[linux.git] / kernel / softlockup.c

/*
 * Detect Soft Lockups
 *
 * started by Ingo Molnar, Copyright (C) 2005, 2006 Red Hat, Inc.
 *
 * 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/nmi.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/freezer.h>
#include <linux/kthread.h>
#include <linux/notifier.h>
#include <linux/module.h>

#include <asm/irq_regs.h>

static DEFINE_SPINLOCK(print_lock);

static DEFINE_PER_CPU(unsigned long, touch_timestamp);
static DEFINE_PER_CPU(unsigned long, print_timestamp);
static DEFINE_PER_CPU(struct task_struct *, watchdog_task);

static int __read_mostly did_panic;
unsigned long __read_mostly softlockup_thresh = 60;

static int
softlock_panic(struct notifier_block *this, unsigned long event, void *ptr)
{
	did_panic = 1;

	return NOTIFY_DONE;
}

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

/*
 * 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 */
}

void touch_softlockup_watchdog(void)
{
	int this_cpu = raw_smp_processor_id();

	__raw_get_cpu_var(touch_timestamp) = get_timestamp(this_cpu);
}
EXPORT_SYMBOL(touch_softlockup_watchdog);

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);

/*
 * This callback runs from the timer interrupt, and checks
 * whether the watchdog thread has hung or not:
 */
void softlockup_tick(void)
{
	int this_cpu = smp_processor_id();
	unsigned long touch_timestamp = per_cpu(touch_timestamp, this_cpu);
	unsigned long print_timestamp;
	struct pt_regs *regs = get_irq_regs();
	unsigned long now;

	if (touch_timestamp == 0) {
		touch_softlockup_watchdog();
		return;
	}

	print_timestamp = per_cpu(print_timestamp, this_cpu);

	/* report at most once a second */
	if ((print_timestamp >= touch_timestamp &&
			print_timestamp < (touch_timestamp + 1)) ||
			did_panic || !per_cpu(watchdog_task, this_cpu)) {
		return;
	}

	/* do not print during early bootup: */
	if (unlikely(system_state != SYSTEM_RUNNING)) {
		touch_softlockup_watchdog();
		return;
	}

	now = get_timestamp(this_cpu);

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

	/* Warn about unreasonable delays: */
	if (now <= (touch_timestamp + softlockup_thresh))
		return;

	per_cpu(print_timestamp, this_cpu) = touch_timestamp;

	spin_lock(&print_lock);
	printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %lus! [%s:%d]\n",
			this_cpu, now - touch_timestamp,
			current->comm, task_pid_nr(current));
	if (regs)
		show_regs(regs);
	else
		dump_stack();
	spin_unlock(&print_lock);
}

/*
 * Have a reasonable limit on the number of tasks checked:
 */
unsigned long __read_mostly sysctl_hung_task_check_count = 1024;

/*
 * Zero means infinite timeout - no checking done:
 */
unsigned long __read_mostly sysctl_hung_task_timeout_secs = 120;

unsigned long __read_mostly sysctl_hung_task_warnings = 10;

/*
 * Only do the hung-tasks check on one CPU:
 */
static int check_cpu __read_mostly = -1;

static void check_hung_task(struct task_struct *t, unsigned long now)
{
	unsigned long switch_count = t->nvcsw + t->nivcsw;

	if (t->flags & PF_FROZEN)
		return;

	if (switch_count != t->last_switch_count || !t->last_switch_timestamp) {
		t->last_switch_count = switch_count;
		t->last_switch_timestamp = now;
		return;
	}
	if ((long)(now - t->last_switch_timestamp) <
					sysctl_hung_task_timeout_secs)
		return;
	if (sysctl_hung_task_warnings < 0)
		return;
	sysctl_hung_task_warnings--;

	/*
	 * Ok, the task did not get scheduled for more than 2 minutes,
	 * complain:
	 */
	printk(KERN_ERR "INFO: task %s:%d blocked for more than "
			"%ld seconds.\n", t->comm, t->pid,
			sysctl_hung_task_timeout_secs);
	printk(KERN_ERR "\"echo 0 > /proc/sys/kernel/hung_task_timeout_secs\""
			" disables this message.\n");
	sched_show_task(t);
	__debug_show_held_locks(t);

	t->last_switch_timestamp = now;
	touch_nmi_watchdog();
}

/*
 * Check whether a TASK_UNINTERRUPTIBLE does not get woken up for
 * a really long time (120 seconds). If that happens, print out
 * a warning.
 */
static void check_hung_uninterruptible_tasks(int this_cpu)
{
	int max_count = sysctl_hung_task_check_count;
	unsigned long now = get_timestamp(this_cpu);
	struct task_struct *g, *t;

	/*
	 * If the system crashed already then all bets are off,
	 * do not report extra hung tasks:
	 */
	if ((tainted & TAINT_DIE) || did_panic)
		return;

	read_lock(&tasklist_lock);
	do_each_thread(g, t) {
		if (!--max_count)
			goto unlock;
		if (t->state & TASK_UNINTERRUPTIBLE)
			check_hung_task(t, now);
	} while_each_thread(g, t);
 unlock:
	read_unlock(&tasklist_lock);
}

/*
 * The watchdog thread - runs every second and touches the timestamp.
 */
static int watchdog(void *__bind_cpu)
{
	struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
	int this_cpu = (long)__bind_cpu;

	sched_setscheduler(current, SCHED_FIFO, &param);

	/* initialize timestamp */
	touch_softlockup_watchdog();

	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 softlockup_tick().
	 */
	while (!kthread_should_stop()) {
		touch_softlockup_watchdog();
		schedule();

		if (kthread_should_stop())
			break;

		if (this_cpu == check_cpu) {
			if (sysctl_hung_task_timeout_secs)
				check_hung_uninterruptible_tasks(this_cpu);
		}

		set_current_state(TASK_INTERRUPTIBLE);
	}
	__set_current_state(TASK_RUNNING);

	return 0;
}

/*
 * 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;
	struct task_struct *p;

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
		BUG_ON(per_cpu(watchdog_task, hotcpu));
		p = kthread_create(watchdog, hcpu, "watchdog/%d", hotcpu);
		if (IS_ERR(p)) {
			printk(KERN_ERR "watchdog for %i failed\n", hotcpu);
			return NOTIFY_BAD;
		}
		per_cpu(touch_timestamp, hotcpu) = 0;
		per_cpu(watchdog_task, hotcpu) = p;
		kthread_bind(p, hotcpu);
		break;
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
		check_cpu = any_online_cpu(cpu_online_map);
		wake_up_process(per_cpu(watchdog_task, hotcpu));
		break;
#ifdef CONFIG_HOTPLUG_CPU
	case CPU_DOWN_PREPARE:
	case CPU_DOWN_PREPARE_FROZEN:
		if (hotcpu == check_cpu) {
			cpumask_t temp_cpu_online_map = cpu_online_map;

			cpu_clear(hotcpu, temp_cpu_online_map);
			check_cpu = any_online_cpu(temp_cpu_online_map);
		}
		break;

	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),
			     any_online_cpu(cpu_online_map));
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		p = per_cpu(watchdog_task, hotcpu);
		per_cpu(watchdog_task, hotcpu) = NULL;
		kthread_stop(p);
		break;
#endif /* CONFIG_HOTPLUG_CPU */
	}
	return NOTIFY_OK;
}

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

__init void spawn_softlockup_task(void)
{
	void *cpu = (void *)(long)smp_processor_id();
	int err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu);

	BUG_ON(err == NOTIFY_BAD);
	cpu_callback(&cpu_nfb, CPU_ONLINE, cpu);
	register_cpu_notifier(&cpu_nfb);

	atomic_notifier_chain_register(&panic_notifier_list, &panic_block);
}
Commit	Line	Data
8446f1d3 IM	1	/*
	2	* Detect Soft Lockups
	3	*
6687a97d	4	* started by Ingo Molnar, Copyright (C) 2005, 2006 Red Hat, Inc.
8446f1d3 IM	5	*
	6	* this code detects soft lockups: incidents in where on a CPU
	7	* the kernel does not reschedule for 10 seconds or more.
	8	*/
8446f1d3 IM	9	#include <linux/mm.h>
8446f1d3 IM	10	#include <linux/cpu.h>
82a1fcb9	11	#include <linux/nmi.h>
8446f1d3 IM	12	#include <linux/init.h>
8446f1d3 IM	13	#include <linux/delay.h>
83144186	14	#include <linux/freezer.h>
8446f1d3 IM	15	#include <linux/kthread.h>
	16	#include <linux/notifier.h>
	17	#include <linux/module.h>
	18
43581a10 IM	19	#include <asm/irq_regs.h>
43581a10 IM	20
8446f1d3 IM	21	static DEFINE_SPINLOCK(print_lock);
8446f1d3 IM	22
6687a97d IM	23	static DEFINE_PER_CPU(unsigned long, touch_timestamp);
6687a97d IM	24	static DEFINE_PER_CPU(unsigned long, print_timestamp);
8446f1d3 IM	25	static DEFINE_PER_CPU(struct task_struct *, watchdog_task);
8446f1d3 IM	26
90739081 IM	27	static int __read_mostly did_panic;
90739081 IM	28	unsigned long __read_mostly softlockup_thresh = 60;
6687a97d IM	29
	30	static int
	31	softlock_panic(struct notifier_block this, unsigned long event, void ptr)
8446f1d3 IM	32	{
	33	did_panic = 1;
	34
	35	return NOTIFY_DONE;
	36	}
	37
	38	static struct notifier_block panic_block = {
	39	.notifier_call = softlock_panic,
	40	};
	41
966812dc JF	42	/*
	43	* Returns seconds, approximately. We don't need nanosecond
	44	* resolution, and we don't need to waste time with a big divide when
	45	* 2^30ns == 1.074s.
	46	*/
a3b13c23	47	static unsigned long get_timestamp(int this_cpu)
966812dc	48	{
82a1fcb9	49	return cpu_clock(this_cpu) >> 30LL; /* 2^30 ~= 10^9 */
966812dc JF	50	}
966812dc JF	51
8446f1d3 IM	52	void touch_softlockup_watchdog(void)
8446f1d3 IM	53	{
a3b13c23 IM	54	int this_cpu = raw_smp_processor_id();
	55
	56	__raw_get_cpu_var(touch_timestamp) = get_timestamp(this_cpu);
8446f1d3 IM	57	}
	58	EXPORT_SYMBOL(touch_softlockup_watchdog);
	59
04c9167f JF	60	void touch_all_softlockup_watchdogs(void)
	61	{
	62	int cpu;
	63
	64	/* Cause each CPU to re-update its timestamp rather than complain */
	65	for_each_online_cpu(cpu)
	66	per_cpu(touch_timestamp, cpu) = 0;
	67	}
	68	EXPORT_SYMBOL(touch_all_softlockup_watchdogs);
	69
8446f1d3 IM	70	/*
	71	* This callback runs from the timer interrupt, and checks
	72	* whether the watchdog thread has hung or not:
	73	*/
6687a97d	74	void softlockup_tick(void)
8446f1d3 IM	75	{
8446f1d3 IM	76	int this_cpu = smp_processor_id();
6687a97d	77	unsigned long touch_timestamp = per_cpu(touch_timestamp, this_cpu);
966812dc	78	unsigned long print_timestamp;
43581a10	79	struct pt_regs *regs = get_irq_regs();
966812dc	80	unsigned long now;
8446f1d3	81
04c9167f JF	82	if (touch_timestamp == 0) {
04c9167f JF	83	touch_softlockup_watchdog();
966812dc	84	return;
04c9167f	85	}
966812dc JF	86
	87	print_timestamp = per_cpu(print_timestamp, this_cpu);
	88
	89	/* report at most once a second */
a115d5ca IM	90	if ((print_timestamp >= touch_timestamp &&
	91	print_timestamp < (touch_timestamp + 1)) \|\|
	92	did_panic \|\| !per_cpu(watchdog_task, this_cpu)) {
8446f1d3	93	return;
a115d5ca	94	}
8446f1d3	95
6687a97d IM	96	/* do not print during early bootup: */
	97	if (unlikely(system_state != SYSTEM_RUNNING)) {
	98	touch_softlockup_watchdog();
8446f1d3	99	return;
6687a97d	100	}
8446f1d3	101
a3b13c23	102	now = get_timestamp(this_cpu);
966812dc	103
ed50d6cb PZ	104	/* Wake up the high-prio watchdog task every second: */
	105	if (now > (touch_timestamp + 1))
	106	wake_up_process(per_cpu(watchdog_task, this_cpu));
	107
82a1fcb9	108	/* Warn about unreasonable delays: */
c4f3b63f	109	if (now <= (touch_timestamp + softlockup_thresh))
43581a10	110	return;
8446f1d3	111
43581a10 IM	112	per_cpu(print_timestamp, this_cpu) = touch_timestamp;
	113
	114	spin_lock(&print_lock);
c4f3b63f RT	115	printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %lus! [%s:%d]\n",
c4f3b63f RT	116	this_cpu, now - touch_timestamp,
ba25f9dc	117	current->comm, task_pid_nr(current));
43581a10 IM	118	if (regs)
	119	show_regs(regs);
	120	else
6687a97d	121	dump_stack();
43581a10	122	spin_unlock(&print_lock);
8446f1d3 IM	123	}
8446f1d3 IM	124
82a1fcb9 IM	125	/*
	126	* Have a reasonable limit on the number of tasks checked:
	127	*/
90739081	128	unsigned long __read_mostly sysctl_hung_task_check_count = 1024;
82a1fcb9 IM	129
	130	/*
	131	* Zero means infinite timeout - no checking done:
	132	*/
90739081	133	unsigned long __read_mostly sysctl_hung_task_timeout_secs = 120;
82a1fcb9	134
90739081	135	unsigned long __read_mostly sysctl_hung_task_warnings = 10;
82a1fcb9 IM	136
	137	/*
	138	* Only do the hung-tasks check on one CPU:
	139	*/
	140	static int check_cpu __read_mostly = -1;
	141
	142	static void check_hung_task(struct task_struct *t, unsigned long now)
	143	{
	144	unsigned long switch_count = t->nvcsw + t->nivcsw;
	145
	146	if (t->flags & PF_FROZEN)
	147	return;
	148
	149	if (switch_count != t->last_switch_count \|\| !t->last_switch_timestamp) {
	150	t->last_switch_count = switch_count;
	151	t->last_switch_timestamp = now;
	152	return;
	153	}
	154	if ((long)(now - t->last_switch_timestamp) <
	155	sysctl_hung_task_timeout_secs)
	156	return;
	157	if (sysctl_hung_task_warnings < 0)
	158	return;
	159	sysctl_hung_task_warnings--;
	160
	161	/*
	162	* Ok, the task did not get scheduled for more than 2 minutes,
	163	* complain:
	164	*/
	165	printk(KERN_ERR "INFO: task %s:%d blocked for more than "
	166	"%ld seconds.\n", t->comm, t->pid,
	167	sysctl_hung_task_timeout_secs);
	168	printk(KERN_ERR "\"echo 0 > /proc/sys/kernel/hung_task_timeout_secs\""
	169	" disables this message.\n");
	170	sched_show_task(t);
	171	__debug_show_held_locks(t);
	172
	173	t->last_switch_timestamp = now;
	174	touch_nmi_watchdog();
	175	}
	176
	177	/*
	178	* Check whether a TASK_UNINTERRUPTIBLE does not get woken up for
	179	* a really long time (120 seconds). If that happens, print out
	180	* a warning.
	181	*/
	182	static void check_hung_uninterruptible_tasks(int this_cpu)
	183	{
	184	int max_count = sysctl_hung_task_check_count;
	185	unsigned long now = get_timestamp(this_cpu);
	186	struct task_struct g, t;
	187
	188	/*
	189	* If the system crashed already then all bets are off,
	190	* do not report extra hung tasks:
	191	*/
	192	if ((tainted & TAINT_DIE) \|\| did_panic)
	193	return;
	194
	195	read_lock(&tasklist_lock);
	196	do_each_thread(g, t) {
	197	if (!--max_count)
ed50d6cb	198	goto unlock;
82a1fcb9 IM	199	if (t->state & TASK_UNINTERRUPTIBLE)
	200	check_hung_task(t, now);
	201	} while_each_thread(g, t);
ed50d6cb	202	unlock:
82a1fcb9 IM	203	read_unlock(&tasklist_lock);
	204	}
	205
8446f1d3 IM	206	/*
	207	* The watchdog thread - runs every second and touches the timestamp.
	208	*/
a5f2ce3c	209	static int watchdog(void *__bind_cpu)
8446f1d3	210	{
02fb6149	211	struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
82a1fcb9	212	int this_cpu = (long)__bind_cpu;
8446f1d3 IM	213
8446f1d3 IM	214	sched_setscheduler(current, SCHED_FIFO, &param);
8446f1d3	215
966812dc JF	216	/* initialize timestamp */
	217	touch_softlockup_watchdog();
	218
7be2a03e	219	set_current_state(TASK_INTERRUPTIBLE);
8446f1d3	220	/*
6687a97d	221	* Run briefly once per second to reset the softlockup timestamp.
82a1fcb9	222	* If this gets delayed for more than 60 seconds then the
6687a97d	223	* debug-printout triggers in softlockup_tick().
8446f1d3 IM	224	*/
8446f1d3 IM	225	while (!kthread_should_stop()) {
8446f1d3	226	touch_softlockup_watchdog();
ed50d6cb PZ	227	schedule();
	228
	229	if (kthread_should_stop())
	230	break;
82a1fcb9	231
7be2a03e DA	232	if (this_cpu == check_cpu) {
	233	if (sysctl_hung_task_timeout_secs)
	234	check_hung_uninterruptible_tasks(this_cpu);
	235	}
ed50d6cb	236
7be2a03e	237	set_current_state(TASK_INTERRUPTIBLE);
8446f1d3	238	}
7be2a03e	239	__set_current_state(TASK_RUNNING);
8446f1d3 IM	240
	241	return 0;
	242	}
	243
	244	/*
	245	* Create/destroy watchdog threads as CPUs come and go:
	246	*/
8c78f307	247	static int __cpuinit
8446f1d3 IM	248	cpu_callback(struct notifier_block nfb, unsigned long action, void hcpu)
	249	{
	250	int hotcpu = (unsigned long)hcpu;
	251	struct task_struct *p;
	252
	253	switch (action) {
	254	case CPU_UP_PREPARE:
8bb78442	255	case CPU_UP_PREPARE_FROZEN:
8446f1d3 IM	256	BUG_ON(per_cpu(watchdog_task, hotcpu));
	257	p = kthread_create(watchdog, hcpu, "watchdog/%d", hotcpu);
	258	if (IS_ERR(p)) {
a5f2ce3c	259	printk(KERN_ERR "watchdog for %i failed\n", hotcpu);
8446f1d3 IM	260	return NOTIFY_BAD;
8446f1d3 IM	261	}
a5f2ce3c IM	262	per_cpu(touch_timestamp, hotcpu) = 0;
a5f2ce3c IM	263	per_cpu(watchdog_task, hotcpu) = p;
8446f1d3	264	kthread_bind(p, hotcpu);
a5f2ce3c	265	break;
8446f1d3	266	case CPU_ONLINE:
8bb78442	267	case CPU_ONLINE_FROZEN:
82a1fcb9	268	check_cpu = any_online_cpu(cpu_online_map);
8446f1d3 IM	269	wake_up_process(per_cpu(watchdog_task, hotcpu));
	270	break;
	271	#ifdef CONFIG_HOTPLUG_CPU
82a1fcb9 IM	272	case CPU_DOWN_PREPARE:
	273	case CPU_DOWN_PREPARE_FROZEN:
	274	if (hotcpu == check_cpu) {
	275	cpumask_t temp_cpu_online_map = cpu_online_map;
	276
	277	cpu_clear(hotcpu, temp_cpu_online_map);
	278	check_cpu = any_online_cpu(temp_cpu_online_map);
	279	}
	280	break;
ed50d6cb PZ	281
	282	case CPU_UP_CANCELED:
	283	case CPU_UP_CANCELED_FROZEN:
	284	if (!per_cpu(watchdog_task, hotcpu))
	285	break;
	286	/* Unbind so it can run. Fall thru. */
	287	kthread_bind(per_cpu(watchdog_task, hotcpu),
	288	any_online_cpu(cpu_online_map));
8446f1d3	289	case CPU_DEAD:
8bb78442	290	case CPU_DEAD_FROZEN:
8446f1d3 IM	291	p = per_cpu(watchdog_task, hotcpu);
	292	per_cpu(watchdog_task, hotcpu) = NULL;
	293	kthread_stop(p);
	294	break;
	295	#endif /* CONFIG_HOTPLUG_CPU */
a5f2ce3c	296	}
8446f1d3 IM	297	return NOTIFY_OK;
	298	}
	299
8c78f307	300	static struct notifier_block __cpuinitdata cpu_nfb = {
8446f1d3 IM	301	.notifier_call = cpu_callback
	302	};
	303
	304	__init void spawn_softlockup_task(void)
	305	{
	306	void cpu = (void )(long)smp_processor_id();
07dccf33	307	int err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu);
8446f1d3	308
07dccf33	309	BUG_ON(err == NOTIFY_BAD);
8446f1d3 IM	310	cpu_callback(&cpu_nfb, CPU_ONLINE, cpu);
	311	register_cpu_notifier(&cpu_nfb);
	312
e041c683	313	atomic_notifier_chain_register(&panic_notifier_list, &panic_block);
8446f1d3	314	}