softlockup.c 4.5 KB
Newer Older
I
Ingo Molnar 已提交
1 2 3
/*
 * Detect Soft Lockups
 *
4
 * started by Ingo Molnar, Copyright (C) 2005, 2006 Red Hat, Inc.
I
Ingo Molnar 已提交
5 6 7 8 9 10 11 12 13 14 15 16 17 18
 *
 * this code detects soft lockups: incidents in where on a CPU
 * the kernel does not reschedule for 10 seconds or more.
 */
#include <linux/mm.h>
#include <linux/cpu.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/kthread.h>
#include <linux/notifier.h>
#include <linux/module.h>

static DEFINE_SPINLOCK(print_lock);

19 20
static DEFINE_PER_CPU(unsigned long, touch_timestamp);
static DEFINE_PER_CPU(unsigned long, print_timestamp);
I
Ingo Molnar 已提交
21 22 23
static DEFINE_PER_CPU(struct task_struct *, watchdog_task);

static int did_panic = 0;
24 25 26

static int
softlock_panic(struct notifier_block *this, unsigned long event, void *ptr)
I
Ingo Molnar 已提交
27 28 29 30 31 32 33 34 35 36
{
	did_panic = 1;

	return NOTIFY_DONE;
}

static struct notifier_block panic_block = {
	.notifier_call = softlock_panic,
};

37 38 39 40 41 42 43 44 45 46
/*
 * Returns seconds, approximately.  We don't need nanosecond
 * resolution, and we don't need to waste time with a big divide when
 * 2^30ns == 1.074s.
 */
static unsigned long get_timestamp(void)
{
	return sched_clock() >> 30;  /* 2^30 ~= 10^9 */
}

I
Ingo Molnar 已提交
47 48
void touch_softlockup_watchdog(void)
{
49
	__raw_get_cpu_var(touch_timestamp) = get_timestamp();
I
Ingo Molnar 已提交
50 51 52 53 54 55 56
}
EXPORT_SYMBOL(touch_softlockup_watchdog);

/*
 * This callback runs from the timer interrupt, and checks
 * whether the watchdog thread has hung or not:
 */
57
void softlockup_tick(void)
I
Ingo Molnar 已提交
58 59
{
	int this_cpu = smp_processor_id();
60
	unsigned long touch_timestamp = per_cpu(touch_timestamp, this_cpu);
61 62
	unsigned long print_timestamp;
	unsigned long now;
I
Ingo Molnar 已提交
63

64 65 66 67 68 69 70 71
	/* watchdog task hasn't updated timestamp yet */
	if (touch_timestamp == 0)
		return;

	print_timestamp = per_cpu(print_timestamp, this_cpu);

	/* report at most once a second */
	if (print_timestamp < (touch_timestamp + 1) ||
72 73
		did_panic ||
			!per_cpu(watchdog_task, this_cpu))
I
Ingo Molnar 已提交
74 75
		return;

76 77 78
	/* do not print during early bootup: */
	if (unlikely(system_state != SYSTEM_RUNNING)) {
		touch_softlockup_watchdog();
I
Ingo Molnar 已提交
79
		return;
80
	}
I
Ingo Molnar 已提交
81

82 83
	now = get_timestamp();

84
	/* Wake up the high-prio watchdog task every second: */
85
	if (now > (touch_timestamp + 1))
86 87 88
		wake_up_process(per_cpu(watchdog_task, this_cpu));

	/* Warn about unreasonable 10+ seconds delays: */
89
	if (now > (touch_timestamp + 10)) {
90
		per_cpu(print_timestamp, this_cpu) = touch_timestamp;
I
Ingo Molnar 已提交
91 92 93 94

		spin_lock(&print_lock);
		printk(KERN_ERR "BUG: soft lockup detected on CPU#%d!\n",
			this_cpu);
95
		dump_stack();
I
Ingo Molnar 已提交
96 97 98 99 100 101 102 103 104
		spin_unlock(&print_lock);
	}
}

/*
 * The watchdog thread - runs every second and touches the timestamp.
 */
static int watchdog(void * __bind_cpu)
{
O
Oleg Nesterov 已提交
105
	struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
I
Ingo Molnar 已提交
106 107 108 109

	sched_setscheduler(current, SCHED_FIFO, &param);
	current->flags |= PF_NOFREEZE;

110 111 112
	/* initialize timestamp */
	touch_softlockup_watchdog();

I
Ingo Molnar 已提交
113
	/*
114 115 116
	 * Run briefly once per second to reset the softlockup timestamp.
	 * If this gets delayed for more than 10 seconds then the
	 * debug-printout triggers in softlockup_tick().
I
Ingo Molnar 已提交
117 118
	 */
	while (!kthread_should_stop()) {
119
		set_current_state(TASK_INTERRUPTIBLE);
I
Ingo Molnar 已提交
120
		touch_softlockup_watchdog();
121
		schedule();
I
Ingo Molnar 已提交
122 123 124 125 126 127 128 129
	}

	return 0;
}

/*
 * Create/destroy watchdog threads as CPUs come and go:
 */
130
static int __cpuinit
I
Ingo Molnar 已提交
131 132 133 134 135 136 137 138 139 140 141 142 143
cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu)
{
	int hotcpu = (unsigned long)hcpu;
	struct task_struct *p;

	switch (action) {
	case CPU_UP_PREPARE:
		BUG_ON(per_cpu(watchdog_task, hotcpu));
		p = kthread_create(watchdog, hcpu, "watchdog/%d", hotcpu);
		if (IS_ERR(p)) {
			printk("watchdog for %i failed\n", hotcpu);
			return NOTIFY_BAD;
		}
144
  		per_cpu(touch_timestamp, hotcpu) = 0;
I
Ingo Molnar 已提交
145 146 147 148 149 150 151 152
  		per_cpu(watchdog_task, hotcpu) = p;
		kthread_bind(p, hotcpu);
 		break;
	case CPU_ONLINE:
		wake_up_process(per_cpu(watchdog_task, hotcpu));
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_UP_CANCELED:
153 154
		if (!per_cpu(watchdog_task, hotcpu))
			break;
I
Ingo Molnar 已提交
155
		/* Unbind so it can run.  Fall thru. */
156 157
		kthread_bind(per_cpu(watchdog_task, hotcpu),
			     any_online_cpu(cpu_online_map));
I
Ingo Molnar 已提交
158 159 160 161 162 163 164 165 166 167
	case CPU_DEAD:
		p = per_cpu(watchdog_task, hotcpu);
		per_cpu(watchdog_task, hotcpu) = NULL;
		kthread_stop(p);
		break;
#endif /* CONFIG_HOTPLUG_CPU */
 	}
	return NOTIFY_OK;
}

168
static struct notifier_block __cpuinitdata cpu_nfb = {
I
Ingo Molnar 已提交
169 170 171 172 173 174
	.notifier_call = cpu_callback
};

__init void spawn_softlockup_task(void)
{
	void *cpu = (void *)(long)smp_processor_id();
175
	int err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu);
I
Ingo Molnar 已提交
176

177
	BUG_ON(err == NOTIFY_BAD);
I
Ingo Molnar 已提交
178 179 180
	cpu_callback(&cpu_nfb, CPU_ONLINE, cpu);
	register_cpu_notifier(&cpu_nfb);

181
	atomic_notifier_chain_register(&panic_notifier_list, &panic_block);
I
Ingo Molnar 已提交
182
}