watchdog.c 24.2 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
#include <linux/mm.h>
#include <linux/cpu.h>
#include <linux/nmi.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/sysctl.h>
20
#include <linux/smpboot.h>
21
#include <linux/sched/rt.h>
22 23

#include <asm/irq_regs.h>
24
#include <linux/kvm_para.h>
25 26
#include <linux/perf_event.h>

27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
/*
 * The run state of the lockup detectors is controlled by the content of the
 * 'watchdog_enabled' variable. Each lockup detector has its dedicated bit -
 * bit 0 for the hard lockup detector and bit 1 for the soft lockup detector.
 *
 * 'watchdog_user_enabled', 'nmi_watchdog_enabled' and 'soft_watchdog_enabled'
 * are variables that are only used as an 'interface' between the parameters
 * in /proc/sys/kernel and the internal state bits in 'watchdog_enabled'. The
 * 'watchdog_thresh' variable is handled differently because its value is not
 * boolean, and the lockup detectors are 'suspended' while 'watchdog_thresh'
 * is equal zero.
 */
#define NMI_WATCHDOG_ENABLED_BIT   0
#define SOFT_WATCHDOG_ENABLED_BIT  1
#define NMI_WATCHDOG_ENABLED      (1 << NMI_WATCHDOG_ENABLED_BIT)
#define SOFT_WATCHDOG_ENABLED     (1 << SOFT_WATCHDOG_ENABLED_BIT)

44 45
static DEFINE_MUTEX(watchdog_proc_mutex);

46 47 48 49 50 51 52 53
#ifdef CONFIG_HARDLOCKUP_DETECTOR
static unsigned long __read_mostly watchdog_enabled = SOFT_WATCHDOG_ENABLED|NMI_WATCHDOG_ENABLED;
#else
static unsigned long __read_mostly watchdog_enabled = SOFT_WATCHDOG_ENABLED;
#endif
int __read_mostly nmi_watchdog_enabled;
int __read_mostly soft_watchdog_enabled;
int __read_mostly watchdog_user_enabled;
54
int __read_mostly watchdog_thresh = 10;
55

56 57 58 59 60 61
#ifdef CONFIG_SMP
int __read_mostly sysctl_softlockup_all_cpu_backtrace;
#else
#define sysctl_softlockup_all_cpu_backtrace 0
#endif

62
static int __read_mostly watchdog_running;
63
static u64 __read_mostly sample_period;
64 65 66 67 68 69

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);
70 71
static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts);
static DEFINE_PER_CPU(unsigned long, soft_lockup_hrtimer_cnt);
72
static DEFINE_PER_CPU(struct task_struct *, softlockup_task_ptr_saved);
73
#ifdef CONFIG_HARDLOCKUP_DETECTOR
74 75
static DEFINE_PER_CPU(bool, hard_watchdog_warn);
static DEFINE_PER_CPU(bool, watchdog_nmi_touch);
76 77 78
static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved);
static DEFINE_PER_CPU(struct perf_event *, watchdog_ev);
#endif
79
static unsigned long soft_lockup_nmi_warn;
80 81 82 83 84

/* boot commands */
/*
 * Should we panic when a soft-lockup or hard-lockup occurs:
 */
85
#ifdef CONFIG_HARDLOCKUP_DETECTOR
86 87
static int hardlockup_panic =
			CONFIG_BOOTPARAM_HARDLOCKUP_PANIC_VALUE;
88 89 90 91 92 93 94 95
/*
 * We may not want to enable hard lockup detection by default in all cases,
 * for example when running the kernel as a guest on a hypervisor. In these
 * cases this function can be called to disable hard lockup detection. This
 * function should only be executed once by the boot processor before the
 * kernel command line parameters are parsed, because otherwise it is not
 * possible to override this in hardlockup_panic_setup().
 */
96
void hardlockup_detector_disable(void)
97
{
98
	watchdog_enabled &= ~NMI_WATCHDOG_ENABLED;
99 100
}

101 102 103 104
static int __init hardlockup_panic_setup(char *str)
{
	if (!strncmp(str, "panic", 5))
		hardlockup_panic = 1;
105 106
	else if (!strncmp(str, "nopanic", 7))
		hardlockup_panic = 0;
107
	else if (!strncmp(str, "0", 1))
108 109 110
		watchdog_enabled &= ~NMI_WATCHDOG_ENABLED;
	else if (!strncmp(str, "1", 1))
		watchdog_enabled |= NMI_WATCHDOG_ENABLED;
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
	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)
{
129
	watchdog_enabled = 0;
130 131 132 133 134 135
	return 1;
}
__setup("nowatchdog", nowatchdog_setup);

static int __init nosoftlockup_setup(char *str)
{
136
	watchdog_enabled &= ~SOFT_WATCHDOG_ENABLED;
137 138 139
	return 1;
}
__setup("nosoftlockup", nosoftlockup_setup);
140

141 142 143 144 145 146 147 148 149
#ifdef CONFIG_SMP
static int __init softlockup_all_cpu_backtrace_setup(char *str)
{
	sysctl_softlockup_all_cpu_backtrace =
		!!simple_strtol(str, NULL, 0);
	return 1;
}
__setup("softlockup_all_cpu_backtrace=", softlockup_all_cpu_backtrace_setup);
#endif
150

151 152 153 154 155 156 157
/*
 * 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.
 */
158
static int get_softlockup_thresh(void)
159 160 161
{
	return watchdog_thresh * 2;
}
162 163 164 165 166 167

/*
 * 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.
 */
168
static unsigned long get_timestamp(void)
169
{
170
	return running_clock() >> 30LL;  /* 2^30 ~= 10^9 */
171 172
}

173
static void set_sample_period(void)
174 175
{
	/*
176
	 * convert watchdog_thresh from seconds to ns
177 178 179 180
	 * 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
181
	 */
182
	sample_period = get_softlockup_thresh() * ((u64)NSEC_PER_SEC / 5);
183 184 185 186 187
}

/* Commands for resetting the watchdog */
static void __touch_watchdog(void)
{
188
	__this_cpu_write(watchdog_touch_ts, get_timestamp());
189 190
}

191
void touch_softlockup_watchdog(void)
192
{
193 194 195 196 197
	/*
	 * Preemption can be enabled.  It doesn't matter which CPU's timestamp
	 * gets zeroed here, so use the raw_ operation.
	 */
	raw_cpu_write(watchdog_touch_ts, 0);
198
}
199
EXPORT_SYMBOL(touch_softlockup_watchdog);
200

201
void touch_all_softlockup_watchdogs(void)
202 203 204 205 206 207 208 209 210 211 212 213
{
	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;
}

214
#ifdef CONFIG_HARDLOCKUP_DETECTOR
215 216
void touch_nmi_watchdog(void)
{
217 218 219 220 221 222 223
	/*
	 * Using __raw here because some code paths have
	 * preemption enabled.  If preemption is enabled
	 * then interrupts should be enabled too, in which
	 * case we shouldn't have to worry about the watchdog
	 * going off.
	 */
224
	raw_cpu_write(watchdog_nmi_touch, true);
225
	touch_softlockup_watchdog();
226 227 228
}
EXPORT_SYMBOL(touch_nmi_watchdog);

229 230
#endif

231 232
void touch_softlockup_watchdog_sync(void)
{
233 234
	__this_cpu_write(softlockup_touch_sync, true);
	__this_cpu_write(watchdog_touch_ts, 0);
235 236
}

237
#ifdef CONFIG_HARDLOCKUP_DETECTOR
238
/* watchdog detector functions */
239
static int is_hardlockup(void)
240
{
241
	unsigned long hrint = __this_cpu_read(hrtimer_interrupts);
242

243
	if (__this_cpu_read(hrtimer_interrupts_saved) == hrint)
244 245
		return 1;

246
	__this_cpu_write(hrtimer_interrupts_saved, hrint);
247 248 249 250
	return 0;
}
#endif

251
static int is_softlockup(unsigned long touch_ts)
252
{
253
	unsigned long now = get_timestamp();
254

255 256 257 258 259
	if (watchdog_enabled & SOFT_WATCHDOG_ENABLED) {
		/* Warn about unreasonable delays. */
		if (time_after(now, touch_ts + get_softlockup_thresh()))
			return now - touch_ts;
	}
260 261 262
	return 0;
}

263
#ifdef CONFIG_HARDLOCKUP_DETECTOR
264

265 266 267 268 269 270 271 272 273
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 */
274
static void watchdog_overflow_callback(struct perf_event *event,
275 276 277
		 struct perf_sample_data *data,
		 struct pt_regs *regs)
{
278 279 280
	/* Ensure the watchdog never gets throttled */
	event->hw.interrupts = 0;

281 282
	if (__this_cpu_read(watchdog_nmi_touch) == true) {
		__this_cpu_write(watchdog_nmi_touch, false);
283 284 285 286 287 288 289 290 291
		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
	 */
292 293 294
	if (is_hardlockup()) {
		int this_cpu = smp_processor_id();

295
		/* only print hardlockups once */
296
		if (__this_cpu_read(hard_watchdog_warn) == true)
297 298 299
			return;

		if (hardlockup_panic)
300 301
			panic("Watchdog detected hard LOCKUP on cpu %d",
			      this_cpu);
302
		else
303 304
			WARN(1, "Watchdog detected hard LOCKUP on cpu %d",
			     this_cpu);
305

306
		__this_cpu_write(hard_watchdog_warn, true);
307 308 309
		return;
	}

310
	__this_cpu_write(hard_watchdog_warn, false);
311 312
	return;
}
313 314
#endif /* CONFIG_HARDLOCKUP_DETECTOR */

315 316
static void watchdog_interrupt_count(void)
{
317
	__this_cpu_inc(hrtimer_interrupts);
318
}
319 320 321

static int watchdog_nmi_enable(unsigned int cpu);
static void watchdog_nmi_disable(unsigned int cpu);
322 323 324 325

/* watchdog kicker functions */
static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
{
326
	unsigned long touch_ts = __this_cpu_read(watchdog_touch_ts);
327 328
	struct pt_regs *regs = get_irq_regs();
	int duration;
329
	int softlockup_all_cpu_backtrace = sysctl_softlockup_all_cpu_backtrace;
330 331 332 333 334

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

	/* kick the softlockup detector */
335
	wake_up_process(__this_cpu_read(softlockup_watchdog));
336 337

	/* .. and repeat */
338
	hrtimer_forward_now(hrtimer, ns_to_ktime(sample_period));
339 340

	if (touch_ts == 0) {
341
		if (unlikely(__this_cpu_read(softlockup_touch_sync))) {
342 343 344 345
			/*
			 * If the time stamp was touched atomically
			 * make sure the scheduler tick is up to date.
			 */
346
			__this_cpu_write(softlockup_touch_sync, false);
347 348
			sched_clock_tick();
		}
349 350 351

		/* Clear the guest paused flag on watchdog reset */
		kvm_check_and_clear_guest_paused();
352 353 354 355 356 357 358 359 360 361
		__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
	 */
362
	duration = is_softlockup(touch_ts);
363
	if (unlikely(duration)) {
364 365 366 367 368 369 370 371
		/*
		 * 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;

372
		/* only warn once */
373 374 375 376 377 378 379 380 381 382 383 384 385 386
		if (__this_cpu_read(soft_watchdog_warn) == true) {
			/*
			 * When multiple processes are causing softlockups the
			 * softlockup detector only warns on the first one
			 * because the code relies on a full quiet cycle to
			 * re-arm.  The second process prevents the quiet cycle
			 * and never gets reported.  Use task pointers to detect
			 * this.
			 */
			if (__this_cpu_read(softlockup_task_ptr_saved) !=
			    current) {
				__this_cpu_write(soft_watchdog_warn, false);
				__touch_watchdog();
			}
387
			return HRTIMER_RESTART;
388
		}
389

390 391 392 393 394 395 396 397 398 399 400
		if (softlockup_all_cpu_backtrace) {
			/* Prevent multiple soft-lockup reports if one cpu is already
			 * engaged in dumping cpu back traces
			 */
			if (test_and_set_bit(0, &soft_lockup_nmi_warn)) {
				/* Someone else will report us. Let's give up */
				__this_cpu_write(soft_watchdog_warn, true);
				return HRTIMER_RESTART;
			}
		}

401
		pr_emerg("BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
402
			smp_processor_id(), duration,
403
			current->comm, task_pid_nr(current));
404
		__this_cpu_write(softlockup_task_ptr_saved, current);
405 406 407 408 409 410 411
		print_modules();
		print_irqtrace_events(current);
		if (regs)
			show_regs(regs);
		else
			dump_stack();

412 413 414 415 416 417 418 419 420 421 422
		if (softlockup_all_cpu_backtrace) {
			/* Avoid generating two back traces for current
			 * given that one is already made above
			 */
			trigger_allbutself_cpu_backtrace();

			clear_bit(0, &soft_lockup_nmi_warn);
			/* Barrier to sync with other cpus */
			smp_mb__after_atomic();
		}

Josh Hunt's avatar
Josh Hunt committed
423
		add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
424 425
		if (softlockup_panic)
			panic("softlockup: hung tasks");
426
		__this_cpu_write(soft_watchdog_warn, true);
427
	} else
428
		__this_cpu_write(soft_watchdog_warn, false);
429 430 431 432

	return HRTIMER_RESTART;
}

433 434 435
static void watchdog_set_prio(unsigned int policy, unsigned int prio)
{
	struct sched_param param = { .sched_priority = prio };
436

437 438 439 440
	sched_setscheduler(current, policy, &param);
}

static void watchdog_enable(unsigned int cpu)
441
{
442
	struct hrtimer *hrtimer = raw_cpu_ptr(&watchdog_hrtimer);
443

444 445 446 447
	/* kick off the timer for the hardlockup detector */
	hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	hrtimer->function = watchdog_timer_fn;

448 449
	/* Enable the perf event */
	watchdog_nmi_enable(cpu);
450 451

	/* done here because hrtimer_start can only pin to smp_processor_id() */
452
	hrtimer_start(hrtimer, ns_to_ktime(sample_period),
453 454
		      HRTIMER_MODE_REL_PINNED);

455 456 457 458
	/* initialize timestamp */
	watchdog_set_prio(SCHED_FIFO, MAX_RT_PRIO - 1);
	__touch_watchdog();
}
459

460 461
static void watchdog_disable(unsigned int cpu)
{
462
	struct hrtimer *hrtimer = raw_cpu_ptr(&watchdog_hrtimer);
463

464 465 466 467
	watchdog_set_prio(SCHED_NORMAL, 0);
	hrtimer_cancel(hrtimer);
	/* disable the perf event */
	watchdog_nmi_disable(cpu);
468 469
}

470 471 472 473 474
static void watchdog_cleanup(unsigned int cpu, bool online)
{
	watchdog_disable(cpu);
}

475 476 477 478 479 480 481 482 483
static int watchdog_should_run(unsigned int cpu)
{
	return __this_cpu_read(hrtimer_interrupts) !=
		__this_cpu_read(soft_lockup_hrtimer_cnt);
}

/*
 * The watchdog thread function - touches the timestamp.
 *
484
 * It only runs once every sample_period seconds (4 seconds by
485 486 487 488 489 490 491 492 493
 * 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().
 */
static void watchdog(unsigned int cpu)
{
	__this_cpu_write(soft_lockup_hrtimer_cnt,
			 __this_cpu_read(hrtimer_interrupts));
	__touch_watchdog();
494 495 496 497 498 499 500 501 502 503 504 505 506 507 508

	/*
	 * watchdog_nmi_enable() clears the NMI_WATCHDOG_ENABLED bit in the
	 * failure path. Check for failures that can occur asynchronously -
	 * for example, when CPUs are on-lined - and shut down the hardware
	 * perf event on each CPU accordingly.
	 *
	 * The only non-obvious place this bit can be cleared is through
	 * watchdog_nmi_enable(), so a pr_info() is placed there.  Placing a
	 * pr_info here would be too noisy as it would result in a message
	 * every few seconds if the hardlockup was disabled but the softlockup
	 * enabled.
	 */
	if (!(watchdog_enabled & NMI_WATCHDOG_ENABLED))
		watchdog_nmi_disable(cpu);
509
}
510

511
#ifdef CONFIG_HARDLOCKUP_DETECTOR
512 513 514 515 516 517 518
/*
 * 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;

519
static int watchdog_nmi_enable(unsigned int cpu)
520 521 522 523
{
	struct perf_event_attr *wd_attr;
	struct perf_event *event = per_cpu(watchdog_ev, cpu);

524 525 526
	/* nothing to do if the hard lockup detector is disabled */
	if (!(watchdog_enabled & NMI_WATCHDOG_ENABLED))
		goto out;
527

528 529 530 531 532 533 534 535 536
	/* 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;
537
	wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh);
538 539

	/* Try to register using hardware perf events */
540
	event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback, NULL);
541 542 543 544 545

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

546
	if (!IS_ERR(event)) {
547 548 549
		/* 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");
550 551 552
		goto out_save;
	}

553 554 555 556 557 558 559 560 561 562 563 564
	/*
	 * Disable the hard lockup detector if _any_ CPU fails to set up
	 * set up the hardware perf event. The watchdog() function checks
	 * the NMI_WATCHDOG_ENABLED bit periodically.
	 *
	 * The barriers are for syncing up watchdog_enabled across all the
	 * cpus, as clear_bit() does not use barriers.
	 */
	smp_mb__before_atomic();
	clear_bit(NMI_WATCHDOG_ENABLED_BIT, &watchdog_enabled);
	smp_mb__after_atomic();

565 566 567
	/* skip displaying the same error again */
	if (cpu > 0 && (PTR_ERR(event) == cpu0_err))
		return PTR_ERR(event);
568 569 570

	/* vary the KERN level based on the returned errno */
	if (PTR_ERR(event) == -EOPNOTSUPP)
571
		pr_info("disabled (cpu%i): not supported (no LAPIC?)\n", cpu);
572
	else if (PTR_ERR(event) == -ENOENT)
573
		pr_warn("disabled (cpu%i): hardware events not enabled\n",
574
			 cpu);
575
	else
576 577
		pr_err("disabled (cpu%i): unable to create perf event: %ld\n",
			cpu, PTR_ERR(event));
578 579 580

	pr_info("Shutting down hard lockup detector on all cpus\n");

581
	return PTR_ERR(event);
582 583 584 585 586 587 588 589 590 591

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

592
static void watchdog_nmi_disable(unsigned int cpu)
593 594 595 596 597 598 599 600 601 602
{
	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);
	}
603 604 605 606
	if (cpu == 0) {
		/* watchdog_nmi_enable() expects this to be zero initially. */
		cpu0_err = 0;
	}
607
}
608 609 610 611 612

void watchdog_nmi_enable_all(void)
{
	int cpu;

613 614 615 616
	mutex_lock(&watchdog_proc_mutex);

	if (!(watchdog_enabled & NMI_WATCHDOG_ENABLED))
		goto unlock;
617 618 619 620 621

	get_online_cpus();
	for_each_online_cpu(cpu)
		watchdog_nmi_enable(cpu);
	put_online_cpus();
622 623

unlock:
624
	mutex_unlock(&watchdog_proc_mutex);
625 626 627 628 629 630
}

void watchdog_nmi_disable_all(void)
{
	int cpu;

631 632
	mutex_lock(&watchdog_proc_mutex);

633
	if (!watchdog_running)
634
		goto unlock;
635 636 637 638 639

	get_online_cpus();
	for_each_online_cpu(cpu)
		watchdog_nmi_disable(cpu);
	put_online_cpus();
640 641 642

unlock:
	mutex_unlock(&watchdog_proc_mutex);
643
}
644
#else
645 646
static int watchdog_nmi_enable(unsigned int cpu) { return 0; }
static void watchdog_nmi_disable(unsigned int cpu) { return; }
647 648
void watchdog_nmi_enable_all(void) {}
void watchdog_nmi_disable_all(void) {}
649
#endif /* CONFIG_HARDLOCKUP_DETECTOR */
650

651 652 653 654 655 656 657 658 659 660 661
static struct smp_hotplug_thread watchdog_threads = {
	.store			= &softlockup_watchdog,
	.thread_should_run	= watchdog_should_run,
	.thread_fn		= watchdog,
	.thread_comm		= "watchdog/%u",
	.setup			= watchdog_enable,
	.cleanup		= watchdog_cleanup,
	.park			= watchdog_disable,
	.unpark			= watchdog_enable,
};

662 663
static void restart_watchdog_hrtimer(void *info)
{
664
	struct hrtimer *hrtimer = raw_cpu_ptr(&watchdog_hrtimer);
665 666 667 668 669 670 671 672 673 674 675 676 677 678
	int ret;

	/*
	 * No need to cancel and restart hrtimer if it is currently executing
	 * because it will reprogram itself with the new period now.
	 * We should never see it unqueued here because we are running per-cpu
	 * with interrupts disabled.
	 */
	ret = hrtimer_try_to_cancel(hrtimer);
	if (ret == 1)
		hrtimer_start(hrtimer, ns_to_ktime(sample_period),
				HRTIMER_MODE_REL_PINNED);
}

679
static void update_watchdog(int cpu)
680 681 682 683 684 685 686 687 688 689
{
	/*
	 * Make sure that perf event counter will adopt to a new
	 * sampling period. Updating the sampling period directly would
	 * be much nicer but we do not have an API for that now so
	 * let's use a big hammer.
	 * Hrtimer will adopt the new period on the next tick but this
	 * might be late already so we have to restart the timer as well.
	 */
	watchdog_nmi_disable(cpu);
690
	smp_call_function_single(cpu, restart_watchdog_hrtimer, NULL, 1);
691 692 693
	watchdog_nmi_enable(cpu);
}

694
static void update_watchdog_all_cpus(void)
695 696 697 698 699
{
	int cpu;

	get_online_cpus();
	for_each_online_cpu(cpu)
700
		update_watchdog(cpu);
701 702 703
	put_online_cpus();
}

704
static int watchdog_enable_all_cpus(void)
705
{
706
	int err = 0;
707

708
	if (!watchdog_running) {
709 710 711 712
		err = smpboot_register_percpu_thread(&watchdog_threads);
		if (err)
			pr_err("Failed to create watchdog threads, disabled\n");
		else
713
			watchdog_running = 1;
714 715 716 717 718 719
	} else {
		/*
		 * Enable/disable the lockup detectors or
		 * change the sample period 'on the fly'.
		 */
		update_watchdog_all_cpus();
720
	}
721 722

	return err;
723 724
}

725 726 727
/* prepare/enable/disable routines */
/* sysctl functions */
#ifdef CONFIG_SYSCTL
728 729
static void watchdog_disable_all_cpus(void)
{
730 731
	if (watchdog_running) {
		watchdog_running = 0;
732
		smpboot_unregister_percpu_thread(&watchdog_threads);
733
	}
734 735 736
}

/*
737 738 739 740 741 742 743 744 745 746 747 748 749 750
 * Update the run state of the lockup detectors.
 */
static int proc_watchdog_update(void)
{
	int err = 0;

	/*
	 * Watchdog threads won't be started if they are already active.
	 * The 'watchdog_running' variable in watchdog_*_all_cpus() takes
	 * care of this. If those threads are already active, the sample
	 * period will be updated and the lockup detectors will be enabled
	 * or disabled 'on the fly'.
	 */
	if (watchdog_enabled && watchdog_thresh)
751
		err = watchdog_enable_all_cpus();
752 753 754 755 756 757 758
	else
		watchdog_disable_all_cpus();

	return err;

}

759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823
/*
 * common function for watchdog, nmi_watchdog and soft_watchdog parameter
 *
 * caller             | table->data points to | 'which' contains the flag(s)
 * -------------------|-----------------------|-----------------------------
 * proc_watchdog      | watchdog_user_enabled | NMI_WATCHDOG_ENABLED or'ed
 *                    |                       | with SOFT_WATCHDOG_ENABLED
 * -------------------|-----------------------|-----------------------------
 * proc_nmi_watchdog  | nmi_watchdog_enabled  | NMI_WATCHDOG_ENABLED
 * -------------------|-----------------------|-----------------------------
 * proc_soft_watchdog | soft_watchdog_enabled | SOFT_WATCHDOG_ENABLED
 */
static int proc_watchdog_common(int which, struct ctl_table *table, int write,
				void __user *buffer, size_t *lenp, loff_t *ppos)
{
	int err, old, new;
	int *watchdog_param = (int *)table->data;

	mutex_lock(&watchdog_proc_mutex);

	/*
	 * If the parameter is being read return the state of the corresponding
	 * bit(s) in 'watchdog_enabled', else update 'watchdog_enabled' and the
	 * run state of the lockup detectors.
	 */
	if (!write) {
		*watchdog_param = (watchdog_enabled & which) != 0;
		err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
	} else {
		err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
		if (err)
			goto out;

		/*
		 * There is a race window between fetching the current value
		 * from 'watchdog_enabled' and storing the new value. During
		 * this race window, watchdog_nmi_enable() can sneak in and
		 * clear the NMI_WATCHDOG_ENABLED bit in 'watchdog_enabled'.
		 * The 'cmpxchg' detects this race and the loop retries.
		 */
		do {
			old = watchdog_enabled;
			/*
			 * If the parameter value is not zero set the
			 * corresponding bit(s), else clear it(them).
			 */
			if (*watchdog_param)
				new = old | which;
			else
				new = old & ~which;
		} while (cmpxchg(&watchdog_enabled, old, new) != old);

		/*
		 * Update the run state of the lockup detectors.
		 * Restore 'watchdog_enabled' on failure.
		 */
		err = proc_watchdog_update();
		if (err)
			watchdog_enabled = old;
	}
out:
	mutex_unlock(&watchdog_proc_mutex);
	return err;
}

824 825 826 827 828 829 830 831 832 833 834 835
/*
 * /proc/sys/kernel/watchdog
 */
int proc_watchdog(struct ctl_table *table, int write,
		  void __user *buffer, size_t *lenp, loff_t *ppos)
{
	return proc_watchdog_common(NMI_WATCHDOG_ENABLED|SOFT_WATCHDOG_ENABLED,
				    table, write, buffer, lenp, ppos);
}

/*
 * /proc/sys/kernel/nmi_watchdog
836
 */
837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852
int proc_nmi_watchdog(struct ctl_table *table, int write,
		      void __user *buffer, size_t *lenp, loff_t *ppos)
{
	return proc_watchdog_common(NMI_WATCHDOG_ENABLED,
				    table, write, buffer, lenp, ppos);
}

/*
 * /proc/sys/kernel/soft_watchdog
 */
int proc_soft_watchdog(struct ctl_table *table, int write,
			void __user *buffer, size_t *lenp, loff_t *ppos)
{
	return proc_watchdog_common(SOFT_WATCHDOG_ENABLED,
				    table, write, buffer, lenp, ppos);
}
853

854 855 856 857 858
/*
 * /proc/sys/kernel/watchdog_thresh
 */
int proc_watchdog_thresh(struct ctl_table *table, int write,
			 void __user *buffer, size_t *lenp, loff_t *ppos)
859
{
860
	int err, old;
861

862
	mutex_lock(&watchdog_proc_mutex);
863

864
	old = ACCESS_ONCE(watchdog_thresh);
865
	err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
866

867
	if (err || !write)
868
		goto out;
869

870
	/*
871 872
	 * Update the sample period.
	 * Restore 'watchdog_thresh' on failure.
873
	 */
874 875 876 877
	set_sample_period();
	err = proc_watchdog_update();
	if (err)
		watchdog_thresh = old;
878 879
out:
	mutex_unlock(&watchdog_proc_mutex);
880
	return err;
881 882 883
}
#endif /* CONFIG_SYSCTL */

884
void __init lockup_detector_init(void)
885
{
886
	set_sample_period();
887

888
	if (watchdog_enabled)
889
		watchdog_enable_all_cpus();
890
}