aboutsummaryrefslogtreecommitdiff
path: root/sys/kern/subr_lock.c
blob: 36ba9125e17650a1640febfb77330767737aabd8 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
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
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
/*-
 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
 *
 * Copyright (c) 2006 John Baldwin <jhb@FreeBSD.org>
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

/*
 * This module holds the global variables and functions used to maintain
 * lock_object structures.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include "opt_ddb.h"
#include "opt_mprof.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/lock_profile.h>
#include <sys/malloc.h>
#include <sys/mutex.h>
#include <sys/pcpu.h>
#include <sys/proc.h>
#include <sys/sbuf.h>
#include <sys/sched.h>
#include <sys/smp.h>
#include <sys/sysctl.h>

#ifdef DDB
#include <ddb/ddb.h>
#endif

#include <machine/cpufunc.h>

/*
 * Uncomment to validate that spin argument to acquire/release routines matches
 * the flag in the lock
 */
//#define	LOCK_PROFILING_DEBUG_SPIN

SDT_PROVIDER_DEFINE(lock);
SDT_PROBE_DEFINE1(lock, , , starvation, "u_int");

CTASSERT(LOCK_CLASS_MAX == 15);

struct lock_class *lock_classes[LOCK_CLASS_MAX + 1] = {
	&lock_class_mtx_spin,
	&lock_class_mtx_sleep,
	&lock_class_sx,
	&lock_class_rm,
	&lock_class_rm_sleepable,
	&lock_class_rw,
	&lock_class_lockmgr,
};

void
lock_init(struct lock_object *lock, struct lock_class *class, const char *name,
    const char *type, int flags)
{
	int i;

	/* Check for double-init and zero object. */
	KASSERT(flags & LO_NEW || !lock_initialized(lock),
	    ("lock \"%s\" %p already initialized", name, lock));

	/* Look up lock class to find its index. */
	for (i = 0; i < LOCK_CLASS_MAX; i++)
		if (lock_classes[i] == class) {
			lock->lo_flags = i << LO_CLASSSHIFT;
			break;
		}
	KASSERT(i < LOCK_CLASS_MAX, ("unknown lock class %p", class));

	/* Initialize the lock object. */
	lock->lo_name = name;
	lock->lo_flags |= flags | LO_INITIALIZED;
	LOCK_LOG_INIT(lock, 0);
	WITNESS_INIT(lock, (type != NULL) ? type : name);
}

void
lock_destroy(struct lock_object *lock)
{

	KASSERT(lock_initialized(lock), ("lock %p is not initialized", lock));
	WITNESS_DESTROY(lock);
	LOCK_LOG_DESTROY(lock, 0);
	lock->lo_flags &= ~LO_INITIALIZED;
}

static SYSCTL_NODE(_debug, OID_AUTO, lock, CTLFLAG_RD | CTLFLAG_MPSAFE, NULL,
    "lock debugging");
static SYSCTL_NODE(_debug_lock, OID_AUTO, delay,
    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL,
    "lock delay");

static u_int __read_mostly starvation_limit = 131072;
SYSCTL_INT(_debug_lock_delay, OID_AUTO, starvation_limit, CTLFLAG_RW,
    &starvation_limit, 0, "");

static u_int __read_mostly restrict_starvation = 0;
SYSCTL_INT(_debug_lock_delay, OID_AUTO, restrict_starvation, CTLFLAG_RW,
    &restrict_starvation, 0, "");

void
lock_delay(struct lock_delay_arg *la)
{
	struct lock_delay_config *lc = la->config;
	u_short i;

	la->delay <<= 1;
	if (__predict_false(la->delay > lc->max))
		la->delay = lc->max;

	for (i = la->delay; i > 0; i--)
		cpu_spinwait();

	la->spin_cnt += la->delay;
	if (__predict_false(la->spin_cnt > starvation_limit)) {
		SDT_PROBE1(lock, , , starvation, la->delay);
		if (restrict_starvation)
			la->delay = lc->base;
	}
}

static u_int
lock_roundup_2(u_int val)
{
	u_int res;

	for (res = 1; res <= val; res <<= 1)
		continue;

	return (res);
}

void
lock_delay_default_init(struct lock_delay_config *lc)
{

	lc->base = 1;
	lc->max = lock_roundup_2(mp_ncpus) * 256;
	if (lc->max > 32678)
		lc->max = 32678;
}

struct lock_delay_config __read_frequently locks_delay;
u_short __read_frequently locks_delay_retries;
u_short __read_frequently locks_delay_loops;

SYSCTL_U16(_debug_lock, OID_AUTO, delay_base, CTLFLAG_RW, &locks_delay.base,
    0, "");
SYSCTL_U16(_debug_lock, OID_AUTO, delay_max, CTLFLAG_RW, &locks_delay.max,
    0, "");
SYSCTL_U16(_debug_lock, OID_AUTO, delay_retries, CTLFLAG_RW, &locks_delay_retries,
    0, "");
SYSCTL_U16(_debug_lock, OID_AUTO, delay_loops, CTLFLAG_RW, &locks_delay_loops,
    0, "");

static void
locks_delay_init(void *arg __unused)
{

	lock_delay_default_init(&locks_delay);
	locks_delay_retries = 10;
	locks_delay_loops = max(10000, locks_delay.max);
}
LOCK_DELAY_SYSINIT(locks_delay_init);

#ifdef DDB
DB_SHOW_COMMAND(lock, db_show_lock)
{
	struct lock_object *lock;
	struct lock_class *class;

	if (!have_addr)
		return;
	lock = (struct lock_object *)addr;
	if (LO_CLASSINDEX(lock) > LOCK_CLASS_MAX) {
		db_printf("Unknown lock class: %d\n", LO_CLASSINDEX(lock));
		return;
	}
	class = LOCK_CLASS(lock);
	db_printf(" class: %s\n", class->lc_name);
	db_printf(" name: %s\n", lock->lo_name);
	class->lc_ddb_show(lock);
}
#endif

#ifdef LOCK_PROFILING

/*
 * One object per-thread for each lock the thread owns.  Tracks individual
 * lock instances.
 */
struct lock_profile_object {
	LIST_ENTRY(lock_profile_object) lpo_link;
	struct lock_object *lpo_obj;
	const char	*lpo_file;
	int		lpo_line;
	uint16_t	lpo_ref;
	uint16_t	lpo_cnt;
	uint64_t	lpo_acqtime;
	uint64_t	lpo_waittime;
	u_int		lpo_contest_locking;
};

/*
 * One lock_prof for each (file, line, lock object) triple.
 */
struct lock_prof {
	SLIST_ENTRY(lock_prof) link;
	struct lock_class *class;
	const char	*file;
	const char	*name;
	int		line;
	int		ticks;
	uintmax_t	cnt_wait_max;
	uintmax_t	cnt_max;
	uintmax_t	cnt_tot;
	uintmax_t	cnt_wait;
	uintmax_t	cnt_cur;
	uintmax_t	cnt_contest_locking;
};

SLIST_HEAD(lphead, lock_prof);

#define	LPROF_HASH_SIZE		4096
#define	LPROF_HASH_MASK		(LPROF_HASH_SIZE - 1)
#define	LPROF_CACHE_SIZE	4096

/*
 * Array of objects and profs for each type of object for each cpu.  Spinlocks
 * are handled separately because a thread may be preempted and acquire a
 * spinlock while in the lock profiling code of a non-spinlock.  In this way
 * we only need a critical section to protect the per-cpu lists.
 */
struct lock_prof_type {
	struct lphead		lpt_lpalloc;
	struct lpohead		lpt_lpoalloc;
	struct lphead		lpt_hash[LPROF_HASH_SIZE];
	struct lock_prof	lpt_prof[LPROF_CACHE_SIZE];
	struct lock_profile_object lpt_objs[LPROF_CACHE_SIZE];
};

struct lock_prof_cpu {
	struct lock_prof_type	lpc_types[2]; /* One for spin one for other. */
};

DPCPU_DEFINE_STATIC(struct lock_prof_cpu, lp);
#define	LP_CPU_SELF	(DPCPU_PTR(lp))
#define	LP_CPU(cpu)	(DPCPU_ID_PTR((cpu), lp))

volatile int __read_mostly lock_prof_enable;
int __read_mostly lock_contested_only;
static volatile int lock_prof_resetting;

#define LPROF_SBUF_SIZE		256

static int lock_prof_rejected;
static int lock_prof_skipspin;

#ifndef USE_CPU_NANOSECONDS
uint64_t
nanoseconds(void)
{
	struct bintime bt;
	uint64_t ns;

	binuptime(&bt);
	/* From bintime2timespec */
	ns = bt.sec * (uint64_t)1000000000;
	ns += ((uint64_t)1000000000 * (uint32_t)(bt.frac >> 32)) >> 32;
	return (ns);
}
#endif

static void
lock_prof_init_type(struct lock_prof_type *type)
{
	int i;

	SLIST_INIT(&type->lpt_lpalloc);
	LIST_INIT(&type->lpt_lpoalloc);
	for (i = 0; i < LPROF_CACHE_SIZE; i++) {
		SLIST_INSERT_HEAD(&type->lpt_lpalloc, &type->lpt_prof[i],
		    link);
		LIST_INSERT_HEAD(&type->lpt_lpoalloc, &type->lpt_objs[i],
		    lpo_link);
	}
}

static void
lock_prof_init(void *arg)
{
	int cpu;

	CPU_FOREACH(cpu) {
		lock_prof_init_type(&LP_CPU(cpu)->lpc_types[0]);
		lock_prof_init_type(&LP_CPU(cpu)->lpc_types[1]);
	}
}
SYSINIT(lockprof, SI_SUB_SMP, SI_ORDER_ANY, lock_prof_init, NULL);

static void
lock_prof_reset_wait(void)
{

	/*
	 * Spin relinquishing our cpu so that quiesce_all_cpus may
	 * complete.
	 */
	while (lock_prof_resetting)
		sched_relinquish(curthread);
}

static void
lock_prof_reset(void)
{
	struct lock_prof_cpu *lpc;
	int enabled, i, cpu;

	/*
	 * We not only race with acquiring and releasing locks but also
	 * thread exit.  To be certain that threads exit without valid head
	 * pointers they must see resetting set before enabled is cleared.
	 * Otherwise a lock may not be removed from a per-thread list due
	 * to disabled being set but not wait for reset() to remove it below.
	 */
	atomic_store_rel_int(&lock_prof_resetting, 1);
	enabled = lock_prof_enable;
	lock_prof_enable = 0;
	/*
	 * This both publishes lock_prof_enable as disabled and makes sure
	 * everyone else reads it if they are not far enough. We wait for the
	 * rest down below.
	 */
	cpus_fence_seq_cst();
	quiesce_all_critical();
	/*
	 * Some objects may have migrated between CPUs.  Clear all links
	 * before we zero the structures.  Some items may still be linked
	 * into per-thread lists as well.
	 */
	CPU_FOREACH(cpu) {
		lpc = LP_CPU(cpu);
		for (i = 0; i < LPROF_CACHE_SIZE; i++) {
			LIST_REMOVE(&lpc->lpc_types[0].lpt_objs[i], lpo_link);
			LIST_REMOVE(&lpc->lpc_types[1].lpt_objs[i], lpo_link);
		}
	}
	CPU_FOREACH(cpu) {
		lpc = LP_CPU(cpu);
		bzero(lpc, sizeof(*lpc));
		lock_prof_init_type(&lpc->lpc_types[0]);
		lock_prof_init_type(&lpc->lpc_types[1]);
	}
	/*
	 * Paired with the fence from cpus_fence_seq_cst()
	 */
	atomic_store_rel_int(&lock_prof_resetting, 0);
	lock_prof_enable = enabled;
}

static void
lock_prof_output(struct lock_prof *lp, struct sbuf *sb)
{
	const char *p;

	for (p = lp->file; p != NULL && strncmp(p, "../", 3) == 0; p += 3);
	sbuf_printf(sb,
	    "%8ju %9ju %11ju %11ju %11ju %6ju %6ju %2ju %6ju %s:%d (%s:%s)\n",
	    lp->cnt_max / 1000, lp->cnt_wait_max / 1000, lp->cnt_tot / 1000,
	    lp->cnt_wait / 1000, lp->cnt_cur,
	    lp->cnt_cur == 0 ? (uintmax_t)0 :
	    lp->cnt_tot / (lp->cnt_cur * 1000),
	    lp->cnt_cur == 0 ? (uintmax_t)0 :
	    lp->cnt_wait / (lp->cnt_cur * 1000),
	    (uintmax_t)0, lp->cnt_contest_locking,
	    p, lp->line, lp->class->lc_name, lp->name);
}

static void
lock_prof_sum(struct lock_prof *match, struct lock_prof *dst, int hash,
    int spin, int t)
{
	struct lock_prof_type *type;
	struct lock_prof *l;
	int cpu;

	dst->file = match->file;
	dst->line = match->line;
	dst->class = match->class;
	dst->name = match->name;

	CPU_FOREACH(cpu) {
		type = &LP_CPU(cpu)->lpc_types[spin];
		SLIST_FOREACH(l, &type->lpt_hash[hash], link) {
			if (l->ticks == t)
				continue;
			if (l->file != match->file || l->line != match->line ||
			    l->name != match->name)
				continue;
			l->ticks = t;
			if (l->cnt_max > dst->cnt_max)
				dst->cnt_max = l->cnt_max;
			if (l->cnt_wait_max > dst->cnt_wait_max)
				dst->cnt_wait_max = l->cnt_wait_max;
			dst->cnt_tot += l->cnt_tot;
			dst->cnt_wait += l->cnt_wait;
			dst->cnt_cur += l->cnt_cur;
			dst->cnt_contest_locking += l->cnt_contest_locking;
		}
	}
}

static void
lock_prof_type_stats(struct lock_prof_type *type, struct sbuf *sb, int spin,
    int t)
{
	struct lock_prof *l;
	int i;

	for (i = 0; i < LPROF_HASH_SIZE; ++i) {
		SLIST_FOREACH(l, &type->lpt_hash[i], link) {
			struct lock_prof lp = {};

			if (l->ticks == t)
				continue;
			lock_prof_sum(l, &lp, i, spin, t);
			lock_prof_output(&lp, sb);
		}
	}
}

static int
dump_lock_prof_stats(SYSCTL_HANDLER_ARGS)
{
	struct sbuf *sb;
	int error, cpu, t;
	int enabled;

	error = sysctl_wire_old_buffer(req, 0);
	if (error != 0)
		return (error);
	sb = sbuf_new_for_sysctl(NULL, NULL, LPROF_SBUF_SIZE, req);
	sbuf_printf(sb, "\n%8s %9s %11s %11s %11s %6s %6s %2s %6s %s\n",
	    "max", "wait_max", "total", "wait_total", "count", "avg", "wait_avg", "cnt_hold", "cnt_lock", "name");
	enabled = lock_prof_enable;
	lock_prof_enable = 0;
	/*
	 * See the comment in lock_prof_reset
	 */
	cpus_fence_seq_cst();
	quiesce_all_critical();
	t = ticks;
	CPU_FOREACH(cpu) {
		lock_prof_type_stats(&LP_CPU(cpu)->lpc_types[0], sb, 0, t);
		lock_prof_type_stats(&LP_CPU(cpu)->lpc_types[1], sb, 1, t);
	}
	atomic_thread_fence_rel();
	lock_prof_enable = enabled;

	error = sbuf_finish(sb);
	/* Output a trailing NUL. */
	if (error == 0)
		error = SYSCTL_OUT(req, "", 1);
	sbuf_delete(sb);
	return (error);
}

static int
enable_lock_prof(SYSCTL_HANDLER_ARGS)
{
	int error, v;

	v = lock_prof_enable;
	error = sysctl_handle_int(oidp, &v, v, req);
	if (error)
		return (error);
	if (req->newptr == NULL)
		return (error);
	if (v == lock_prof_enable)
		return (0);
	if (v == 1)
		lock_prof_reset();
	lock_prof_enable = !!v;

	return (0);
}

static int
reset_lock_prof_stats(SYSCTL_HANDLER_ARGS)
{
	int error, v;

	v = 0;
	error = sysctl_handle_int(oidp, &v, 0, req);
	if (error)
		return (error);
	if (req->newptr == NULL)
		return (error);
	if (v == 0)
		return (0);
	lock_prof_reset();

	return (0);
}

static struct lock_prof *
lock_profile_lookup(struct lock_object *lo, int spin, const char *file,
    int line)
{
	const char *unknown = "(unknown)";
	struct lock_prof_type *type;
	struct lock_prof *lp;
	struct lphead *head;
	const char *p;
	u_int hash;

	p = file;
	if (p == NULL || *p == '\0')
		p = unknown;
	hash = (uintptr_t)lo->lo_name * 31 + (uintptr_t)p * 31 + line;
	hash &= LPROF_HASH_MASK;
	type = &LP_CPU_SELF->lpc_types[spin];
	head = &type->lpt_hash[hash];
	SLIST_FOREACH(lp, head, link) {
		if (lp->line == line && lp->file == p &&
		    lp->name == lo->lo_name)
			return (lp);
	}
	lp = SLIST_FIRST(&type->lpt_lpalloc);
	if (lp == NULL) {
		lock_prof_rejected++;
		return (lp);
	}
	SLIST_REMOVE_HEAD(&type->lpt_lpalloc, link);
	lp->file = p;
	lp->line = line;
	lp->class = LOCK_CLASS(lo);
	lp->name = lo->lo_name;
	SLIST_INSERT_HEAD(&type->lpt_hash[hash], lp, link);
	return (lp);
}

static struct lock_profile_object *
lock_profile_object_lookup(struct lock_object *lo, int spin, const char *file,
    int line)
{
	struct lock_profile_object *l;
	struct lock_prof_type *type;
	struct lpohead *head;

	head = &curthread->td_lprof[spin];
	LIST_FOREACH(l, head, lpo_link)
		if (l->lpo_obj == lo && l->lpo_file == file &&
		    l->lpo_line == line)
			return (l);
	type = &LP_CPU_SELF->lpc_types[spin];
	l = LIST_FIRST(&type->lpt_lpoalloc);
	if (l == NULL) {
		lock_prof_rejected++;
		return (NULL);
	}
	LIST_REMOVE(l, lpo_link);
	l->lpo_obj = lo;
	l->lpo_file = file;
	l->lpo_line = line;
	l->lpo_cnt = 0;
	LIST_INSERT_HEAD(head, l, lpo_link);

	return (l);
}

void
lock_profile_obtain_lock_success(struct lock_object *lo, bool spin,
    int contested, uint64_t waittime, const char *file, int line)
{
	struct lock_profile_object *l;

#ifdef LOCK_PROFILING_DEBUG_SPIN
	bool is_spin = (LOCK_CLASS(lo)->lc_flags & LC_SPINLOCK);
	if ((spin && !is_spin) || (!spin && is_spin))
		printf("%s: lock %s spin mismatch (arg %d, flag %d)\n", __func__,
		    lo->lo_name, spin, is_spin);
#endif

	/* don't reset the timer when/if recursing */
	if (!lock_prof_enable || (lo->lo_flags & LO_NOPROFILE))
		return;
	if (lock_contested_only && !contested)
		return;
	if (spin && lock_prof_skipspin == 1)
		return;

	if (SCHEDULER_STOPPED())
		return;

	critical_enter();
	/* Recheck enabled now that we're in a critical section. */
	if (lock_prof_enable == 0)
		goto out;
	l = lock_profile_object_lookup(lo, spin, file, line);
	if (l == NULL)
		goto out;
	l->lpo_cnt++;
	if (++l->lpo_ref > 1)
		goto out;
	l->lpo_contest_locking = contested;
	l->lpo_acqtime = nanoseconds(); 
	if (waittime && (l->lpo_acqtime > waittime))
		l->lpo_waittime = l->lpo_acqtime - waittime;
	else
		l->lpo_waittime = 0;
out:
	/*
	 * Paired with cpus_fence_seq_cst().
	 */
	atomic_thread_fence_rel();
	critical_exit();
}

void
lock_profile_thread_exit(struct thread *td)
{
#ifdef INVARIANTS
	struct lock_profile_object *l;

	MPASS(curthread->td_critnest == 0);
#endif
	/*
	 * If lock profiling was disabled we have to wait for reset to
	 * clear our pointers before we can exit safely.
	 */
	lock_prof_reset_wait();
#ifdef INVARIANTS
	LIST_FOREACH(l, &td->td_lprof[0], lpo_link)
		printf("thread still holds lock acquired at %s:%d\n",
		    l->lpo_file, l->lpo_line);
	LIST_FOREACH(l, &td->td_lprof[1], lpo_link)
		printf("thread still holds lock acquired at %s:%d\n",
		    l->lpo_file, l->lpo_line);
#endif
	MPASS(LIST_FIRST(&td->td_lprof[0]) == NULL);
	MPASS(LIST_FIRST(&td->td_lprof[1]) == NULL);
}

void
lock_profile_release_lock(struct lock_object *lo, bool spin)
{
	struct lock_profile_object *l;
	struct lock_prof_type *type;
	struct lock_prof *lp;
	uint64_t curtime, holdtime;
	struct lpohead *head;

#ifdef LOCK_PROFILING_DEBUG_SPIN
	bool is_spin = (LOCK_CLASS(lo)->lc_flags & LC_SPINLOCK);
	if ((spin && !is_spin) || (!spin && is_spin))
		printf("%s: lock %s spin mismatch (arg %d, flag %d)\n", __func__,
		    lo->lo_name, spin, is_spin);
#endif

	if (lo->lo_flags & LO_NOPROFILE)
		return;
	head = &curthread->td_lprof[spin];
	if (LIST_FIRST(head) == NULL)
		return;
	if (SCHEDULER_STOPPED())
		return;
	critical_enter();
	/* Recheck enabled now that we're in a critical section. */
	if (lock_prof_enable == 0 && lock_prof_resetting == 1)
		goto out;
	/*
	 * If lock profiling is not enabled we still want to remove the
	 * lpo from our queue.
	 */
	LIST_FOREACH(l, head, lpo_link)
		if (l->lpo_obj == lo)
			break;
	if (l == NULL)
		goto out;
	if (--l->lpo_ref > 0)
		goto out;
	lp = lock_profile_lookup(lo, spin, l->lpo_file, l->lpo_line);
	if (lp == NULL)
		goto release;
	curtime = nanoseconds();
	if (curtime < l->lpo_acqtime)
		goto release;
	holdtime = curtime - l->lpo_acqtime;

	/*
	 * Record if the lock has been held longer now than ever
	 * before.
	 */
	if (holdtime > lp->cnt_max)
		lp->cnt_max = holdtime;
	if (l->lpo_waittime > lp->cnt_wait_max)
		lp->cnt_wait_max = l->lpo_waittime;
	lp->cnt_tot += holdtime;
	lp->cnt_wait += l->lpo_waittime;
	lp->cnt_contest_locking += l->lpo_contest_locking;
	lp->cnt_cur += l->lpo_cnt;
release:
	LIST_REMOVE(l, lpo_link);
	type = &LP_CPU_SELF->lpc_types[spin];
	LIST_INSERT_HEAD(&type->lpt_lpoalloc, l, lpo_link);
out:
	/*
	 * Paired with cpus_fence_seq_cst().
	 */
	atomic_thread_fence_rel();
	critical_exit();
}

static SYSCTL_NODE(_debug_lock, OID_AUTO, prof,
    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL,
    "lock profiling");
SYSCTL_INT(_debug_lock_prof, OID_AUTO, skipspin, CTLFLAG_RW,
    &lock_prof_skipspin, 0, "Skip profiling on spinlocks.");
SYSCTL_INT(_debug_lock_prof, OID_AUTO, rejected, CTLFLAG_RD,
    &lock_prof_rejected, 0, "Number of rejected profiling records");
SYSCTL_INT(_debug_lock_prof, OID_AUTO, contested_only, CTLFLAG_RW,
    &lock_contested_only, 0, "Only profile contested acquires");
SYSCTL_PROC(_debug_lock_prof, OID_AUTO, stats,
    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
    dump_lock_prof_stats, "A",
    "Lock profiling statistics");
SYSCTL_PROC(_debug_lock_prof, OID_AUTO, reset,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, NULL, 0,
    reset_lock_prof_stats, "I",
    "Reset lock profiling statistics");
SYSCTL_PROC(_debug_lock_prof, OID_AUTO, enable,
    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, 0,
    enable_lock_prof, "I",
    "Enable lock profiling");

#endif