watchdog.c 15.4 KB
Newer Older
1 2 3 4 5
/*
 * Detect hard and soft lockups on a system
 *
 * started by Don Zickus, Copyright (C) 2010 Red Hat, Inc.
 *
6 7 8
 * Note: Most of this code is borrowed heavily from the original softlockup
 * detector, so thanks to Ingo for the initial implementation.
 * Some chunks also taken from the old x86-specific nmi watchdog code, thanks
9 10 11
 * to those contributors as well.
 */

12 13
#define pr_fmt(fmt) "NMI watchdog: " fmt

14 15 16 17 18 19 20 21 22 23 24 25 26
#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>
27
#include <linux/kvm_para.h>
28 29
#include <linux/perf_event.h>

30
int watchdog_enabled = 1;
31
int __read_mostly watchdog_thresh = 10;
32 33 34 35 36 37

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);
38
#ifdef CONFIG_HARDLOCKUP_DETECTOR
39 40
static DEFINE_PER_CPU(bool, hard_watchdog_warn);
static DEFINE_PER_CPU(bool, watchdog_nmi_touch);
41 42 43 44 45 46 47 48 49
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

/* boot commands */
/*
 * Should we panic when a soft-lockup or hard-lockup occurs:
 */
50
#ifdef CONFIG_HARDLOCKUP_DETECTOR
51 52
static int hardlockup_panic =
			CONFIG_BOOTPARAM_HARDLOCKUP_PANIC_VALUE;
53 54 55 56 57

static int __init hardlockup_panic_setup(char *str)
{
	if (!strncmp(str, "panic", 5))
		hardlockup_panic = 1;
58 59
	else if (!strncmp(str, "nopanic", 7))
		hardlockup_panic = 0;
60
	else if (!strncmp(str, "0", 1))
61
		watchdog_enabled = 0;
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
	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)
{
80
	watchdog_enabled = 0;
81 82 83 84 85 86 87
	return 1;
}
__setup("nowatchdog", nowatchdog_setup);

/* deprecated */
static int __init nosoftlockup_setup(char *str)
{
88
	watchdog_enabled = 0;
89 90 91 92 93
	return 1;
}
__setup("nosoftlockup", nosoftlockup_setup);
/*  */

94 95 96 97 98 99 100
/*
 * Hard-lockup warnings should be triggered after just a few seconds. Soft-
 * lockups can have false positives under extreme conditions. So we generally
 * want a higher threshold for soft lockups than for hard lockups. So we couple
 * the thresholds with a factor: we make the soft threshold twice the amount of
 * time the hard threshold is.
 */
101
static int get_softlockup_thresh(void)
102 103 104
{
	return watchdog_thresh * 2;
}
105 106 107 108 109 110 111 112 113 114 115 116 117 118

/*
 * 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)
{
	/*
119
	 * convert watchdog_thresh from seconds to ns
120 121 122 123
	 * the divide by 5 is to give hrtimer several chances (two
	 * or three with the current relation between the soft
	 * and hard thresholds) to increment before the
	 * hardlockup detector generates a warning
124
	 */
125
	return get_softlockup_thresh() * (NSEC_PER_SEC / 5);
126 127 128 129 130
}

/* Commands for resetting the watchdog */
static void __touch_watchdog(void)
{
131
	int this_cpu = smp_processor_id();
132

133
	__this_cpu_write(watchdog_touch_ts, get_timestamp(this_cpu));
134 135
}

136
void touch_softlockup_watchdog(void)
137
{
138
	__this_cpu_write(watchdog_touch_ts, 0);
139
}
140
EXPORT_SYMBOL(touch_softlockup_watchdog);
141

142
void touch_all_softlockup_watchdogs(void)
143 144 145 146 147 148 149 150 151 152 153 154
{
	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;
}

155
#ifdef CONFIG_HARDLOCKUP_DETECTOR
156 157
void touch_nmi_watchdog(void)
{
158 159 160 161 162 163 164 165
	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;
		}
	}
166
	touch_softlockup_watchdog();
167 168 169
}
EXPORT_SYMBOL(touch_nmi_watchdog);

170 171
#endif

172 173 174 175 176 177
void touch_softlockup_watchdog_sync(void)
{
	__raw_get_cpu_var(softlockup_touch_sync) = true;
	__raw_get_cpu_var(watchdog_touch_ts) = 0;
}

178
#ifdef CONFIG_HARDLOCKUP_DETECTOR
179
/* watchdog detector functions */
180
static int is_hardlockup(void)
181
{
182
	unsigned long hrint = __this_cpu_read(hrtimer_interrupts);
183

184
	if (__this_cpu_read(hrtimer_interrupts_saved) == hrint)
185 186
		return 1;

187
	__this_cpu_write(hrtimer_interrupts_saved, hrint);
188 189 190 191
	return 0;
}
#endif

192
static int is_softlockup(unsigned long touch_ts)
193
{
194
	unsigned long now = get_timestamp(smp_processor_id());
195 196

	/* Warn about unreasonable delays: */
197
	if (time_after(now, touch_ts + get_softlockup_thresh()))
198 199 200 201 202
		return now - touch_ts;

	return 0;
}

203
#ifdef CONFIG_HARDLOCKUP_DETECTOR
204

205 206 207 208 209 210 211 212 213
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 */
214
static void watchdog_overflow_callback(struct perf_event *event,
215 216 217
		 struct perf_sample_data *data,
		 struct pt_regs *regs)
{
218 219 220
	/* Ensure the watchdog never gets throttled */
	event->hw.interrupts = 0;

221 222
	if (__this_cpu_read(watchdog_nmi_touch) == true) {
		__this_cpu_write(watchdog_nmi_touch, false);
223 224 225 226 227 228 229 230 231
		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
	 */
232 233 234
	if (is_hardlockup()) {
		int this_cpu = smp_processor_id();

235
		/* only print hardlockups once */
236
		if (__this_cpu_read(hard_watchdog_warn) == true)
237 238 239 240 241 242 243
			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);

244
		__this_cpu_write(hard_watchdog_warn, true);
245 246 247
		return;
	}

248
	__this_cpu_write(hard_watchdog_warn, false);
249 250 251 252
	return;
}
static void watchdog_interrupt_count(void)
{
253
	__this_cpu_inc(hrtimer_interrupts);
254 255 256
}
#else
static inline void watchdog_interrupt_count(void) { return; }
257
#endif /* CONFIG_HARDLOCKUP_DETECTOR */
258 259 260 261

/* watchdog kicker functions */
static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
{
262
	unsigned long touch_ts = __this_cpu_read(watchdog_touch_ts);
263 264 265 266 267 268 269
	struct pt_regs *regs = get_irq_regs();
	int duration;

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

	/* kick the softlockup detector */
270
	wake_up_process(__this_cpu_read(softlockup_watchdog));
271 272 273 274 275

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

	if (touch_ts == 0) {
276
		if (unlikely(__this_cpu_read(softlockup_touch_sync))) {
277 278 279 280
			/*
			 * If the time stamp was touched atomically
			 * make sure the scheduler tick is up to date.
			 */
281
			__this_cpu_write(softlockup_touch_sync, false);
282 283
			sched_clock_tick();
		}
284 285 286

		/* Clear the guest paused flag on watchdog reset */
		kvm_check_and_clear_guest_paused();
287 288 289 290 291 292 293 294 295 296
		__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
	 */
297
	duration = is_softlockup(touch_ts);
298
	if (unlikely(duration)) {
299 300 301 302 303 304 305 306
		/*
		 * If a virtual machine is stopped by the host it can look to
		 * the watchdog like a soft lockup, check to see if the host
		 * stopped the vm before we issue the warning
		 */
		if (kvm_check_and_clear_guest_paused())
			return HRTIMER_RESTART;

307
		/* only warn once */
308
		if (__this_cpu_read(soft_watchdog_warn) == true)
309 310
			return HRTIMER_RESTART;

311
		printk(KERN_EMERG "BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
312
			smp_processor_id(), duration,
313 314 315 316 317 318 319 320 321 322
			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");
323
		__this_cpu_write(soft_watchdog_warn, true);
324
	} else
325
		__this_cpu_write(soft_watchdog_warn, false);
326 327 328 329 330 331 332 333

	return HRTIMER_RESTART;
}


/*
 * The watchdog thread - touches the timestamp.
 */
334
static int watchdog(void *unused)
335
{
336
	struct sched_param param = { .sched_priority = 0 };
337
	struct hrtimer *hrtimer = &__raw_get_cpu_var(watchdog_hrtimer);
338 339 340 341 342 343 344 345 346 347 348

	/* 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);
	/*
349 350 351 352 353
	 * Run briefly (kicked by the hrtimer callback function) once every
	 * get_sample_period() seconds (4 seconds by default) to reset the
	 * softlockup timestamp. If this gets delayed for more than
	 * 2*watchdog_thresh seconds then the debug-printout triggers in
	 * watchdog_timer_fn().
354 355 356 357 358 359 360 361 362 363
	 */
	while (!kthread_should_stop()) {
		__touch_watchdog();
		schedule();

		if (kthread_should_stop())
			break;

		set_current_state(TASK_INTERRUPTIBLE);
	}
364 365 366 367
	/*
	 * Drop the policy/priority elevation during thread exit to avoid a
	 * scheduling latency spike.
	 */
368
	__set_current_state(TASK_RUNNING);
369
	sched_setscheduler(current, SCHED_NORMAL, &param);
370 371 372 373
	return 0;
}


374
#ifdef CONFIG_HARDLOCKUP_DETECTOR
375 376 377 378 379 380 381
/*
 * People like the simple clean cpu node info on boot.
 * Reduce the watchdog noise by only printing messages
 * that are different from what cpu0 displayed.
 */
static unsigned long cpu0_err;

382 383 384 385 386 387 388 389 390 391 392 393 394 395
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;

	wd_attr = &wd_hw_attr;
396
	wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh);
397 398

	/* Try to register using hardware perf events */
399
	event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback, NULL);
400 401 402 403 404

	/* save cpu0 error for future comparision */
	if (cpu == 0 && IS_ERR(event))
		cpu0_err = PTR_ERR(event);

405
	if (!IS_ERR(event)) {
406 407 408
		/* only print for cpu0 or different than cpu0 */
		if (cpu == 0 || cpu0_err)
			pr_info("enabled on all CPUs, permanently consumes one hw-PMU counter.\n");
409 410 411
		goto out_save;
	}

412 413 414
	/* skip displaying the same error again */
	if (cpu > 0 && (PTR_ERR(event) == cpu0_err))
		return PTR_ERR(event);
415 416 417

	/* vary the KERN level based on the returned errno */
	if (PTR_ERR(event) == -EOPNOTSUPP)
418
		pr_info("disabled (cpu%i): not supported (no LAPIC?)\n", cpu);
419
	else if (PTR_ERR(event) == -ENOENT)
420 421
		pr_warning("disabled (cpu%i): hardware events not enabled\n",
			 cpu);
422
	else
423 424
		pr_err("disabled (cpu%i): unable to create perf event: %ld\n",
			cpu, PTR_ERR(event));
425
	return PTR_ERR(event);
426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451

	/* 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; }
452
#endif /* CONFIG_HARDLOCKUP_DETECTOR */
453 454

/* prepare/enable/disable routines */
455
static void watchdog_prepare_cpu(int cpu)
456 457 458 459 460 461 462 463 464 465 466
{
	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;
}

static int watchdog_enable(int cpu)
{
	struct task_struct *p = per_cpu(softlockup_watchdog, cpu);
467
	int err = 0;
468 469

	/* enable the perf event */
470
	err = watchdog_nmi_enable(cpu);
471 472

	/* Regardless of err above, fall through and start softlockup */
473 474 475

	/* create the watchdog thread */
	if (!p) {
476
		struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
477
		p = kthread_create_on_node(watchdog, NULL, cpu_to_node(cpu), "watchdog/%d", cpu);
478
		if (IS_ERR(p)) {
479
			pr_err("softlockup watchdog for %i failed\n", cpu);
480
			if (!err) {
481 482
				/* if hardlockup hasn't already set this */
				err = PTR_ERR(p);
483 484 485
				/* and disable the perf event */
				watchdog_nmi_disable(cpu);
			}
486
			goto out;
487
		}
488
		sched_setscheduler(p, SCHED_FIFO, &param);
489 490 491 492 493 494
		kthread_bind(p, cpu);
		per_cpu(watchdog_touch_ts, cpu) = 0;
		per_cpu(softlockup_watchdog, cpu) = p;
		wake_up_process(p);
	}

495 496
out:
	return err;
497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
}

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);
	}
}

520 521
/* sysctl functions */
#ifdef CONFIG_SYSCTL
522 523 524
static void watchdog_enable_all_cpus(void)
{
	int cpu;
525 526

	watchdog_enabled = 0;
527 528

	for_each_online_cpu(cpu)
529 530 531 532
		if (!watchdog_enable(cpu))
			/* if any cpu succeeds, watchdog is considered
			   enabled for the system */
			watchdog_enabled = 1;
533

534
	if (!watchdog_enabled)
535
		pr_err("failed to be enabled on some cpus\n");
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551

}

static void watchdog_disable_all_cpus(void)
{
	int cpu;

	for_each_online_cpu(cpu)
		watchdog_disable(cpu);

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


/*
552
 * proc handler for /proc/sys/kernel/nmi_watchdog,watchdog_thresh
553 554
 */

555 556
int proc_dowatchdog(struct ctl_table *table, int write,
		    void __user *buffer, size_t *lenp, loff_t *ppos)
557
{
558
	int ret;
559

560
	ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
561 562 563
	if (ret || !write)
		goto out;

564
	if (watchdog_enabled && watchdog_thresh)
565 566 567 568 569 570
		watchdog_enable_all_cpus();
	else
		watchdog_disable_all_cpus();

out:
	return ret;
571 572 573 574 575 576 577
}
#endif /* CONFIG_SYSCTL */


/*
 * Create/destroy watchdog threads as CPUs come and go:
 */
578
static int __cpuinit
579 580 581 582 583 584 585
cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu)
{
	int hotcpu = (unsigned long)hcpu;

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
586
		watchdog_prepare_cpu(hotcpu);
587 588 589
		break;
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
590
		if (watchdog_enabled)
591
			watchdog_enable(hotcpu);
592 593 594 595 596 597 598 599 600 601 602 603
		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 */
	}
604 605 606 607 608 609 610

	/*
	 * hardlockup and softlockup are not important enough
	 * to block cpu bring up.  Just always succeed and
	 * rely on printk output to flag problems.
	 */
	return NOTIFY_OK;
611 612
}

613
static struct notifier_block __cpuinitdata cpu_nfb = {
614 615 616
	.notifier_call = cpu_callback
};

617
void __init lockup_detector_init(void)
618 619 620 621 622
{
	void *cpu = (void *)(long)smp_processor_id();
	int err;

	err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu);
623
	WARN_ON(notifier_to_errno(err));
624 625 626 627

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

628
	return;
629
}