[net-next-2.6.git] / kernel / watchdog.c

/*
 * Detect hard and soft lockups on a system
 *
 * started by Don Zickus, Copyright (C) 2010 Red Hat, Inc.
 *
 * this code detects hard lockups: incidents in where on a CPU
 * the kernel does not respond to anything except NMI.
 *
 * Note: Most of this code is borrowed heavily from softlockup.c,
 * so thanks to Ingo for the initial implementation.
 * Some chunks also taken from arch/x86/kernel/apic/nmi.c, thanks
 * to those contributors as well.
 */

#include <linux/mm.h>
#include <linux/cpu.h>
#include <linux/nmi.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/freezer.h>
#include <linux/kthread.h>
#include <linux/lockdep.h>
#include <linux/notifier.h>
#include <linux/module.h>
#include <linux/sysctl.h>

#include <asm/irq_regs.h>
#include <linux/perf_event.h>

int watchdog_enabled;
int __read_mostly softlockup_thresh = 60;

static DEFINE_PER_CPU(unsigned long, watchdog_touch_ts);
static DEFINE_PER_CPU(struct task_struct *, softlockup_watchdog);
static DEFINE_PER_CPU(struct hrtimer, watchdog_hrtimer);
static DEFINE_PER_CPU(bool, softlockup_touch_sync);
static DEFINE_PER_CPU(bool, soft_watchdog_warn);
#ifdef CONFIG_HARDLOCKUP_DETECTOR
static DEFINE_PER_CPU(bool, hard_watchdog_warn);
static DEFINE_PER_CPU(bool, watchdog_nmi_touch);
static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts);
static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved);
static DEFINE_PER_CPU(struct perf_event *, watchdog_ev);
#endif

static int no_watchdog;


/* boot commands */
/*
 * Should we panic when a soft-lockup or hard-lockup occurs:
 */
#ifdef CONFIG_HARDLOCKUP_DETECTOR
static int hardlockup_panic;

static int __init hardlockup_panic_setup(char *str)
{
	if (!strncmp(str, "panic", 5))
		hardlockup_panic = 1;
	return 1;
}
__setup("nmi_watchdog=", hardlockup_panic_setup);
#endif

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);

static int __init nowatchdog_setup(char *str)
{
	no_watchdog = 1;
	return 1;
}
__setup("nowatchdog", nowatchdog_setup);

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


/*
 * 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(int this_cpu)
{
	return cpu_clock(this_cpu) >> 30LL;  /* 2^30 ~= 10^9 */
}

static unsigned long get_sample_period(void)
{
	/*
	 * convert softlockup_thresh from seconds to ns
	 * the divide by 5 is to give hrtimer 5 chances to
	 * increment before the hardlockup detector generates
	 * a warning
	 */
	return softlockup_thresh / 5 * NSEC_PER_SEC;
}

/* Commands for resetting the watchdog */
static void __touch_watchdog(void)
{
	int this_cpu = smp_processor_id();

	__get_cpu_var(watchdog_touch_ts) = get_timestamp(this_cpu);
}

void touch_softlockup_watchdog(void)
{
	__raw_get_cpu_var(watchdog_touch_ts) = 0;
}
EXPORT_SYMBOL(touch_softlockup_watchdog);

void touch_all_softlockup_watchdogs(void)
{
	int cpu;

	/*
	 * this is done lockless
	 * do we care if a 0 races with a timestamp?
	 * all it means is the softlock check starts one cycle later
	 */
	for_each_online_cpu(cpu)
		per_cpu(watchdog_touch_ts, cpu) = 0;
}

#ifdef CONFIG_HARDLOCKUP_DETECTOR
void touch_nmi_watchdog(void)
{
	if (watchdog_enabled) {
		unsigned cpu;

		for_each_present_cpu(cpu) {
			if (per_cpu(watchdog_nmi_touch, cpu) != true)
				per_cpu(watchdog_nmi_touch, cpu) = true;
		}
	}
	touch_softlockup_watchdog();
}
EXPORT_SYMBOL(touch_nmi_watchdog);

#endif

void touch_softlockup_watchdog_sync(void)
{
	__raw_get_cpu_var(softlockup_touch_sync) = true;
	__raw_get_cpu_var(watchdog_touch_ts) = 0;
}

#ifdef CONFIG_HARDLOCKUP_DETECTOR
/* watchdog detector functions */
static int is_hardlockup(void)
{
	unsigned long hrint = __get_cpu_var(hrtimer_interrupts);

	if (__get_cpu_var(hrtimer_interrupts_saved) == hrint)
		return 1;

	__get_cpu_var(hrtimer_interrupts_saved) = hrint;
	return 0;
}
#endif

static int is_softlockup(unsigned long touch_ts)
{
	unsigned long now = get_timestamp(smp_processor_id());

	/* Warn about unreasonable delays: */
	if (time_after(now, touch_ts + softlockup_thresh))
		return now - touch_ts;

	return 0;
}

#ifdef CONFIG_HARDLOCKUP_DETECTOR
static struct perf_event_attr wd_hw_attr = {
	.type		= PERF_TYPE_HARDWARE,
	.config		= PERF_COUNT_HW_CPU_CYCLES,
	.size		= sizeof(struct perf_event_attr),
	.pinned		= 1,
	.disabled	= 1,
};

/* Callback function for perf event subsystem */
static void watchdog_overflow_callback(struct perf_event *event, int nmi,
		 struct perf_sample_data *data,
		 struct pt_regs *regs)
{
	/* Ensure the watchdog never gets throttled */
	event->hw.interrupts = 0;

	if (__get_cpu_var(watchdog_nmi_touch) == true) {
		__get_cpu_var(watchdog_nmi_touch) = false;
		return;
	}

	/* check for a hardlockup
	 * This is done by making sure our timer interrupt
	 * is incrementing.  The timer interrupt should have
	 * fired multiple times before we overflow'd.  If it hasn't
	 * then this is a good indication the cpu is stuck
	 */
	if (is_hardlockup()) {
		int this_cpu = smp_processor_id();

		/* only print hardlockups once */
		if (__get_cpu_var(hard_watchdog_warn) == true)
			return;

		if (hardlockup_panic)
			panic("Watchdog detected hard LOCKUP on cpu %d", this_cpu);
		else
			WARN(1, "Watchdog detected hard LOCKUP on cpu %d", this_cpu);

		__get_cpu_var(hard_watchdog_warn) = true;
		return;
	}

	__get_cpu_var(hard_watchdog_warn) = false;
	return;
}
static void watchdog_interrupt_count(void)
{
	__get_cpu_var(hrtimer_interrupts)++;
}
#else
static inline void watchdog_interrupt_count(void) { return; }
#endif /* CONFIG_HARDLOCKUP_DETECTOR */

/* watchdog kicker functions */
static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
{
	unsigned long touch_ts = __get_cpu_var(watchdog_touch_ts);
	struct pt_regs *regs = get_irq_regs();
	int duration;

	/* kick the hardlockup detector */
	watchdog_interrupt_count();

	/* kick the softlockup detector */
	wake_up_process(__get_cpu_var(softlockup_watchdog));

	/* .. and repeat */
	hrtimer_forward_now(hrtimer, ns_to_ktime(get_sample_period()));

	if (touch_ts == 0) {
		if (unlikely(__get_cpu_var(softlockup_touch_sync))) {
			/*
			 * If the time stamp was touched atomically
			 * make sure the scheduler tick is up to date.
			 */
			__get_cpu_var(softlockup_touch_sync) = false;
			sched_clock_tick();
		}
		__touch_watchdog();
		return HRTIMER_RESTART;
	}

	/* check for a softlockup
	 * This is done by making sure a high priority task is
	 * being scheduled.  The task touches the watchdog to
	 * indicate it is getting cpu time.  If it hasn't then
	 * this is a good indication some task is hogging the cpu
	 */
	duration = is_softlockup(touch_ts);
	if (unlikely(duration)) {
		/* only warn once */
		if (__get_cpu_var(soft_watchdog_warn) == true)
			return HRTIMER_RESTART;

		printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
			smp_processor_id(), duration,
			current->comm, task_pid_nr(current));
		print_modules();
		print_irqtrace_events(current);
		if (regs)
			show_regs(regs);
		else
			dump_stack();

		if (softlockup_panic)
			panic("softlockup: hung tasks");
		__get_cpu_var(soft_watchdog_warn) = true;
	} else
		__get_cpu_var(soft_watchdog_warn) = false;

	return HRTIMER_RESTART;
}


/*
 * The watchdog thread - touches the timestamp.
 */
static int watchdog(void *unused)
{
	struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
	struct hrtimer *hrtimer = &__raw_get_cpu_var(watchdog_hrtimer);

	sched_setscheduler(current, SCHED_FIFO, &param);

	/* initialize timestamp */
	__touch_watchdog();

	/* kick off the timer for the hardlockup detector */
	/* done here because hrtimer_start can only pin to smp_processor_id() */
	hrtimer_start(hrtimer, ns_to_ktime(get_sample_period()),
		      HRTIMER_MODE_REL_PINNED);

	set_current_state(TASK_INTERRUPTIBLE);
	/*
	 * Run briefly once per second to reset the softlockup timestamp.
	 * If this gets delayed for more than 60 seconds then the
	 * debug-printout triggers in watchdog_timer_fn().
	 */
	while (!kthread_should_stop()) {
		__touch_watchdog();
		schedule();

		if (kthread_should_stop())
			break;

		set_current_state(TASK_INTERRUPTIBLE);
	}
	__set_current_state(TASK_RUNNING);

	return 0;
}


#ifdef CONFIG_HARDLOCKUP_DETECTOR
static int watchdog_nmi_enable(int cpu)
{
	struct perf_event_attr *wd_attr;
	struct perf_event *event = per_cpu(watchdog_ev, cpu);

	/* is it already setup and enabled? */
	if (event && event->state > PERF_EVENT_STATE_OFF)
		goto out;

	/* it is setup but not enabled */
	if (event != NULL)
		goto out_enable;

	/* Try to register using hardware perf events */
	wd_attr = &wd_hw_attr;
	wd_attr->sample_period = hw_nmi_get_sample_period();
	event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback);
	if (!IS_ERR(event)) {
		printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n");
		goto out_save;
	}

	printk(KERN_ERR "NMI watchdog failed to create perf event on cpu%i: %p\n", cpu, event);
	return PTR_ERR(event);

	/* success path */
out_save:
	per_cpu(watchdog_ev, cpu) = event;
out_enable:
	perf_event_enable(per_cpu(watchdog_ev, cpu));
out:
	return 0;
}

static void watchdog_nmi_disable(int cpu)
{
	struct perf_event *event = per_cpu(watchdog_ev, cpu);

	if (event) {
		perf_event_disable(event);
		per_cpu(watchdog_ev, cpu) = NULL;

		/* should be in cleanup, but blocks oprofile */
		perf_event_release_kernel(event);
	}
	return;
}
#else
static int watchdog_nmi_enable(int cpu) { return 0; }
static void watchdog_nmi_disable(int cpu) { return; }
#endif /* CONFIG_HARDLOCKUP_DETECTOR */

/* prepare/enable/disable routines */
static int watchdog_prepare_cpu(int cpu)
{
	struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, cpu);

	WARN_ON(per_cpu(softlockup_watchdog, cpu));
	hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	hrtimer->function = watchdog_timer_fn;

	return 0;
}

static int watchdog_enable(int cpu)
{
	struct task_struct *p = per_cpu(softlockup_watchdog, cpu);
	int err;

	/* enable the perf event */
	err = watchdog_nmi_enable(cpu);
	if (err)
		return err;

	/* create the watchdog thread */
	if (!p) {
		p = kthread_create(watchdog, (void *)(unsigned long)cpu, "watchdog/%d", cpu);
		if (IS_ERR(p)) {
			printk(KERN_ERR "softlockup watchdog for %i failed\n", cpu);
			return PTR_ERR(p);
		}
		kthread_bind(p, cpu);
		per_cpu(watchdog_touch_ts, cpu) = 0;
		per_cpu(softlockup_watchdog, cpu) = p;
		wake_up_process(p);
	}

	/* if any cpu succeeds, watchdog is considered enabled for the system */
	watchdog_enabled = 1;

	return 0;
}

static void watchdog_disable(int cpu)
{
	struct task_struct *p = per_cpu(softlockup_watchdog, cpu);
	struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, cpu);

	/*
	 * cancel the timer first to stop incrementing the stats
	 * and waking up the kthread
	 */
	hrtimer_cancel(hrtimer);

	/* disable the perf event */
	watchdog_nmi_disable(cpu);

	/* stop the watchdog thread */
	if (p) {
		per_cpu(softlockup_watchdog, cpu) = NULL;
		kthread_stop(p);
	}
}

static void watchdog_enable_all_cpus(void)
{
	int cpu;
	int result = 0;

	for_each_online_cpu(cpu)
		result += watchdog_enable(cpu);

	if (result)
		printk(KERN_ERR "watchdog: failed to be enabled on some cpus\n");

}

static void watchdog_disable_all_cpus(void)
{
	int cpu;

	if (no_watchdog)
		return;

	for_each_online_cpu(cpu)
		watchdog_disable(cpu);

	/* if all watchdogs are disabled, then they are disabled for the system */
	watchdog_enabled = 0;
}


/* sysctl functions */
#ifdef CONFIG_SYSCTL
/*
 * proc handler for /proc/sys/kernel/nmi_watchdog
 */

int proc_dowatchdog_enabled(struct ctl_table *table, int write,
		     void __user *buffer, size_t *length, loff_t *ppos)
{
	proc_dointvec(table, write, buffer, length, ppos);

	if (watchdog_enabled)
		watchdog_enable_all_cpus();
	else
		watchdog_disable_all_cpus();
	return 0;
}

int proc_dowatchdog_thresh(struct ctl_table *table, int write,
			     void __user *buffer,
			     size_t *lenp, loff_t *ppos)
{
	return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
}
#endif /* CONFIG_SYSCTL */


/*
 * Create/destroy watchdog threads as CPUs come and go:
 */
static int __cpuinit
cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu)
{
	int hotcpu = (unsigned long)hcpu;
	int err = 0;

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
		err = watchdog_prepare_cpu(hotcpu);
		break;
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
		err = watchdog_enable(hotcpu);
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_UP_CANCELED:
	case CPU_UP_CANCELED_FROZEN:
		watchdog_disable(hotcpu);
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		watchdog_disable(hotcpu);
		break;
#endif /* CONFIG_HOTPLUG_CPU */
	}
	return notifier_from_errno(err);
}

static struct notifier_block __cpuinitdata cpu_nfb = {
	.notifier_call = cpu_callback
};

static int __init spawn_watchdog_task(void)
{
	void *cpu = (void *)(long)smp_processor_id();
	int err;

	if (no_watchdog)
		return 0;

	err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu);
	WARN_ON(notifier_to_errno(err));

	cpu_callback(&cpu_nfb, CPU_ONLINE, cpu);
	register_cpu_notifier(&cpu_nfb);

	return 0;
}
early_initcall(spawn_watchdog_task);
Commit	Line	Data
58687acb DZ	1	/*
	2	* Detect hard and soft lockups on a system
	3	*
	4	* started by Don Zickus, Copyright (C) 2010 Red Hat, Inc.
	5	*
	6	* this code detects hard lockups: incidents in where on a CPU
	7	* the kernel does not respond to anything except NMI.
	8	*
	9	* Note: Most of this code is borrowed heavily from softlockup.c,
	10	* so thanks to Ingo for the initial implementation.
	11	* Some chunks also taken from arch/x86/kernel/apic/nmi.c, thanks
	12	* to those contributors as well.
	13	*/
	14
	15	#include <linux/mm.h>
	16	#include <linux/cpu.h>
	17	#include <linux/nmi.h>
	18	#include <linux/init.h>
	19	#include <linux/delay.h>
	20	#include <linux/freezer.h>
	21	#include <linux/kthread.h>
	22	#include <linux/lockdep.h>
	23	#include <linux/notifier.h>
	24	#include <linux/module.h>
	25	#include <linux/sysctl.h>
	26
	27	#include <asm/irq_regs.h>
	28	#include <linux/perf_event.h>
	29
	30	int watchdog_enabled;
	31	int __read_mostly softlockup_thresh = 60;
	32
	33	static DEFINE_PER_CPU(unsigned long, watchdog_touch_ts);
	34	static DEFINE_PER_CPU(struct task_struct *, softlockup_watchdog);
	35	static DEFINE_PER_CPU(struct hrtimer, watchdog_hrtimer);
	36	static DEFINE_PER_CPU(bool, softlockup_touch_sync);
58687acb	37	static DEFINE_PER_CPU(bool, soft_watchdog_warn);
23637d47	38	#ifdef CONFIG_HARDLOCKUP_DETECTOR
cafcd80d DZ	39	static DEFINE_PER_CPU(bool, hard_watchdog_warn);
cafcd80d DZ	40	static DEFINE_PER_CPU(bool, watchdog_nmi_touch);
58687acb DZ	41	static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts);
	42	static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved);
	43	static DEFINE_PER_CPU(struct perf_event *, watchdog_ev);
	44	#endif
	45
433039e9	46	static int no_watchdog;
58687acb DZ	47
	48
	49	/* boot commands */
	50	/*
	51	* Should we panic when a soft-lockup or hard-lockup occurs:
	52	*/
23637d47	53	#ifdef CONFIG_HARDLOCKUP_DETECTOR
58687acb DZ	54	static int hardlockup_panic;
	55
	56	static int __init hardlockup_panic_setup(char *str)
	57	{
	58	if (!strncmp(str, "panic", 5))
	59	hardlockup_panic = 1;
	60	return 1;
	61	}
	62	__setup("nmi_watchdog=", hardlockup_panic_setup);
	63	#endif
	64
	65	unsigned int __read_mostly softlockup_panic =
	66	CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC_VALUE;
	67
	68	static int __init softlockup_panic_setup(char *str)
	69	{
	70	softlockup_panic = simple_strtoul(str, NULL, 0);
	71
	72	return 1;
	73	}
	74	__setup("softlockup_panic=", softlockup_panic_setup);
	75
	76	static int __init nowatchdog_setup(char *str)
	77	{
	78	no_watchdog = 1;
	79	return 1;
	80	}
	81	__setup("nowatchdog", nowatchdog_setup);
	82
	83	/* deprecated */
	84	static int __init nosoftlockup_setup(char *str)
	85	{
	86	no_watchdog = 1;
	87	return 1;
	88	}
	89	__setup("nosoftlockup", nosoftlockup_setup);
	90	/* */
	91
	92
	93	/*
	94	* Returns seconds, approximately. We don't need nanosecond
	95	* resolution, and we don't need to waste time with a big divide when
	96	* 2^30ns == 1.074s.
	97	*/
	98	static unsigned long get_timestamp(int this_cpu)
	99	{
	100	return cpu_clock(this_cpu) >> 30LL; /* 2^30 ~= 10^9 */
	101	}
	102
	103	static unsigned long get_sample_period(void)
	104	{
	105	/*
	106	* convert softlockup_thresh from seconds to ns
	107	* the divide by 5 is to give hrtimer 5 chances to
	108	* increment before the hardlockup detector generates
	109	* a warning
	110	*/
	111	return softlockup_thresh / 5 * NSEC_PER_SEC;
	112	}
	113
	114	/* Commands for resetting the watchdog */
	115	static void __touch_watchdog(void)
	116	{
26e09c6e	117	int this_cpu = smp_processor_id();
58687acb DZ	118
	119	__get_cpu_var(watchdog_touch_ts) = get_timestamp(this_cpu);
	120	}
	121
332fbdbc	122	void touch_softlockup_watchdog(void)
58687acb	123	{
68d3f1d8	124	__raw_get_cpu_var(watchdog_touch_ts) = 0;
58687acb	125	}
0167c781	126	EXPORT_SYMBOL(touch_softlockup_watchdog);
58687acb	127
332fbdbc	128	void touch_all_softlockup_watchdogs(void)
58687acb DZ	129	{
	130	int cpu;
	131
	132	/*
	133	* this is done lockless
	134	* do we care if a 0 races with a timestamp?
	135	* all it means is the softlock check starts one cycle later
	136	*/
	137	for_each_online_cpu(cpu)
	138	per_cpu(watchdog_touch_ts, cpu) = 0;
	139	}
	140
cafcd80d	141	#ifdef CONFIG_HARDLOCKUP_DETECTOR
58687acb DZ	142	void touch_nmi_watchdog(void)
58687acb DZ	143	{
68d3f1d8 DZ	144	if (watchdog_enabled) {
	145	unsigned cpu;
	146
	147	for_each_present_cpu(cpu) {
	148	if (per_cpu(watchdog_nmi_touch, cpu) != true)
	149	per_cpu(watchdog_nmi_touch, cpu) = true;
	150	}
	151	}
332fbdbc	152	touch_softlockup_watchdog();
58687acb DZ	153	}
	154	EXPORT_SYMBOL(touch_nmi_watchdog);
	155
cafcd80d DZ	156	#endif
cafcd80d DZ	157
58687acb DZ	158	void touch_softlockup_watchdog_sync(void)
	159	{
	160	__raw_get_cpu_var(softlockup_touch_sync) = true;
	161	__raw_get_cpu_var(watchdog_touch_ts) = 0;
	162	}
	163
23637d47	164	#ifdef CONFIG_HARDLOCKUP_DETECTOR
58687acb	165	/* watchdog detector functions */
26e09c6e	166	static int is_hardlockup(void)
58687acb	167	{
26e09c6e	168	unsigned long hrint = __get_cpu_var(hrtimer_interrupts);
58687acb	169
26e09c6e	170	if (__get_cpu_var(hrtimer_interrupts_saved) == hrint)
58687acb DZ	171	return 1;
58687acb DZ	172
26e09c6e	173	__get_cpu_var(hrtimer_interrupts_saved) = hrint;
58687acb DZ	174	return 0;
	175	}
	176	#endif
	177
26e09c6e	178	static int is_softlockup(unsigned long touch_ts)
58687acb	179	{
26e09c6e	180	unsigned long now = get_timestamp(smp_processor_id());
58687acb DZ	181
	182	/* Warn about unreasonable delays: */
	183	if (time_after(now, touch_ts + softlockup_thresh))
	184	return now - touch_ts;
	185
	186	return 0;
	187	}
	188
23637d47	189	#ifdef CONFIG_HARDLOCKUP_DETECTOR
58687acb DZ	190	static struct perf_event_attr wd_hw_attr = {
	191	.type = PERF_TYPE_HARDWARE,
	192	.config = PERF_COUNT_HW_CPU_CYCLES,
	193	.size = sizeof(struct perf_event_attr),
	194	.pinned = 1,
	195	.disabled = 1,
	196	};
	197
	198	/* Callback function for perf event subsystem */
277b1998	199	static void watchdog_overflow_callback(struct perf_event *event, int nmi,
58687acb DZ	200	struct perf_sample_data *data,
	201	struct pt_regs *regs)
	202	{
c6db67cd PZ	203	/* Ensure the watchdog never gets throttled */
	204	event->hw.interrupts = 0;
	205
d7c54733 DZ	206	if (__get_cpu_var(watchdog_nmi_touch) == true) {
d7c54733 DZ	207	__get_cpu_var(watchdog_nmi_touch) = false;
58687acb DZ	208	return;
	209	}
	210
	211	/* check for a hardlockup
	212	* This is done by making sure our timer interrupt
	213	* is incrementing. The timer interrupt should have
	214	* fired multiple times before we overflow'd. If it hasn't
	215	* then this is a good indication the cpu is stuck
	216	*/
26e09c6e DZ	217	if (is_hardlockup()) {
	218	int this_cpu = smp_processor_id();
	219
58687acb DZ	220	/* only print hardlockups once */
	221	if (__get_cpu_var(hard_watchdog_warn) == true)
	222	return;
	223
	224	if (hardlockup_panic)
	225	panic("Watchdog detected hard LOCKUP on cpu %d", this_cpu);
	226	else
	227	WARN(1, "Watchdog detected hard LOCKUP on cpu %d", this_cpu);
	228
	229	__get_cpu_var(hard_watchdog_warn) = true;
	230	return;
	231	}
	232
	233	__get_cpu_var(hard_watchdog_warn) = false;
	234	return;
	235	}
	236	static void watchdog_interrupt_count(void)
	237	{
	238	__get_cpu_var(hrtimer_interrupts)++;
	239	}
	240	#else
	241	static inline void watchdog_interrupt_count(void) { return; }
23637d47	242	#endif /* CONFIG_HARDLOCKUP_DETECTOR */
58687acb DZ	243
	244	/* watchdog kicker functions */
	245	static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
	246	{
58687acb DZ	247	unsigned long touch_ts = __get_cpu_var(watchdog_touch_ts);
	248	struct pt_regs *regs = get_irq_regs();
	249	int duration;
	250
	251	/* kick the hardlockup detector */
	252	watchdog_interrupt_count();
	253
	254	/* kick the softlockup detector */
	255	wake_up_process(__get_cpu_var(softlockup_watchdog));
	256
	257	/* .. and repeat */
	258	hrtimer_forward_now(hrtimer, ns_to_ktime(get_sample_period()));
	259
	260	if (touch_ts == 0) {
26e09c6e	261	if (unlikely(__get_cpu_var(softlockup_touch_sync))) {
58687acb DZ	262	/*
	263	* If the time stamp was touched atomically
	264	* make sure the scheduler tick is up to date.
	265	*/
26e09c6e	266	__get_cpu_var(softlockup_touch_sync) = false;
58687acb DZ	267	sched_clock_tick();
	268	}
	269	__touch_watchdog();
	270	return HRTIMER_RESTART;
	271	}
	272
	273	/* check for a softlockup
	274	* This is done by making sure a high priority task is
	275	* being scheduled. The task touches the watchdog to
	276	* indicate it is getting cpu time. If it hasn't then
	277	* this is a good indication some task is hogging the cpu
	278	*/
26e09c6e	279	duration = is_softlockup(touch_ts);
58687acb DZ	280	if (unlikely(duration)) {
	281	/* only warn once */
	282	if (__get_cpu_var(soft_watchdog_warn) == true)
	283	return HRTIMER_RESTART;
	284
	285	printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
26e09c6e	286	smp_processor_id(), duration,
58687acb DZ	287	current->comm, task_pid_nr(current));
	288	print_modules();
	289	print_irqtrace_events(current);
	290	if (regs)
	291	show_regs(regs);
	292	else
	293	dump_stack();
	294
	295	if (softlockup_panic)
	296	panic("softlockup: hung tasks");
	297	__get_cpu_var(soft_watchdog_warn) = true;
	298	} else
	299	__get_cpu_var(soft_watchdog_warn) = false;
	300
	301	return HRTIMER_RESTART;
	302	}
	303
	304
	305	/*
	306	* The watchdog thread - touches the timestamp.
	307	*/
26e09c6e	308	static int watchdog(void *unused)
58687acb DZ	309	{
58687acb DZ	310	struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
26e09c6e	311	struct hrtimer *hrtimer = &__raw_get_cpu_var(watchdog_hrtimer);
58687acb DZ	312
	313	sched_setscheduler(current, SCHED_FIFO, &param);
	314
	315	/* initialize timestamp */
	316	__touch_watchdog();
	317
	318	/* kick off the timer for the hardlockup detector */
	319	/* done here because hrtimer_start can only pin to smp_processor_id() */
	320	hrtimer_start(hrtimer, ns_to_ktime(get_sample_period()),
	321	HRTIMER_MODE_REL_PINNED);
	322
	323	set_current_state(TASK_INTERRUPTIBLE);
	324	/*
	325	* Run briefly once per second to reset the softlockup timestamp.
	326	* If this gets delayed for more than 60 seconds then the
26e09c6e	327	* debug-printout triggers in watchdog_timer_fn().
58687acb DZ	328	*/
	329	while (!kthread_should_stop()) {
	330	__touch_watchdog();
	331	schedule();
	332
	333	if (kthread_should_stop())
	334	break;
	335
	336	set_current_state(TASK_INTERRUPTIBLE);
	337	}
	338	__set_current_state(TASK_RUNNING);
	339
	340	return 0;
	341	}
	342
	343
23637d47	344	#ifdef CONFIG_HARDLOCKUP_DETECTOR
58687acb DZ	345	static int watchdog_nmi_enable(int cpu)
	346	{
	347	struct perf_event_attr *wd_attr;
	348	struct perf_event *event = per_cpu(watchdog_ev, cpu);
	349
	350	/* is it already setup and enabled? */
	351	if (event && event->state > PERF_EVENT_STATE_OFF)
	352	goto out;
	353
	354	/* it is setup but not enabled */
	355	if (event != NULL)
	356	goto out_enable;
	357
	358	/* Try to register using hardware perf events */
	359	wd_attr = &wd_hw_attr;
	360	wd_attr->sample_period = hw_nmi_get_sample_period();
38a81da2	361	event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback);
58687acb DZ	362	if (!IS_ERR(event)) {
	363	printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n");
	364	goto out_save;
	365	}
	366
	367	printk(KERN_ERR "NMI watchdog failed to create perf event on cpu%i: %p\n", cpu, event);
eac24335	368	return PTR_ERR(event);
58687acb DZ	369
	370	/* success path */
	371	out_save:
	372	per_cpu(watchdog_ev, cpu) = event;
	373	out_enable:
	374	perf_event_enable(per_cpu(watchdog_ev, cpu));
	375	out:
	376	return 0;
	377	}
	378
	379	static void watchdog_nmi_disable(int cpu)
	380	{
	381	struct perf_event *event = per_cpu(watchdog_ev, cpu);
	382
	383	if (event) {
	384	perf_event_disable(event);
	385	per_cpu(watchdog_ev, cpu) = NULL;
	386
	387	/* should be in cleanup, but blocks oprofile */
	388	perf_event_release_kernel(event);
	389	}
	390	return;
	391	}
	392	#else
	393	static int watchdog_nmi_enable(int cpu) { return 0; }
	394	static void watchdog_nmi_disable(int cpu) { return; }
23637d47	395	#endif /* CONFIG_HARDLOCKUP_DETECTOR */
58687acb DZ	396
	397	/* prepare/enable/disable routines */
	398	static int watchdog_prepare_cpu(int cpu)
	399	{
	400	struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, cpu);
	401
	402	WARN_ON(per_cpu(softlockup_watchdog, cpu));
	403	hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	404	hrtimer->function = watchdog_timer_fn;
	405
	406	return 0;
	407	}
	408
	409	static int watchdog_enable(int cpu)
	410	{
	411	struct task_struct *p = per_cpu(softlockup_watchdog, cpu);
eac24335	412	int err;
58687acb DZ	413
58687acb DZ	414	/* enable the perf event */
eac24335 AM	415	err = watchdog_nmi_enable(cpu);
	416	if (err)
	417	return err;
58687acb DZ	418
	419	/* create the watchdog thread */
	420	if (!p) {
	421	p = kthread_create(watchdog, (void *)(unsigned long)cpu, "watchdog/%d", cpu);
	422	if (IS_ERR(p)) {
	423	printk(KERN_ERR "softlockup watchdog for %i failed\n", cpu);
eac24335	424	return PTR_ERR(p);
58687acb DZ	425	}
	426	kthread_bind(p, cpu);
	427	per_cpu(watchdog_touch_ts, cpu) = 0;
	428	per_cpu(softlockup_watchdog, cpu) = p;
	429	wake_up_process(p);
	430	}
	431
68d3f1d8 DZ	432	/* if any cpu succeeds, watchdog is considered enabled for the system */
	433	watchdog_enabled = 1;
	434
58687acb DZ	435	return 0;
	436	}
	437
	438	static void watchdog_disable(int cpu)
	439	{
	440	struct task_struct *p = per_cpu(softlockup_watchdog, cpu);
	441	struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, cpu);
	442
	443	/*
	444	* cancel the timer first to stop incrementing the stats
	445	* and waking up the kthread
	446	*/
	447	hrtimer_cancel(hrtimer);
	448
	449	/* disable the perf event */
	450	watchdog_nmi_disable(cpu);
	451
	452	/* stop the watchdog thread */
	453	if (p) {
	454	per_cpu(softlockup_watchdog, cpu) = NULL;
	455	kthread_stop(p);
	456	}
58687acb DZ	457	}
	458
	459	static void watchdog_enable_all_cpus(void)
	460	{
	461	int cpu;
eb703f98	462	int result = 0;
58687acb DZ	463
	464	for_each_online_cpu(cpu)
	465	result += watchdog_enable(cpu);
	466
	467	if (result)
	468	printk(KERN_ERR "watchdog: failed to be enabled on some cpus\n");
	469
	470	}
	471
	472	static void watchdog_disable_all_cpus(void)
	473	{
	474	int cpu;
	475
d9ca07a0 SE	476	if (no_watchdog)
	477	return;
	478
58687acb DZ	479	for_each_online_cpu(cpu)
	480	watchdog_disable(cpu);
	481
	482	/* if all watchdogs are disabled, then they are disabled for the system */
	483	watchdog_enabled = 0;
	484	}
	485
	486
	487	/* sysctl functions */
	488	#ifdef CONFIG_SYSCTL
	489	/*
	490	* proc handler for /proc/sys/kernel/nmi_watchdog
	491	*/
	492
	493	int proc_dowatchdog_enabled(struct ctl_table *table, int write,
	494	void __user buffer, size_t length, loff_t *ppos)
	495	{
	496	proc_dointvec(table, write, buffer, length, ppos);
	497
	498	if (watchdog_enabled)
	499	watchdog_enable_all_cpus();
	500	else
	501	watchdog_disable_all_cpus();
	502	return 0;
	503	}
	504
	505	int proc_dowatchdog_thresh(struct ctl_table *table, int write,
	506	void __user *buffer,
	507	size_t lenp, loff_t ppos)
	508	{
	509	return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
	510	}
58687acb DZ	511	#endif /* CONFIG_SYSCTL */
	512
	513
	514	/*
	515	* Create/destroy watchdog threads as CPUs come and go:
	516	*/
	517	static int __cpuinit
	518	cpu_callback(struct notifier_block nfb, unsigned long action, void hcpu)
	519	{
	520	int hotcpu = (unsigned long)hcpu;
eac24335	521	int err = 0;
58687acb DZ	522
	523	switch (action) {
	524	case CPU_UP_PREPARE:
	525	case CPU_UP_PREPARE_FROZEN:
eac24335	526	err = watchdog_prepare_cpu(hotcpu);
58687acb DZ	527	break;
	528	case CPU_ONLINE:
	529	case CPU_ONLINE_FROZEN:
eac24335	530	err = watchdog_enable(hotcpu);
58687acb DZ	531	break;
	532	#ifdef CONFIG_HOTPLUG_CPU
	533	case CPU_UP_CANCELED:
	534	case CPU_UP_CANCELED_FROZEN:
	535	watchdog_disable(hotcpu);
	536	break;
	537	case CPU_DEAD:
	538	case CPU_DEAD_FROZEN:
	539	watchdog_disable(hotcpu);
	540	break;
	541	#endif /* CONFIG_HOTPLUG_CPU */
	542	}
eac24335	543	return notifier_from_errno(err);
58687acb DZ	544	}
	545
	546	static struct notifier_block __cpuinitdata cpu_nfb = {
	547	.notifier_call = cpu_callback
	548	};
	549
	550	static int __init spawn_watchdog_task(void)
	551	{
	552	void cpu = (void )(long)smp_processor_id();
	553	int err;
	554
	555	if (no_watchdog)
	556	return 0;
	557
	558	err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu);
eac24335	559	WARN_ON(notifier_to_errno(err));
58687acb DZ	560
	561	cpu_callback(&cpu_nfb, CPU_ONLINE, cpu);
	562	register_cpu_notifier(&cpu_nfb);
	563
58687acb DZ	564	return 0;
	565	}
	566	early_initcall(spawn_watchdog_task);