watchdog.c 29.4 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
#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
#include <linux/tick.h>
23 24

#include <asm/irq_regs.h>
25
#include <linux/kvm_para.h>
26
#include <linux/perf_event.h>
27
#include <linux/kthread.h>
28

29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
/*
 * 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)

46 47
static DEFINE_MUTEX(watchdog_proc_mutex);

48 49 50 51 52 53 54 55
#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;
56
int __read_mostly watchdog_thresh = 10;
57

58 59
#ifdef CONFIG_SMP
int __read_mostly sysctl_softlockup_all_cpu_backtrace;
60
int __read_mostly sysctl_hardlockup_all_cpu_backtrace;
61 62
#else
#define sysctl_softlockup_all_cpu_backtrace 0
63
#define sysctl_hardlockup_all_cpu_backtrace 0
64
#endif
65 66 67 68 69 70
static struct cpumask watchdog_cpumask __read_mostly;
unsigned long *watchdog_cpumask_bits = cpumask_bits(&watchdog_cpumask);

/* Helper for online, unparked cpus. */
#define for_each_watchdog_cpu(cpu) \
	for_each_cpu_and((cpu), cpu_online_mask, &watchdog_cpumask)
71

72 73 74 75 76
/*
 * The 'watchdog_running' variable is set to 1 when the watchdog threads
 * are registered/started and is set to 0 when the watchdog threads are
 * unregistered/stopped, so it is an indicator whether the threads exist.
 */
77
static int __read_mostly watchdog_running;
78 79 80 81 82 83 84 85 86 87 88 89 90 91
/*
 * If a subsystem has a need to deactivate the watchdog temporarily, it
 * can use the suspend/resume interface to achieve this. The content of
 * the 'watchdog_suspended' variable reflects this state. Existing threads
 * are parked/unparked by the lockup_detector_{suspend|resume} functions
 * (see comment blocks pertaining to those functions for further details).
 *
 * 'watchdog_suspended' also prevents threads from being registered/started
 * or unregistered/stopped via parameters in /proc/sys/kernel, so the state
 * of 'watchdog_running' cannot change while the watchdog is deactivated
 * temporarily (see related code in 'proc' handlers).
 */
static int __read_mostly watchdog_suspended;

92
static u64 __read_mostly sample_period;
93 94 95 96 97 98

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);
99 100
static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts);
static DEFINE_PER_CPU(unsigned long, soft_lockup_hrtimer_cnt);
101
static DEFINE_PER_CPU(struct task_struct *, softlockup_task_ptr_saved);
102
#ifdef CONFIG_HARDLOCKUP_DETECTOR
103 104
static DEFINE_PER_CPU(bool, hard_watchdog_warn);
static DEFINE_PER_CPU(bool, watchdog_nmi_touch);
105 106 107
static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved);
static DEFINE_PER_CPU(struct perf_event *, watchdog_ev);
#endif
108
static unsigned long soft_lockup_nmi_warn;
109 110 111 112 113

/* boot commands */
/*
 * Should we panic when a soft-lockup or hard-lockup occurs:
 */
114
#ifdef CONFIG_HARDLOCKUP_DETECTOR
115
unsigned int __read_mostly hardlockup_panic =
116
			CONFIG_BOOTPARAM_HARDLOCKUP_PANIC_VALUE;
117
static unsigned long hardlockup_allcpu_dumped;
118 119 120 121 122 123 124 125
/*
 * 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().
 */
126
void hardlockup_detector_disable(void)
127
{
128
	watchdog_enabled &= ~NMI_WATCHDOG_ENABLED;
129 130
}

131 132 133 134
static int __init hardlockup_panic_setup(char *str)
{
	if (!strncmp(str, "panic", 5))
		hardlockup_panic = 1;
135 136
	else if (!strncmp(str, "nopanic", 7))
		hardlockup_panic = 0;
137
	else if (!strncmp(str, "0", 1))
138 139 140
		watchdog_enabled &= ~NMI_WATCHDOG_ENABLED;
	else if (!strncmp(str, "1", 1))
		watchdog_enabled |= NMI_WATCHDOG_ENABLED;
141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158
	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)
{
159
	watchdog_enabled = 0;
160 161 162 163 164 165
	return 1;
}
__setup("nowatchdog", nowatchdog_setup);

static int __init nosoftlockup_setup(char *str)
{
166
	watchdog_enabled &= ~SOFT_WATCHDOG_ENABLED;
167 168 169
	return 1;
}
__setup("nosoftlockup", nosoftlockup_setup);
170

171 172 173 174 175 176 177 178
#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);
179 180 181 182 183 184 185
static int __init hardlockup_all_cpu_backtrace_setup(char *str)
{
	sysctl_hardlockup_all_cpu_backtrace =
		!!simple_strtol(str, NULL, 0);
	return 1;
}
__setup("hardlockup_all_cpu_backtrace=", hardlockup_all_cpu_backtrace_setup);
186
#endif
187

188 189 190 191 192 193 194
/*
 * 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.
 */
195
static int get_softlockup_thresh(void)
196 197 198
{
	return watchdog_thresh * 2;
}
199 200 201 202 203 204

/*
 * 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.
 */
205
static unsigned long get_timestamp(void)
206
{
207
	return running_clock() >> 30LL;  /* 2^30 ~= 10^9 */
208 209
}

210
static void set_sample_period(void)
211 212
{
	/*
213
	 * convert watchdog_thresh from seconds to ns
214 215 216 217
	 * 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
218
	 */
219
	sample_period = get_softlockup_thresh() * ((u64)NSEC_PER_SEC / 5);
220 221 222 223 224
}

/* Commands for resetting the watchdog */
static void __touch_watchdog(void)
{
225
	__this_cpu_write(watchdog_touch_ts, get_timestamp());
226 227
}

228 229 230 231 232 233 234 235 236
/**
 * touch_softlockup_watchdog_sched - touch watchdog on scheduler stalls
 *
 * Call when the scheduler may have stalled for legitimate reasons
 * preventing the watchdog task from executing - e.g. the scheduler
 * entering idle state.  This should only be used for scheduler events.
 * Use touch_softlockup_watchdog() for everything else.
 */
void touch_softlockup_watchdog_sched(void)
237
{
238 239 240 241 242
	/*
	 * 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);
243
}
244 245 246 247 248

void touch_softlockup_watchdog(void)
{
	touch_softlockup_watchdog_sched();
}
249
EXPORT_SYMBOL(touch_softlockup_watchdog);
250

251
void touch_all_softlockup_watchdogs(void)
252 253 254 255 256 257 258 259
{
	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
	 */
260
	for_each_watchdog_cpu(cpu)
261 262 263
		per_cpu(watchdog_touch_ts, cpu) = 0;
}

264
#ifdef CONFIG_HARDLOCKUP_DETECTOR
265 266
void touch_nmi_watchdog(void)
{
267 268 269 270 271 272 273
	/*
	 * 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.
	 */
274
	raw_cpu_write(watchdog_nmi_touch, true);
275
	touch_softlockup_watchdog();
276 277 278
}
EXPORT_SYMBOL(touch_nmi_watchdog);

279 280
#endif

281 282
void touch_softlockup_watchdog_sync(void)
{
283 284
	__this_cpu_write(softlockup_touch_sync, true);
	__this_cpu_write(watchdog_touch_ts, 0);
285 286
}

287
#ifdef CONFIG_HARDLOCKUP_DETECTOR
288
/* watchdog detector functions */
289
static bool is_hardlockup(void)
290
{
291
	unsigned long hrint = __this_cpu_read(hrtimer_interrupts);
292

293
	if (__this_cpu_read(hrtimer_interrupts_saved) == hrint)
294
		return true;
295

296
	__this_cpu_write(hrtimer_interrupts_saved, hrint);
297
	return false;
298 299 300
}
#endif

301
static int is_softlockup(unsigned long touch_ts)
302
{
303
	unsigned long now = get_timestamp();
304

305
	if ((watchdog_enabled & SOFT_WATCHDOG_ENABLED) && watchdog_thresh){
306 307 308 309
		/* Warn about unreasonable delays. */
		if (time_after(now, touch_ts + get_softlockup_thresh()))
			return now - touch_ts;
	}
310 311 312
	return 0;
}

313
#ifdef CONFIG_HARDLOCKUP_DETECTOR
314

315 316 317 318 319 320 321 322 323
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 */
324
static void watchdog_overflow_callback(struct perf_event *event,
325 326 327
		 struct perf_sample_data *data,
		 struct pt_regs *regs)
{
328 329 330
	/* Ensure the watchdog never gets throttled */
	event->hw.interrupts = 0;

331 332
	if (__this_cpu_read(watchdog_nmi_touch) == true) {
		__this_cpu_write(watchdog_nmi_touch, false);
333 334 335 336 337 338 339 340 341
		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
	 */
342 343
	if (is_hardlockup()) {
		int this_cpu = smp_processor_id();
344
		struct pt_regs *regs = get_irq_regs();
345

346
		/* only print hardlockups once */
347
		if (__this_cpu_read(hard_watchdog_warn) == true)
348 349
			return;

350 351 352 353 354
		pr_emerg("Watchdog detected hard LOCKUP on cpu %d", this_cpu);
		print_modules();
		print_irqtrace_events(current);
		if (regs)
			show_regs(regs);
355
		else
356 357 358 359 360 361 362 363 364 365 366 367
			dump_stack();

		/*
		 * Perform all-CPU dump only once to avoid multiple hardlockups
		 * generating interleaving traces
		 */
		if (sysctl_hardlockup_all_cpu_backtrace &&
				!test_and_set_bit(0, &hardlockup_allcpu_dumped))
			trigger_allbutself_cpu_backtrace();

		if (hardlockup_panic)
			panic("Hard LOCKUP");
368

369
		__this_cpu_write(hard_watchdog_warn, true);
370 371 372
		return;
	}

373
	__this_cpu_write(hard_watchdog_warn, false);
374 375
	return;
}
376 377
#endif /* CONFIG_HARDLOCKUP_DETECTOR */

378 379
static void watchdog_interrupt_count(void)
{
380
	__this_cpu_inc(hrtimer_interrupts);
381
}
382 383 384

static int watchdog_nmi_enable(unsigned int cpu);
static void watchdog_nmi_disable(unsigned int cpu);
385

386 387 388
static int watchdog_enable_all_cpus(void);
static void watchdog_disable_all_cpus(void);

389 390 391
/* watchdog kicker functions */
static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
{
392
	unsigned long touch_ts = __this_cpu_read(watchdog_touch_ts);
393 394
	struct pt_regs *regs = get_irq_regs();
	int duration;
395
	int softlockup_all_cpu_backtrace = sysctl_softlockup_all_cpu_backtrace;
396 397 398 399 400

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

	/* kick the softlockup detector */
401
	wake_up_process(__this_cpu_read(softlockup_watchdog));
402 403

	/* .. and repeat */
404
	hrtimer_forward_now(hrtimer, ns_to_ktime(sample_period));
405 406

	if (touch_ts == 0) {
407
		if (unlikely(__this_cpu_read(softlockup_touch_sync))) {
408 409 410 411
			/*
			 * If the time stamp was touched atomically
			 * make sure the scheduler tick is up to date.
			 */
412
			__this_cpu_write(softlockup_touch_sync, false);
413 414
			sched_clock_tick();
		}
415 416 417

		/* Clear the guest paused flag on watchdog reset */
		kvm_check_and_clear_guest_paused();
418 419 420 421 422 423 424 425 426 427
		__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
	 */
428
	duration = is_softlockup(touch_ts);
429
	if (unlikely(duration)) {
430 431 432 433 434 435 436 437
		/*
		 * 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;

438
		/* only warn once */
439 440 441 442 443 444 445 446 447 448 449 450 451 452
		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();
			}
453
			return HRTIMER_RESTART;
454
		}
455

456 457 458 459 460 461 462 463 464 465 466
		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;
			}
		}

467
		pr_emerg("BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
468
			smp_processor_id(), duration,
469
			current->comm, task_pid_nr(current));
470
		__this_cpu_write(softlockup_task_ptr_saved, current);
471 472 473 474 475 476 477
		print_modules();
		print_irqtrace_events(current);
		if (regs)
			show_regs(regs);
		else
			dump_stack();

478 479 480 481 482 483 484 485 486 487 488
		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
489
		add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
490 491
		if (softlockup_panic)
			panic("softlockup: hung tasks");
492
		__this_cpu_write(soft_watchdog_warn, true);
493
	} else
494
		__this_cpu_write(soft_watchdog_warn, false);
495 496 497 498

	return HRTIMER_RESTART;
}

499 500 501
static void watchdog_set_prio(unsigned int policy, unsigned int prio)
{
	struct sched_param param = { .sched_priority = prio };
502

503 504 505 506
	sched_setscheduler(current, policy, &param);
}

static void watchdog_enable(unsigned int cpu)
507
{
508
	struct hrtimer *hrtimer = raw_cpu_ptr(&watchdog_hrtimer);
509

510 511 512 513
	/* kick off the timer for the hardlockup detector */
	hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
	hrtimer->function = watchdog_timer_fn;

514 515
	/* Enable the perf event */
	watchdog_nmi_enable(cpu);
516 517

	/* done here because hrtimer_start can only pin to smp_processor_id() */
518
	hrtimer_start(hrtimer, ns_to_ktime(sample_period),
519 520
		      HRTIMER_MODE_REL_PINNED);

521 522 523 524
	/* initialize timestamp */
	watchdog_set_prio(SCHED_FIFO, MAX_RT_PRIO - 1);
	__touch_watchdog();
}
525

526 527
static void watchdog_disable(unsigned int cpu)
{
528
	struct hrtimer *hrtimer = raw_cpu_ptr(&watchdog_hrtimer);
529

530 531 532 533
	watchdog_set_prio(SCHED_NORMAL, 0);
	hrtimer_cancel(hrtimer);
	/* disable the perf event */
	watchdog_nmi_disable(cpu);
534 535
}

536 537 538 539 540
static void watchdog_cleanup(unsigned int cpu, bool online)
{
	watchdog_disable(cpu);
}

541 542 543 544 545 546 547 548 549
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.
 *
550
 * It only runs once every sample_period seconds (4 seconds by
551 552 553 554 555 556 557 558 559
 * 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();
560 561 562 563 564 565 566 567 568 569 570 571 572 573 574

	/*
	 * 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);
575
}
576

577
#ifdef CONFIG_HARDLOCKUP_DETECTOR
578 579 580 581 582 583 584
/*
 * 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;

585
static int watchdog_nmi_enable(unsigned int cpu)
586 587 588 589
{
	struct perf_event_attr *wd_attr;
	struct perf_event *event = per_cpu(watchdog_ev, cpu);

590 591 592
	/* nothing to do if the hard lockup detector is disabled */
	if (!(watchdog_enabled & NMI_WATCHDOG_ENABLED))
		goto out;
593

594 595 596 597 598 599 600 601 602
	/* 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;
603
	wd_attr->sample_period = hw_nmi_get_sample_period(watchdog_thresh);
604 605

	/* Try to register using hardware perf events */
606
	event = perf_event_create_kernel_counter(wd_attr, cpu, NULL, watchdog_overflow_callback, NULL);
607 608 609 610 611

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

612
	if (!IS_ERR(event)) {
613 614 615
		/* 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");
616 617 618
		goto out_save;
	}

619 620 621 622 623 624 625 626 627 628 629 630
	/*
	 * 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();

631 632 633
	/* skip displaying the same error again */
	if (cpu > 0 && (PTR_ERR(event) == cpu0_err))
		return PTR_ERR(event);
634 635 636

	/* vary the KERN level based on the returned errno */
	if (PTR_ERR(event) == -EOPNOTSUPP)
637
		pr_info("disabled (cpu%i): not supported (no LAPIC?)\n", cpu);
638
	else if (PTR_ERR(event) == -ENOENT)
639
		pr_warn("disabled (cpu%i): hardware events not enabled\n",
640
			 cpu);
641
	else
642 643
		pr_err("disabled (cpu%i): unable to create perf event: %ld\n",
			cpu, PTR_ERR(event));
644 645 646

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

647
	return PTR_ERR(event);
648 649 650 651 652 653 654 655 656 657

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

658
static void watchdog_nmi_disable(unsigned int cpu)
659 660 661 662 663 664 665 666 667 668
{
	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);
	}
669 670 671 672
	if (cpu == 0) {
		/* watchdog_nmi_enable() expects this to be zero initially. */
		cpu0_err = 0;
	}
673
}
674

675
#else
676 677
static int watchdog_nmi_enable(unsigned int cpu) { return 0; }
static void watchdog_nmi_disable(unsigned int cpu) { return; }
678
#endif /* CONFIG_HARDLOCKUP_DETECTOR */
679

680 681 682 683 684 685 686 687 688 689 690
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,
};

691 692
/*
 * park all watchdog threads that are specified in 'watchdog_cpumask'
693 694 695 696 697 698
 *
 * This function returns an error if kthread_park() of a watchdog thread
 * fails. In this situation, the watchdog threads of some CPUs can already
 * be parked and the watchdog threads of other CPUs can still be runnable.
 * Callers are expected to handle this special condition as appropriate in
 * their context.
699 700 701
 *
 * This function may only be called in a context that is protected against
 * races with CPU hotplug - for example, via get_online_cpus().
702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717
 */
static int watchdog_park_threads(void)
{
	int cpu, ret = 0;

	for_each_watchdog_cpu(cpu) {
		ret = kthread_park(per_cpu(softlockup_watchdog, cpu));
		if (ret)
			break;
	}

	return ret;
}

/*
 * unpark all watchdog threads that are specified in 'watchdog_cpumask'
718 719 720
 *
 * This function may only be called in a context that is protected against
 * races with CPU hotplug - for example, via get_online_cpus().
721 722 723 724 725 726 727 728 729
 */
static void watchdog_unpark_threads(void)
{
	int cpu;

	for_each_watchdog_cpu(cpu)
		kthread_unpark(per_cpu(softlockup_watchdog, cpu));
}

730 731 732
/*
 * Suspend the hard and soft lockup detector by parking the watchdog threads.
 */
733
int lockup_detector_suspend(void)
734 735 736
{
	int ret = 0;

737
	get_online_cpus();
738 739 740 741 742 743
	mutex_lock(&watchdog_proc_mutex);
	/*
	 * Multiple suspend requests can be active in parallel (counted by
	 * the 'watchdog_suspended' variable). If the watchdog threads are
	 * running, the first caller takes care that they will be parked.
	 * The state of 'watchdog_running' cannot change while a suspend
744
	 * request is active (see related code in 'proc' handlers).
745 746 747 748 749 750
	 */
	if (watchdog_running && !watchdog_suspended)
		ret = watchdog_park_threads();

	if (ret == 0)
		watchdog_suspended++;
751 752 753 754 755
	else {
		watchdog_disable_all_cpus();
		pr_err("Failed to suspend lockup detectors, disabled\n");
		watchdog_enabled = 0;
	}
756 757 758 759 760 761 762 763 764

	mutex_unlock(&watchdog_proc_mutex);

	return ret;
}

/*
 * Resume the hard and soft lockup detector by unparking the watchdog threads.
 */
765
void lockup_detector_resume(void)
766 767 768 769 770 771 772 773 774 775 776 777
{
	mutex_lock(&watchdog_proc_mutex);

	watchdog_suspended--;
	/*
	 * The watchdog threads are unparked if they were previously running
	 * and if there is no more active suspend request.
	 */
	if (watchdog_running && !watchdog_suspended)
		watchdog_unpark_threads();

	mutex_unlock(&watchdog_proc_mutex);
778
	put_online_cpus();
779 780
}

781
static int update_watchdog_all_cpus(void)
782
{
783 784 785 786 787 788
	int ret;

	ret = watchdog_park_threads();
	if (ret)
		return ret;

789
	watchdog_unpark_threads();
790 791

	return 0;
792 793
}

794
static int watchdog_enable_all_cpus(void)
795
{
796
	int err = 0;
797

798
	if (!watchdog_running) {
799 800
		err = smpboot_register_percpu_thread_cpumask(&watchdog_threads,
							     &watchdog_cpumask);
801 802
		if (err)
			pr_err("Failed to create watchdog threads, disabled\n");
803
		else
804
			watchdog_running = 1;
805 806 807 808 809
	} else {
		/*
		 * Enable/disable the lockup detectors or
		 * change the sample period 'on the fly'.
		 */
810 811 812 813 814 815
		err = update_watchdog_all_cpus();

		if (err) {
			watchdog_disable_all_cpus();
			pr_err("Failed to update lockup detectors, disabled\n");
		}
816
	}
817

818 819 820
	if (err)
		watchdog_enabled = 0;

821
	return err;
822 823 824 825
}

static void watchdog_disable_all_cpus(void)
{
826 827
	if (watchdog_running) {
		watchdog_running = 0;
828
		smpboot_unregister_percpu_thread(&watchdog_threads);
829
	}
830 831
}

832 833
#ifdef CONFIG_SYSCTL

834
/*
835 836 837 838 839 840 841 842 843 844 845 846 847 848
 * 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)
849
		err = watchdog_enable_all_cpus();
850 851 852 853 854 855 856
	else
		watchdog_disable_all_cpus();

	return err;

}

857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874
/*
 * 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;

875
	get_online_cpus();
876 877
	mutex_lock(&watchdog_proc_mutex);

878 879 880 881 882 883
	if (watchdog_suspended) {
		/* no parameter changes allowed while watchdog is suspended */
		err = -EAGAIN;
		goto out;
	}

884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916
	/*
	 * 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);

		/*
917 918 919 920 921
		 * Update the run state of the lockup detectors. There is _no_
		 * need to check the value returned by proc_watchdog_update()
		 * and to restore the previous value of 'watchdog_enabled' as
		 * both lockup detectors are disabled if proc_watchdog_update()
		 * returns an error.
922 923 924 925 926
		 */
		err = proc_watchdog_update();
	}
out:
	mutex_unlock(&watchdog_proc_mutex);
927
	put_online_cpus();
928 929 930
	return err;
}

931 932 933 934 935 936 937 938 939 940 941 942
/*
 * /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
943
 */
944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959
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);
}
960

961 962 963 964 965
/*
 * /proc/sys/kernel/watchdog_thresh
 */
int proc_watchdog_thresh(struct ctl_table *table, int write,
			 void __user *buffer, size_t *lenp, loff_t *ppos)
966
{
967
	int err, old;
968

969
	get_online_cpus();
970
	mutex_lock(&watchdog_proc_mutex);
971

972 973 974 975 976 977
	if (watchdog_suspended) {
		/* no parameter changes allowed while watchdog is suspended */
		err = -EAGAIN;
		goto out;
	}

978
	old = ACCESS_ONCE(watchdog_thresh);
979
	err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
980

981
	if (err || !write)
982
		goto out;
983

984
	/*
985
	 * Update the sample period. Restore on failure.
986
	 */
987 988
	set_sample_period();
	err = proc_watchdog_update();
989
	if (err) {
990
		watchdog_thresh = old;
991 992
		set_sample_period();
	}
993 994
out:
	mutex_unlock(&watchdog_proc_mutex);
995
	put_online_cpus();
996
	return err;
997
}
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009

/*
 * The cpumask is the mask of possible cpus that the watchdog can run
 * on, not the mask of cpus it is actually running on.  This allows the
 * user to specify a mask that will include cpus that have not yet
 * been brought online, if desired.
 */
int proc_watchdog_cpumask(struct ctl_table *table, int write,
			  void __user *buffer, size_t *lenp, loff_t *ppos)
{
	int err;

1010
	get_online_cpus();
1011
	mutex_lock(&watchdog_proc_mutex);
1012 1013 1014 1015 1016 1017 1018

	if (watchdog_suspended) {
		/* no parameter changes allowed while watchdog is suspended */
		err = -EAGAIN;
		goto out;
	}

1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
	err = proc_do_large_bitmap(table, write, buffer, lenp, ppos);
	if (!err && write) {
		/* Remove impossible cpus to keep sysctl output cleaner. */
		cpumask_and(&watchdog_cpumask, &watchdog_cpumask,
			    cpu_possible_mask);

		if (watchdog_running) {
			/*
			 * Failure would be due to being unable to allocate
			 * a temporary cpumask, so we are likely not in a
			 * position to do much else to make things better.
			 */
			if (smpboot_update_cpumask_percpu_thread(
				    &watchdog_threads, &watchdog_cpumask) != 0)
				pr_err("cpumask update failed\n");
		}
	}
1036
out:
1037
	mutex_unlock(&watchdog_proc_mutex);
1038
	put_online_cpus();
1039 1040 1041
	return err;
}

1042 1043
#endif /* CONFIG_SYSCTL */

1044
void __init lockup_detector_init(void)
1045
{
1046
	set_sample_period();
1047

1048 1049
#ifdef CONFIG_NO_HZ_FULL
	if (tick_nohz_full_enabled()) {
1050 1051
		pr_info("Disabling watchdog on nohz_full cores by default\n");
		cpumask_copy(&watchdog_cpumask, housekeeping_mask);
1052 1053 1054 1055 1056 1057
	} else
		cpumask_copy(&watchdog_cpumask, cpu_possible_mask);
#else
	cpumask_copy(&watchdog_cpumask, cpu_possible_mask);
#endif

1058
	if (watchdog_enabled)
1059
		watchdog_enable_all_cpus();
1060
}