softlockup.c 6.58 KB
Newer Older
Ingo Molnar's avatar
Ingo Molnar committed
1 2 3
/*
 * Detect Soft Lockups
 *
4
 * started by Ingo Molnar, Copyright (C) 2005, 2006 Red Hat, Inc.
Ingo Molnar's avatar
Ingo Molnar committed
5 6 7 8 9 10
 *
 * 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>
11
#include <linux/nmi.h>
Ingo Molnar's avatar
Ingo Molnar committed
12 13
#include <linux/init.h>
#include <linux/delay.h>
14
#include <linux/freezer.h>
Ingo Molnar's avatar
Ingo Molnar committed
15
#include <linux/kthread.h>
Vegard Nossum's avatar
Vegard Nossum committed
16
#include <linux/lockdep.h>
Ingo Molnar's avatar
Ingo Molnar committed
17 18
#include <linux/notifier.h>
#include <linux/module.h>
19
#include <linux/sysctl.h>
Ingo Molnar's avatar
Ingo Molnar committed
20

Ingo Molnar's avatar
Ingo Molnar committed
21 22
#include <asm/irq_regs.h>

Ingo Molnar's avatar
Ingo Molnar committed
23 24
static DEFINE_SPINLOCK(print_lock);

25 26
static DEFINE_PER_CPU(unsigned long, touch_timestamp);
static DEFINE_PER_CPU(unsigned long, print_timestamp);
Ingo Molnar's avatar
Ingo Molnar committed
27 28
static DEFINE_PER_CPU(struct task_struct *, watchdog_task);

Ingo Molnar's avatar
Ingo Molnar committed
29
static int __read_mostly did_panic;
30
int __read_mostly softlockup_thresh = 60;
31

32 33 34 35 36 37 38 39 40 41 42 43 44 45 46
/*
 * Should we panic (and reboot, if panic_timeout= is set) when a
 * soft-lockup occurs:
 */
unsigned int __read_mostly softlockup_panic =
				CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC_VALUE;

static int __init softlockup_panic_setup(char *str)
{
	softlockup_panic = simple_strtoul(str, NULL, 0);

	return 1;
}
__setup("softlockup_panic=", softlockup_panic_setup);

47 48
static int
softlock_panic(struct notifier_block *this, unsigned long event, void *ptr)
Ingo Molnar's avatar
Ingo Molnar committed
49 50 51 52 53 54 55 56 57 58
{
	did_panic = 1;

	return NOTIFY_DONE;
}

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

59 60 61 62 63
/*
 * 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.
 */
64
static unsigned long get_timestamp(int this_cpu)
65
{
66
	return cpu_clock(this_cpu) >> 30LL;  /* 2^30 ~= 10^9 */
67 68
}

69
static void __touch_softlockup_watchdog(void)
Ingo Molnar's avatar
Ingo Molnar committed
70
{
71 72 73
	int this_cpu = raw_smp_processor_id();

	__raw_get_cpu_var(touch_timestamp) = get_timestamp(this_cpu);
Ingo Molnar's avatar
Ingo Molnar committed
74
}
75 76 77 78 79

void touch_softlockup_watchdog(void)
{
	__raw_get_cpu_var(touch_timestamp) = 0;
}
Ingo Molnar's avatar
Ingo Molnar committed
80 81
EXPORT_SYMBOL(touch_softlockup_watchdog);

82 83 84 85 86 87 88 89 90 91
void touch_all_softlockup_watchdogs(void)
{
	int cpu;

	/* Cause each CPU to re-update its timestamp rather than complain */
	for_each_online_cpu(cpu)
		per_cpu(touch_timestamp, cpu) = 0;
}
EXPORT_SYMBOL(touch_all_softlockup_watchdogs);

92 93 94 95 96 97 98 99
int proc_dosoftlockup_thresh(struct ctl_table *table, int write,
			     struct file *filp, void __user *buffer,
			     size_t *lenp, loff_t *ppos)
{
	touch_all_softlockup_watchdogs();
	return proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos);
}

Ingo Molnar's avatar
Ingo Molnar committed
100 101 102 103
/*
 * This callback runs from the timer interrupt, and checks
 * whether the watchdog thread has hung or not:
 */
104
void softlockup_tick(void)
Ingo Molnar's avatar
Ingo Molnar committed
105 106
{
	int this_cpu = smp_processor_id();
107
	unsigned long touch_timestamp = per_cpu(touch_timestamp, this_cpu);
108
	unsigned long print_timestamp;
Ingo Molnar's avatar
Ingo Molnar committed
109
	struct pt_regs *regs = get_irq_regs();
110
	unsigned long now;
Ingo Molnar's avatar
Ingo Molnar committed
111

112 113 114 115 116 117 118 119
	/* Is detection switched off? */
	if (!per_cpu(watchdog_task, this_cpu) || softlockup_thresh <= 0) {
		/* Be sure we don't false trigger if switched back on */
		if (touch_timestamp)
			per_cpu(touch_timestamp, this_cpu) = 0;
		return;
	}

120
	if (touch_timestamp == 0) {
121
		__touch_softlockup_watchdog();
122
		return;
123
	}
124 125 126 127

	print_timestamp = per_cpu(print_timestamp, this_cpu);

	/* report at most once a second */
128
	if (print_timestamp == touch_timestamp || did_panic)
Ingo Molnar's avatar
Ingo Molnar committed
129 130
		return;

131 132
	/* do not print during early bootup: */
	if (unlikely(system_state != SYSTEM_RUNNING)) {
133
		__touch_softlockup_watchdog();
Ingo Molnar's avatar
Ingo Molnar committed
134
		return;
135
	}
Ingo Molnar's avatar
Ingo Molnar committed
136

137
	now = get_timestamp(this_cpu);
138

139 140 141 142 143
	/*
	 * Wake up the high-prio watchdog task twice per
	 * threshold timespan.
	 */
	if (now > touch_timestamp + softlockup_thresh/2)
Peter Zijlstra's avatar
Peter Zijlstra committed
144 145
		wake_up_process(per_cpu(watchdog_task, this_cpu));

146
	/* Warn about unreasonable delays: */
147
	if (now <= (touch_timestamp + softlockup_thresh))
Ingo Molnar's avatar
Ingo Molnar committed
148
		return;
Ingo Molnar's avatar
Ingo Molnar committed
149

Ingo Molnar's avatar
Ingo Molnar committed
150 151 152
	per_cpu(print_timestamp, this_cpu) = touch_timestamp;

	spin_lock(&print_lock);
153 154
	printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %lus! [%s:%d]\n",
			this_cpu, now - touch_timestamp,
155
			current->comm, task_pid_nr(current));
156
	print_modules();
Vegard Nossum's avatar
Vegard Nossum committed
157
	print_irqtrace_events(current);
Ingo Molnar's avatar
Ingo Molnar committed
158 159 160
	if (regs)
		show_regs(regs);
	else
161
		dump_stack();
Ingo Molnar's avatar
Ingo Molnar committed
162
	spin_unlock(&print_lock);
163 164 165

	if (softlockup_panic)
		panic("softlockup: hung tasks");
Ingo Molnar's avatar
Ingo Molnar committed
166 167 168 169 170
}

/*
 * The watchdog thread - runs every second and touches the timestamp.
 */
171
static int watchdog(void *__bind_cpu)
Ingo Molnar's avatar
Ingo Molnar committed
172
{
Oleg Nesterov's avatar
Oleg Nesterov committed
173
	struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
Ingo Molnar's avatar
Ingo Molnar committed
174 175 176

	sched_setscheduler(current, SCHED_FIFO, &param);

177
	/* initialize timestamp */
178
	__touch_softlockup_watchdog();
179

180
	set_current_state(TASK_INTERRUPTIBLE);
Ingo Molnar's avatar
Ingo Molnar committed
181
	/*
182
	 * Run briefly once per second to reset the softlockup timestamp.
183
	 * If this gets delayed for more than 60 seconds then the
184
	 * debug-printout triggers in softlockup_tick().
Ingo Molnar's avatar
Ingo Molnar committed
185 186
	 */
	while (!kthread_should_stop()) {
187
		__touch_softlockup_watchdog();
Peter Zijlstra's avatar
Peter Zijlstra committed
188 189 190 191
		schedule();

		if (kthread_should_stop())
			break;
192

193
		set_current_state(TASK_INTERRUPTIBLE);
Ingo Molnar's avatar
Ingo Molnar committed
194
	}
195
	__set_current_state(TASK_RUNNING);
Ingo Molnar's avatar
Ingo Molnar committed
196 197 198 199 200 201 202

	return 0;
}

/*
 * Create/destroy watchdog threads as CPUs come and go:
 */
203
static int __cpuinit
Ingo Molnar's avatar
Ingo Molnar committed
204 205 206 207 208 209 210
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:
211
	case CPU_UP_PREPARE_FROZEN:
Ingo Molnar's avatar
Ingo Molnar committed
212 213 214
		BUG_ON(per_cpu(watchdog_task, hotcpu));
		p = kthread_create(watchdog, hcpu, "watchdog/%d", hotcpu);
		if (IS_ERR(p)) {
215
			printk(KERN_ERR "watchdog for %i failed\n", hotcpu);
Ingo Molnar's avatar
Ingo Molnar committed
216 217
			return NOTIFY_BAD;
		}
218 219
		per_cpu(touch_timestamp, hotcpu) = 0;
		per_cpu(watchdog_task, hotcpu) = p;
Ingo Molnar's avatar
Ingo Molnar committed
220
		kthread_bind(p, hotcpu);
221
		break;
Ingo Molnar's avatar
Ingo Molnar committed
222
	case CPU_ONLINE:
223
	case CPU_ONLINE_FROZEN:
Ingo Molnar's avatar
Ingo Molnar committed
224 225 226
		wake_up_process(per_cpu(watchdog_task, hotcpu));
		break;
#ifdef CONFIG_HOTPLUG_CPU
Peter Zijlstra's avatar
Peter Zijlstra committed
227 228 229 230 231 232
	case CPU_UP_CANCELED:
	case CPU_UP_CANCELED_FROZEN:
		if (!per_cpu(watchdog_task, hotcpu))
			break;
		/* Unbind so it can run.  Fall thru. */
		kthread_bind(per_cpu(watchdog_task, hotcpu),
233
			     cpumask_any(cpu_online_mask));
Ingo Molnar's avatar
Ingo Molnar committed
234
	case CPU_DEAD:
235
	case CPU_DEAD_FROZEN:
Ingo Molnar's avatar
Ingo Molnar committed
236 237 238 239 240
		p = per_cpu(watchdog_task, hotcpu);
		per_cpu(watchdog_task, hotcpu) = NULL;
		kthread_stop(p);
		break;
#endif /* CONFIG_HOTPLUG_CPU */
241
	}
Ingo Molnar's avatar
Ingo Molnar committed
242 243 244
	return NOTIFY_OK;
}

245
static struct notifier_block __cpuinitdata cpu_nfb = {
Ingo Molnar's avatar
Ingo Molnar committed
246 247 248
	.notifier_call = cpu_callback
};

249 250 251 252 253 254 255 256 257 258
static int __initdata nosoftlockup;

static int __init nosoftlockup_setup(char *str)
{
	nosoftlockup = 1;
	return 1;
}
__setup("nosoftlockup", nosoftlockup_setup);

static int __init spawn_softlockup_task(void)
Ingo Molnar's avatar
Ingo Molnar committed
259 260
{
	void *cpu = (void *)(long)smp_processor_id();
261
	int err;
Ingo Molnar's avatar
Ingo Molnar committed
262

263 264 265 266 267 268 269 270
	if (nosoftlockup)
		return 0;

	err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu);
	if (err == NOTIFY_BAD) {
		BUG();
		return 1;
	}
Ingo Molnar's avatar
Ingo Molnar committed
271 272 273
	cpu_callback(&cpu_nfb, CPU_ONLINE, cpu);
	register_cpu_notifier(&cpu_nfb);

274
	atomic_notifier_chain_register(&panic_notifier_list, &panic_block);
275 276

	return 0;
Ingo Molnar's avatar
Ingo Molnar committed
277
}
278
early_initcall(spawn_softlockup_task);