups on the local CPU. When interrupts
50  * The soft-NMI checker will compare the heartbeat timestamp for this CPU
61  * CPU clears their pending bit in their heartbeat timer. When the bitmask
62  * becomes empty, the last CPU to clear its pending bit updates a global
65  * In the heartbeat timer, if any CPU notices that the global timestamp has
67  * means the CPU(s) with their bit still set in the pending mask have had
71  * watchdog to detect an unresponsive CPU and pull it out of its stuck
117 	int cpu = raw_smp_processor_id();  in wd_lockup_ipi()  local
120 	pr_emerg("CPU %d Hard LOCKUP\n", cpu);  in wd_lockup_ipi()
121 	pr_emerg("CPU %d TB:%lld, last heartbeat TB:%lld (%lldms ago)\n",  in wd_lockup_ipi()
122 		 cpu, tb, per_cpu(wd_timer_tb, cpu),  in wd_lockup_ipi()
123 		 tb_to_ns(tb - per_cpu(wd_timer_tb, cpu)) / 1000000);  in wd_lockup_ipi()
145 static void set_cpu_stuck(int cpu, u64 tb)  in set_cpu_stuck()  argument
147 	set_cpumask_stuck(cpumask_of(cpu), tb);  in set_cpu_stuck()
150 static void watchdog_smp_panic(int cpu, u64 tb)  in watchdog_smp_panic()  argument
159 	if (cpumask_test_cpu(cpu, &wd_smp_cpus_pending))  in watchdog_smp_panic()
164 	pr_emerg("CPU %d detected hard LOCKUP on other CPUs %*pbl\n",  in watchdog_smp_panic()
165 		 cpu, cpumask_pr_args(&wd_smp_cpus_pending));  in watchdog_smp_panic()
166 	pr_emerg("CPU %d TB:%lld, last SMP heartbeat TB:%lld (%lldms ago)\n",  in watchdog_smp_panic()
167 		 cpu, tb, wd_smp_last_reset_tb,  in watchdog_smp_panic()
176 			if (c == cpu)  in watchdog_smp_panic()
199 static void wd_smp_clear_cpu_pending(int cpu, u64 tb)  in wd_smp_clear_cpu_pending()  argument
201 	if (!cpumask_test_cpu(cpu, &wd_smp_cpus_pending)) {  in wd_smp_clear_cpu_pending()
202 		if (unlikely(cpumask_test_cpu(cpu, &wd_smp_cpus_stuck))) {  in wd_smp_clear_cpu_pending()
208 			pr_emerg("CPU %d became unstuck TB:%lld\n",  in wd_smp_clear_cpu_pending()
209 				 cpu, tb);  in wd_smp_clear_cpu_pending()
216 			cpumask_clear_cpu(cpu, &wd_smp_cpus_stuck);  in wd_smp_clear_cpu_pending()
221 	cpumask_clear_cpu(cpu, &wd_smp_cpus_pending);  in wd_smp_clear_cpu_pending()
236 static void watchdog_timer_interrupt(int cpu)  in watchdog_timer_interrupt()  argument
240 	per_cpu(wd_timer_tb, cpu) = tb;  in watchdog_timer_interrupt()
242 	wd_smp_clear_cpu_pending(cpu, tb);  in watchdog_timer_interrupt()
245 		watchdog_smp_panic(cpu, tb);  in watchdog_timer_interrupt()
251 	int cpu = raw_smp_processor_id();  in DEFINE_INTERRUPT_HANDLER_NMI()  local
257 	if (!cpumask_test_cpu(cpu, &wd_cpus_enabled))  in DEFINE_INTERRUPT_HANDLER_NMI()
263 	if (tb - per_cpu(wd_timer_tb, cpu) >= wd_panic_timeout_tb) {  in DEFINE_INTERRUPT_HANDLER_NMI()
265 		if (cpumask_test_cpu(cpu, &wd_smp_cpus_stuck)) {  in DEFINE_INTERRUPT_HANDLER_NMI()
269 		set_cpu_stuck(cpu, tb);  in DEFINE_INTERRUPT_HANDLER_NMI()
271 		pr_emerg("CPU %d self-detected hard LOCKUP @ %pS\n",  in DEFINE_INTERRUPT_HANDLER_NMI()
272 			 cpu, (void *)regs->nip);  in DEFINE_INTERRUPT_HANDLER_NMI()
273 		pr_emerg("CPU %d TB:%lld, last heartbeat TB:%lld (%lldms ago)\n",  in DEFINE_INTERRUPT_HANDLER_NMI()
274 			 cpu, tb, per_cpu(wd_timer_tb, cpu),  in DEFINE_INTERRUPT_HANDLER_NMI()
275 			 tb_to_ns(tb - per_cpu(wd_timer_tb, cpu)) / 1000000);  in DEFINE_INTERRUPT_HANDLER_NMI()
296 	int cpu = smp_processor_id();  in watchdog_timer_fn()  local
301 	if (!cpumask_test_cpu(cpu, &watchdog_cpumask))  in watchdog_timer_fn()
304 	watchdog_timer_interrupt(cpu);  in watchdog_timer_fn()
314 	int cpu = smp_processor_id();  in arch_touch_nmi_watchdog()  local
317 	if (tb - per_cpu(wd_timer_tb, cpu) >= ticks) {  in arch_touch_nmi_watchdog()
318 		per_cpu(wd_timer_tb, cpu) = tb;  in arch_touch_nmi_watchdog()
319 		wd_smp_clear_cpu_pending(cpu, tb);  in arch_touch_nmi_watchdog()
327 	int cpu = smp_processor_id();  in start_watchdog()  local
330 	if (cpumask_test_cpu(cpu, &wd_cpus_enabled)) {  in start_watchdog()
338 	if (!cpumask_test_cpu(cpu, &watchdog_cpumask))  in start_watchdog()
342 	cpumask_set_cpu(cpu, &wd_cpus_enabled);  in start_watchdog()
344 		cpumask_set_cpu(cpu, &wd_smp_cpus_pending);  in start_watchdog()
357 static int start_watchdog_on_cpu(unsigned int cpu)  in start_watchdog_on_cpu()  argument
359 	return smp_call_function_single(cpu, start_watchdog, NULL, true);  in start_watchdog_on_cpu()
365 	int cpu = smp_processor_id();  in stop_watchdog()  local
368 	if (!cpumask_test_cpu(cpu, &wd_cpus_enabled))  in stop_watchdog()
369 		return; /* Can happen in CPU unplug case */  in stop_watchdog()
374 	cpumask_clear_cpu(cpu, &wd_cpus_enabled);  in stop_watchdog()
377 	wd_smp_clear_cpu_pending(cpu, get_tb());  in stop_watchdog()
380 static int stop_watchdog_on_cpu(unsigned int cpu)  in stop_watchdog_on_cpu()  argument
382 	return smp_call_function_single(cpu, stop_watchdog, NULL, true);  in stop_watchdog_on_cpu()
398 	int cpu;  in watchdog_nmi_stop()  local
400 	for_each_cpu(cpu, &wd_cpus_enabled)  in watchdog_nmi_stop()
401 		stop_watchdog_on_cpu(cpu);  in watchdog_nmi_stop()
406 	int cpu;  in watchdog_nmi_start()  local
409 	for_each_cpu_and(cpu, cpu_online_mask, &watchdog_cpumask)  in watchdog_nmi_start()
410 		start_watchdog_on_cpu(cpu);  in watchdog_nmi_start()