lockdep.c 126.3 KB
Newer Older
I
Ingo Molnar 已提交
1 2 3 4 5 6 7
/*
 * kernel/lockdep.c
 *
 * Runtime locking correctness validator
 *
 * Started by Ingo Molnar:
 *
P
Peter Zijlstra 已提交
8
 *  Copyright (C) 2006,2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
9
 *  Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra
I
Ingo Molnar 已提交
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
 *
 * this code maps all the lock dependencies as they occur in a live kernel
 * and will warn about the following classes of locking bugs:
 *
 * - lock inversion scenarios
 * - circular lock dependencies
 * - hardirq/softirq safe/unsafe locking bugs
 *
 * Bugs are reported even if the current locking scenario does not cause
 * any deadlock at this point.
 *
 * I.e. if anytime in the past two locks were taken in a different order,
 * even if it happened for another task, even if those were different
 * locks (but of the same class as this lock), this code will detect it.
 *
 * Thanks to Arjan van de Ven for coming up with the initial idea of
 * mapping lock dependencies runtime.
 */
28
#define DISABLE_BRANCH_PROFILING
I
Ingo Molnar 已提交
29 30
#include <linux/mutex.h>
#include <linux/sched.h>
31
#include <linux/sched/clock.h>
32
#include <linux/sched/task.h>
33
#include <linux/sched/mm.h>
I
Ingo Molnar 已提交
34 35 36 37 38 39 40 41 42 43
#include <linux/delay.h>
#include <linux/module.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/spinlock.h>
#include <linux/kallsyms.h>
#include <linux/interrupt.h>
#include <linux/stacktrace.h>
#include <linux/debug_locks.h>
#include <linux/irqflags.h>
44
#include <linux/utsname.h>
P
Peter Zijlstra 已提交
45
#include <linux/hash.h>
46
#include <linux/ftrace.h>
P
Peter Zijlstra 已提交
47
#include <linux/stringify.h>
48
#include <linux/bitops.h>
49
#include <linux/gfp.h>
50
#include <linux/kmemcheck.h>
51
#include <linux/random.h>
52
#include <linux/jhash.h>
P
Peter Zijlstra 已提交
53

I
Ingo Molnar 已提交
54 55 56 57
#include <asm/sections.h>

#include "lockdep_internals.h"

58
#define CREATE_TRACE_POINTS
59
#include <trace/events/lock.h>
60

61 62 63 64
#ifdef CONFIG_LOCKDEP_CROSSRELEASE
#include <linux/slab.h>
#endif

P
Peter Zijlstra 已提交
65 66 67 68 69 70 71 72 73 74 75 76 77 78
#ifdef CONFIG_PROVE_LOCKING
int prove_locking = 1;
module_param(prove_locking, int, 0644);
#else
#define prove_locking 0
#endif

#ifdef CONFIG_LOCK_STAT
int lock_stat = 1;
module_param(lock_stat, int, 0644);
#else
#define lock_stat 0
#endif

79 80 81 82 83 84 85 86 87
static int crossrelease_fullstack;
static int __init allow_crossrelease_fullstack(char *str)
{
	crossrelease_fullstack = 1;
	return 0;
}

early_param("crossrelease_fullstack", allow_crossrelease_fullstack);

I
Ingo Molnar 已提交
88
/*
89 90
 * lockdep_lock: protects the lockdep graph, the hashes and the
 *               class/list/hash allocators.
I
Ingo Molnar 已提交
91 92 93
 *
 * This is one of the rare exceptions where it's justified
 * to use a raw spinlock - we really dont want the spinlock
94
 * code to recurse back into the lockdep code...
I
Ingo Molnar 已提交
95
 */
96
static arch_spinlock_t lockdep_lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
97 98 99

static int graph_lock(void)
{
100
	arch_spin_lock(&lockdep_lock);
101 102 103 104 105 106 107
	/*
	 * Make sure that if another CPU detected a bug while
	 * walking the graph we dont change it (while the other
	 * CPU is busy printing out stuff with the graph lock
	 * dropped already)
	 */
	if (!debug_locks) {
108
		arch_spin_unlock(&lockdep_lock);
109 110
		return 0;
	}
111 112
	/* prevent any recursions within lockdep from causing deadlocks */
	current->lockdep_recursion++;
113 114 115 116 117
	return 1;
}

static inline int graph_unlock(void)
{
P
Peter Zijlstra 已提交
118 119 120 121 122
	if (debug_locks && !arch_spin_is_locked(&lockdep_lock)) {
		/*
		 * The lockdep graph lock isn't locked while we expect it to
		 * be, we're confused now, bye!
		 */
123
		return DEBUG_LOCKS_WARN_ON(1);
P
Peter Zijlstra 已提交
124
	}
125

126
	current->lockdep_recursion--;
127
	arch_spin_unlock(&lockdep_lock);
128 129 130 131 132 133 134 135 136 137 138
	return 0;
}

/*
 * Turn lock debugging off and return with 0 if it was off already,
 * and also release the graph lock:
 */
static inline int debug_locks_off_graph_unlock(void)
{
	int ret = debug_locks_off();

139
	arch_spin_unlock(&lockdep_lock);
140 141 142

	return ret;
}
I
Ingo Molnar 已提交
143 144

unsigned long nr_list_entries;
P
Peter Zijlstra 已提交
145
static struct lock_list list_entries[MAX_LOCKDEP_ENTRIES];
I
Ingo Molnar 已提交
146 147 148 149 150 151 152 153 154 155

/*
 * All data structures here are protected by the global debug_lock.
 *
 * Mutex key structs only get allocated, once during bootup, and never
 * get freed - this significantly simplifies the debugging code.
 */
unsigned long nr_lock_classes;
static struct lock_class lock_classes[MAX_LOCKDEP_KEYS];

D
Dave Jones 已提交
156 157 158
static inline struct lock_class *hlock_class(struct held_lock *hlock)
{
	if (!hlock->class_idx) {
P
Peter Zijlstra 已提交
159 160 161
		/*
		 * Someone passed in garbage, we give up.
		 */
D
Dave Jones 已提交
162 163 164 165 166 167
		DEBUG_LOCKS_WARN_ON(1);
		return NULL;
	}
	return lock_classes + hlock->class_idx - 1;
}

P
Peter Zijlstra 已提交
168
#ifdef CONFIG_LOCK_STAT
169
static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
P
Peter Zijlstra 已提交
170

171 172
static inline u64 lockstat_clock(void)
{
173
	return local_clock();
174 175
}

P
Peter Zijlstra 已提交
176
static int lock_point(unsigned long points[], unsigned long ip)
P
Peter Zijlstra 已提交
177 178 179
{
	int i;

P
Peter Zijlstra 已提交
180 181 182
	for (i = 0; i < LOCKSTAT_POINTS; i++) {
		if (points[i] == 0) {
			points[i] = ip;
P
Peter Zijlstra 已提交
183 184
			break;
		}
P
Peter Zijlstra 已提交
185
		if (points[i] == ip)
P
Peter Zijlstra 已提交
186 187 188 189 190 191
			break;
	}

	return i;
}

192
static void lock_time_inc(struct lock_time *lt, u64 time)
P
Peter Zijlstra 已提交
193 194 195 196
{
	if (time > lt->max)
		lt->max = time;

197
	if (time < lt->min || !lt->nr)
P
Peter Zijlstra 已提交
198 199 200 201 202 203
		lt->min = time;

	lt->total += time;
	lt->nr++;
}

204 205
static inline void lock_time_add(struct lock_time *src, struct lock_time *dst)
{
206 207 208 209 210 211 212 213 214
	if (!src->nr)
		return;

	if (src->max > dst->max)
		dst->max = src->max;

	if (src->min < dst->min || !dst->nr)
		dst->min = src->min;

215 216 217 218 219 220 221 222 223 224 225 226
	dst->total += src->total;
	dst->nr += src->nr;
}

struct lock_class_stats lock_stats(struct lock_class *class)
{
	struct lock_class_stats stats;
	int cpu, i;

	memset(&stats, 0, sizeof(struct lock_class_stats));
	for_each_possible_cpu(cpu) {
		struct lock_class_stats *pcs =
227
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
228 229 230 231

		for (i = 0; i < ARRAY_SIZE(stats.contention_point); i++)
			stats.contention_point[i] += pcs->contention_point[i];

P
Peter Zijlstra 已提交
232 233 234
		for (i = 0; i < ARRAY_SIZE(stats.contending_point); i++)
			stats.contending_point[i] += pcs->contending_point[i];

235 236 237 238 239
		lock_time_add(&pcs->read_waittime, &stats.read_waittime);
		lock_time_add(&pcs->write_waittime, &stats.write_waittime);

		lock_time_add(&pcs->read_holdtime, &stats.read_holdtime);
		lock_time_add(&pcs->write_holdtime, &stats.write_holdtime);
P
Peter Zijlstra 已提交
240 241 242

		for (i = 0; i < ARRAY_SIZE(stats.bounces); i++)
			stats.bounces[i] += pcs->bounces[i];
243 244 245 246 247 248 249 250 251 252 253
	}

	return stats;
}

void clear_lock_stats(struct lock_class *class)
{
	int cpu;

	for_each_possible_cpu(cpu) {
		struct lock_class_stats *cpu_stats =
254
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
255 256 257 258

		memset(cpu_stats, 0, sizeof(struct lock_class_stats));
	}
	memset(class->contention_point, 0, sizeof(class->contention_point));
P
Peter Zijlstra 已提交
259
	memset(class->contending_point, 0, sizeof(class->contending_point));
260 261
}

P
Peter Zijlstra 已提交
262 263
static struct lock_class_stats *get_lock_stats(struct lock_class *class)
{
264
	return &get_cpu_var(cpu_lock_stats)[class - lock_classes];
P
Peter Zijlstra 已提交
265 266 267 268
}

static void put_lock_stats(struct lock_class_stats *stats)
{
269
	put_cpu_var(cpu_lock_stats);
P
Peter Zijlstra 已提交
270 271 272 273 274
}

static void lock_release_holdtime(struct held_lock *hlock)
{
	struct lock_class_stats *stats;
275
	u64 holdtime;
P
Peter Zijlstra 已提交
276 277 278 279

	if (!lock_stat)
		return;

280
	holdtime = lockstat_clock() - hlock->holdtime_stamp;
P
Peter Zijlstra 已提交
281

D
Dave Jones 已提交
282
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
283 284 285 286 287 288 289 290 291 292 293 294
	if (hlock->read)
		lock_time_inc(&stats->read_holdtime, holdtime);
	else
		lock_time_inc(&stats->write_holdtime, holdtime);
	put_lock_stats(stats);
}
#else
static inline void lock_release_holdtime(struct held_lock *hlock)
{
}
#endif

I
Ingo Molnar 已提交
295 296 297 298 299 300 301 302 303 304 305 306
/*
 * We keep a global list of all lock classes. The list only grows,
 * never shrinks. The list is only accessed with the lockdep
 * spinlock lock held.
 */
LIST_HEAD(all_lock_classes);

/*
 * The lockdep classes are in a hash-table as well, for fast lookup:
 */
#define CLASSHASH_BITS		(MAX_LOCKDEP_KEYS_BITS - 1)
#define CLASSHASH_SIZE		(1UL << CLASSHASH_BITS)
P
Peter Zijlstra 已提交
307
#define __classhashfn(key)	hash_long((unsigned long)key, CLASSHASH_BITS)
I
Ingo Molnar 已提交
308 309
#define classhashentry(key)	(classhash_table + __classhashfn((key)))

310
static struct hlist_head classhash_table[CLASSHASH_SIZE];
I
Ingo Molnar 已提交
311 312 313 314 315 316 317

/*
 * We put the lock dependency chains into a hash-table as well, to cache
 * their existence:
 */
#define CHAINHASH_BITS		(MAX_LOCKDEP_CHAINS_BITS-1)
#define CHAINHASH_SIZE		(1UL << CHAINHASH_BITS)
P
Peter Zijlstra 已提交
318
#define __chainhashfn(chain)	hash_long(chain, CHAINHASH_BITS)
I
Ingo Molnar 已提交
319 320
#define chainhashentry(chain)	(chainhash_table + __chainhashfn((chain)))

321
static struct hlist_head chainhash_table[CHAINHASH_SIZE];
I
Ingo Molnar 已提交
322 323 324 325 326 327 328

/*
 * The hash key of the lock dependency chains is a hash itself too:
 * it's a hash of all locks taken up to that lock, including that lock.
 * It's a 64-bit hash, because it's important for the keys to be
 * unique.
 */
329 330 331 332 333 334 335 336
static inline u64 iterate_chain_key(u64 key, u32 idx)
{
	u32 k0 = key, k1 = key >> 32;

	__jhash_mix(idx, k0, k1); /* Macro that modifies arguments! */

	return k0 | (u64)k1 << 32;
}
I
Ingo Molnar 已提交
337

338
void lockdep_off(void)
I
Ingo Molnar 已提交
339 340 341 342 343
{
	current->lockdep_recursion++;
}
EXPORT_SYMBOL(lockdep_off);

344
void lockdep_on(void)
I
Ingo Molnar 已提交
345 346 347 348 349 350 351 352 353 354
{
	current->lockdep_recursion--;
}
EXPORT_SYMBOL(lockdep_on);

/*
 * Debugging switches:
 */

#define VERBOSE			0
355
#define VERY_VERBOSE		0
I
Ingo Molnar 已提交
356 357 358 359 360 361 362 363 364

#if VERBOSE
# define HARDIRQ_VERBOSE	1
# define SOFTIRQ_VERBOSE	1
#else
# define HARDIRQ_VERBOSE	0
# define SOFTIRQ_VERBOSE	0
#endif

365
#if VERBOSE || HARDIRQ_VERBOSE || SOFTIRQ_VERBOSE
I
Ingo Molnar 已提交
366 367 368 369 370
/*
 * Quick filtering for interesting events:
 */
static int class_filter(struct lock_class *class)
{
A
Andi Kleen 已提交
371 372
#if 0
	/* Example */
I
Ingo Molnar 已提交
373
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
374
			!strcmp(class->name, "lockname"))
I
Ingo Molnar 已提交
375 376
		return 1;
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
377
			!strcmp(class->name, "&struct->lockfield"))
I
Ingo Molnar 已提交
378
		return 1;
A
Andi Kleen 已提交
379
#endif
380 381
	/* Filter everything else. 1 would be to allow everything else */
	return 0;
I
Ingo Molnar 已提交
382 383 384 385 386 387 388 389 390 391 392 393 394
}
#endif

static int verbose(struct lock_class *class)
{
#if VERBOSE
	return class_filter(class);
#endif
	return 0;
}

/*
 * Stack-trace: tightly packed array of stack backtrace
395
 * addresses. Protected by the graph_lock.
I
Ingo Molnar 已提交
396 397 398 399
 */
unsigned long nr_stack_trace_entries;
static unsigned long stack_trace[MAX_STACK_TRACE_ENTRIES];

400 401 402 403
static void print_lockdep_off(const char *bug_msg)
{
	printk(KERN_DEBUG "%s\n", bug_msg);
	printk(KERN_DEBUG "turning off the locking correctness validator.\n");
404
#ifdef CONFIG_LOCK_STAT
405
	printk(KERN_DEBUG "Please attach the output of /proc/lock_stat to the bug report\n");
406
#endif
407 408
}

I
Ingo Molnar 已提交
409 410 411 412 413 414
static int save_trace(struct stack_trace *trace)
{
	trace->nr_entries = 0;
	trace->max_entries = MAX_STACK_TRACE_ENTRIES - nr_stack_trace_entries;
	trace->entries = stack_trace + nr_stack_trace_entries;

415 416
	trace->skip = 3;

C
Christoph Hellwig 已提交
417
	save_stack_trace(trace);
I
Ingo Molnar 已提交
418

P
Peter Zijlstra 已提交
419 420 421 422 423 424 425
	/*
	 * Some daft arches put -1 at the end to indicate its a full trace.
	 *
	 * <rant> this is buggy anyway, since it takes a whole extra entry so a
	 * complete trace that maxes out the entries provided will be reported
	 * as incomplete, friggin useless </rant>
	 */
426 427
	if (trace->nr_entries != 0 &&
	    trace->entries[trace->nr_entries-1] == ULONG_MAX)
P
Peter Zijlstra 已提交
428 429
		trace->nr_entries--;

I
Ingo Molnar 已提交
430 431 432 433
	trace->max_entries = trace->nr_entries;

	nr_stack_trace_entries += trace->nr_entries;

P
Peter Zijlstra 已提交
434
	if (nr_stack_trace_entries >= MAX_STACK_TRACE_ENTRIES-1) {
435 436 437
		if (!debug_locks_off_graph_unlock())
			return 0;

438
		print_lockdep_off("BUG: MAX_STACK_TRACE_ENTRIES too low!");
439 440
		dump_stack();

I
Ingo Molnar 已提交
441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
		return 0;
	}

	return 1;
}

unsigned int nr_hardirq_chains;
unsigned int nr_softirq_chains;
unsigned int nr_process_chains;
unsigned int max_lockdep_depth;

#ifdef CONFIG_DEBUG_LOCKDEP
/*
 * Various lockdep statistics:
 */
456
DEFINE_PER_CPU(struct lockdep_stats, lockdep_stats);
I
Ingo Molnar 已提交
457 458 459 460 461 462
#endif

/*
 * Locking printouts:
 */

P
Peter Zijlstra 已提交
463
#define __USAGE(__STATE)						\
P
Peter Zijlstra 已提交
464 465 466 467
	[LOCK_USED_IN_##__STATE] = "IN-"__stringify(__STATE)"-W",	\
	[LOCK_ENABLED_##__STATE] = __stringify(__STATE)"-ON-W",		\
	[LOCK_USED_IN_##__STATE##_READ] = "IN-"__stringify(__STATE)"-R",\
	[LOCK_ENABLED_##__STATE##_READ] = __stringify(__STATE)"-ON-R",
P
Peter Zijlstra 已提交
468

I
Ingo Molnar 已提交
469 470
static const char *usage_str[] =
{
P
Peter Zijlstra 已提交
471 472 473 474
#define LOCKDEP_STATE(__STATE) __USAGE(__STATE)
#include "lockdep_states.h"
#undef LOCKDEP_STATE
	[LOCK_USED] = "INITIAL USE",
I
Ingo Molnar 已提交
475 476 477 478
};

const char * __get_key_name(struct lockdep_subclass_key *key, char *str)
{
A
Alexey Dobriyan 已提交
479
	return kallsyms_lookup((unsigned long)key, NULL, NULL, NULL, str);
I
Ingo Molnar 已提交
480 481
}

482
static inline unsigned long lock_flag(enum lock_usage_bit bit)
I
Ingo Molnar 已提交
483
{
484 485
	return 1UL << bit;
}
I
Ingo Molnar 已提交
486

487 488 489 490 491 492 493 494 495 496
static char get_usage_char(struct lock_class *class, enum lock_usage_bit bit)
{
	char c = '.';

	if (class->usage_mask & lock_flag(bit + 2))
		c = '+';
	if (class->usage_mask & lock_flag(bit)) {
		c = '-';
		if (class->usage_mask & lock_flag(bit + 2))
			c = '?';
I
Ingo Molnar 已提交
497 498
	}

499 500
	return c;
}
501

P
Peter Zijlstra 已提交
502
void get_usage_chars(struct lock_class *class, char usage[LOCK_USAGE_CHARS])
503
{
P
Peter Zijlstra 已提交
504
	int i = 0;
505

P
Peter Zijlstra 已提交
506 507 508 509 510 511 512
#define LOCKDEP_STATE(__STATE) 						\
	usage[i++] = get_usage_char(class, LOCK_USED_IN_##__STATE);	\
	usage[i++] = get_usage_char(class, LOCK_USED_IN_##__STATE##_READ);
#include "lockdep_states.h"
#undef LOCKDEP_STATE

	usage[i] = '\0';
I
Ingo Molnar 已提交
513 514
}

515
static void __print_lock_name(struct lock_class *class)
516 517 518 519
{
	char str[KSYM_NAME_LEN];
	const char *name;

I
Ingo Molnar 已提交
520 521 522
	name = class->name;
	if (!name) {
		name = __get_key_name(class->key, str);
D
Dmitry Vyukov 已提交
523
		printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
524
	} else {
D
Dmitry Vyukov 已提交
525
		printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
526
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
527
			printk(KERN_CONT "#%d", class->name_version);
I
Ingo Molnar 已提交
528
		if (class->subclass)
D
Dmitry Vyukov 已提交
529
			printk(KERN_CONT "/%d", class->subclass);
I
Ingo Molnar 已提交
530
	}
531 532 533 534 535 536 537 538
}

static void print_lock_name(struct lock_class *class)
{
	char usage[LOCK_USAGE_CHARS];

	get_usage_chars(class, usage);

D
Dmitry Vyukov 已提交
539
	printk(KERN_CONT " (");
540
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
541
	printk(KERN_CONT "){%s}", usage);
I
Ingo Molnar 已提交
542 543 544 545 546
}

static void print_lockdep_cache(struct lockdep_map *lock)
{
	const char *name;
547
	char str[KSYM_NAME_LEN];
I
Ingo Molnar 已提交
548 549 550 551 552

	name = lock->name;
	if (!name)
		name = __get_key_name(lock->key->subkeys, str);

D
Dmitry Vyukov 已提交
553
	printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
554 555 556 557
}

static void print_lock(struct held_lock *hlock)
{
558 559 560 561 562 563 564 565 566 567
	/*
	 * We can be called locklessly through debug_show_all_locks() so be
	 * extra careful, the hlock might have been released and cleared.
	 */
	unsigned int class_idx = hlock->class_idx;

	/* Don't re-read hlock->class_idx, can't use READ_ONCE() on bitfields: */
	barrier();

	if (!class_idx || (class_idx - 1) >= MAX_LOCKDEP_KEYS) {
D
Dmitry Vyukov 已提交
568
		printk(KERN_CONT "<RELEASED>\n");
569 570 571 572
		return;
	}

	print_lock_name(lock_classes + class_idx - 1);
D
Dmitry Vyukov 已提交
573 574
	printk(KERN_CONT ", at: [<%p>] %pS\n",
		(void *)hlock->acquire_ip, (void *)hlock->acquire_ip);
I
Ingo Molnar 已提交
575 576 577 578 579 580 581
}

static void lockdep_print_held_locks(struct task_struct *curr)
{
	int i, depth = curr->lockdep_depth;

	if (!depth) {
582
		printk("no locks held by %s/%d.\n", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
583 584 585
		return;
	}
	printk("%d lock%s held by %s/%d:\n",
586
		depth, depth > 1 ? "s" : "", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
587 588 589 590 591 592 593

	for (i = 0; i < depth; i++) {
		printk(" #%d: ", i);
		print_lock(curr->held_locks + i);
	}
}

594
static void print_kernel_ident(void)
P
Peter Zijlstra 已提交
595
{
596
	printk("%s %.*s %s\n", init_utsname()->release,
P
Peter Zijlstra 已提交
597
		(int)strcspn(init_utsname()->version, " "),
598 599
		init_utsname()->version,
		print_tainted());
P
Peter Zijlstra 已提交
600 601 602 603 604 605 606 607 608 609
}

static int very_verbose(struct lock_class *class)
{
#if VERY_VERBOSE
	return class_filter(class);
#endif
	return 0;
}

I
Ingo Molnar 已提交
610
/*
P
Peter Zijlstra 已提交
611
 * Is this the address of a static object:
I
Ingo Molnar 已提交
612
 */
613
#ifdef __KERNEL__
P
Peter Zijlstra 已提交
614
static int static_obj(void *obj)
I
Ingo Molnar 已提交
615
{
P
Peter Zijlstra 已提交
616 617 618 619
	unsigned long start = (unsigned long) &_stext,
		      end   = (unsigned long) &_end,
		      addr  = (unsigned long) obj;

I
Ingo Molnar 已提交
620
	/*
P
Peter Zijlstra 已提交
621
	 * static variable?
I
Ingo Molnar 已提交
622
	 */
P
Peter Zijlstra 已提交
623 624
	if ((addr >= start) && (addr < end))
		return 1;
I
Ingo Molnar 已提交
625

626 627 628
	if (arch_is_kernel_data(addr))
		return 1;

I
Ingo Molnar 已提交
629
	/*
630
	 * in-kernel percpu var?
I
Ingo Molnar 已提交
631
	 */
632 633
	if (is_kernel_percpu_address(addr))
		return 1;
I
Ingo Molnar 已提交
634

P
Peter Zijlstra 已提交
635
	/*
636
	 * module static or percpu var?
P
Peter Zijlstra 已提交
637
	 */
638
	return is_module_address(addr) || is_module_percpu_address(addr);
639
}
640
#endif
641

I
Ingo Molnar 已提交
642
/*
P
Peter Zijlstra 已提交
643 644
 * To make lock name printouts unique, we calculate a unique
 * class->name_version generation counter:
I
Ingo Molnar 已提交
645
 */
P
Peter Zijlstra 已提交
646
static int count_matching_names(struct lock_class *new_class)
I
Ingo Molnar 已提交
647
{
P
Peter Zijlstra 已提交
648 649
	struct lock_class *class;
	int count = 0;
I
Ingo Molnar 已提交
650

P
Peter Zijlstra 已提交
651
	if (!new_class->name)
I
Ingo Molnar 已提交
652 653
		return 0;

654
	list_for_each_entry_rcu(class, &all_lock_classes, lock_entry) {
P
Peter Zijlstra 已提交
655 656 657 658 659
		if (new_class->key - new_class->subclass == class->key)
			return class->name_version;
		if (class->name && !strcmp(class->name, new_class->name))
			count = max(count, class->name_version);
	}
I
Ingo Molnar 已提交
660

P
Peter Zijlstra 已提交
661
	return count + 1;
I
Ingo Molnar 已提交
662 663
}

P
Peter Zijlstra 已提交
664 665 666 667 668 669 670
/*
 * Register a lock's class in the hash-table, if the class is not present
 * yet. Otherwise we look it up. We cache the result in the lock object
 * itself, so actual lookup of the hash should be once per lock object.
 */
static inline struct lock_class *
look_up_lock_class(struct lockdep_map *lock, unsigned int subclass)
I
Ingo Molnar 已提交
671
{
P
Peter Zijlstra 已提交
672
	struct lockdep_subclass_key *key;
673
	struct hlist_head *hash_head;
P
Peter Zijlstra 已提交
674
	struct lock_class *class;
675
	bool is_static = false;
I
Ingo Molnar 已提交
676

677 678 679 680 681 682 683 684 685 686
	if (unlikely(subclass >= MAX_LOCKDEP_SUBCLASSES)) {
		debug_locks_off();
		printk(KERN_ERR
			"BUG: looking up invalid subclass: %u\n", subclass);
		printk(KERN_ERR
			"turning off the locking correctness validator.\n");
		dump_stack();
		return NULL;
	}

P
Peter Zijlstra 已提交
687 688
	/*
	 * Static locks do not have their class-keys yet - for them the key
689 690 691
	 * is the lock object itself. If the lock is in the per cpu area,
	 * the canonical address of the lock (per cpu offset removed) is
	 * used.
P
Peter Zijlstra 已提交
692
	 */
693 694 695 696 697 698 699 700 701 702 703 704 705
	if (unlikely(!lock->key)) {
		unsigned long can_addr, addr = (unsigned long)lock;

		if (__is_kernel_percpu_address(addr, &can_addr))
			lock->key = (void *)can_addr;
		else if (__is_module_percpu_address(addr, &can_addr))
			lock->key = (void *)can_addr;
		else if (static_obj(lock))
			lock->key = (void *)lock;
		else
			return ERR_PTR(-EINVAL);
		is_static = true;
	}
I
Ingo Molnar 已提交
706

P
Peter Zijlstra 已提交
707 708 709 710 711 712
	/*
	 * NOTE: the class-key must be unique. For dynamic locks, a static
	 * lock_class_key variable is passed in through the mutex_init()
	 * (or spin_lock_init()) call - which acts as the key. For static
	 * locks we use the lock object itself as the key.
	 */
P
Peter Zijlstra 已提交
713 714
	BUILD_BUG_ON(sizeof(struct lock_class_key) >
			sizeof(struct lockdep_map));
I
Ingo Molnar 已提交
715

P
Peter Zijlstra 已提交
716
	key = lock->key->subkeys + subclass;
717

P
Peter Zijlstra 已提交
718
	hash_head = classhashentry(key);
719

P
Peter Zijlstra 已提交
720
	/*
721
	 * We do an RCU walk of the hash, see lockdep_free_key_range().
P
Peter Zijlstra 已提交
722
	 */
723 724 725
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return NULL;

726
	hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
P
Peter Zijlstra 已提交
727
		if (class->key == key) {
P
Peter Zijlstra 已提交
728 729 730 731
			/*
			 * Huh! same key, different name? Did someone trample
			 * on some memory? We're most confused.
			 */
P
Peter Zijlstra 已提交
732
			WARN_ON_ONCE(class->name != lock->name);
P
Peter Zijlstra 已提交
733
			return class;
P
Peter Zijlstra 已提交
734 735
		}
	}
I
Ingo Molnar 已提交
736

737
	return is_static || static_obj(lock->key) ? NULL : ERR_PTR(-EINVAL);
I
Ingo Molnar 已提交
738 739
}

740 741 742 743 744 745 746 747 748 749 750 751
#ifdef CONFIG_LOCKDEP_CROSSRELEASE
static void cross_init(struct lockdep_map *lock, int cross);
static int cross_lock(struct lockdep_map *lock);
static int lock_acquire_crosslock(struct held_lock *hlock);
static int lock_release_crosslock(struct lockdep_map *lock);
#else
static inline void cross_init(struct lockdep_map *lock, int cross) {}
static inline int cross_lock(struct lockdep_map *lock) { return 0; }
static inline int lock_acquire_crosslock(struct held_lock *hlock) { return 2; }
static inline int lock_release_crosslock(struct lockdep_map *lock) { return 2; }
#endif

I
Ingo Molnar 已提交
752
/*
P
Peter Zijlstra 已提交
753 754 755
 * Register a lock's class in the hash-table, if the class is not present
 * yet. Otherwise we look it up. We cache the result in the lock object
 * itself, so actual lookup of the hash should be once per lock object.
I
Ingo Molnar 已提交
756
 */
757
static struct lock_class *
P
Peter Zijlstra 已提交
758
register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
I
Ingo Molnar 已提交
759
{
P
Peter Zijlstra 已提交
760
	struct lockdep_subclass_key *key;
761
	struct hlist_head *hash_head;
P
Peter Zijlstra 已提交
762
	struct lock_class *class;
763 764

	DEBUG_LOCKS_WARN_ON(!irqs_disabled());
P
Peter Zijlstra 已提交
765 766

	class = look_up_lock_class(lock, subclass);
767
	if (likely(!IS_ERR_OR_NULL(class)))
768
		goto out_set_class_cache;
P
Peter Zijlstra 已提交
769 770 771

	/*
	 * Debug-check: all keys must be persistent!
772 773
	 */
	if (IS_ERR(class)) {
P
Peter Zijlstra 已提交
774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791
		debug_locks_off();
		printk("INFO: trying to register non-static key.\n");
		printk("the code is fine but needs lockdep annotation.\n");
		printk("turning off the locking correctness validator.\n");
		dump_stack();
		return NULL;
	}

	key = lock->key->subkeys + subclass;
	hash_head = classhashentry(key);

	if (!graph_lock()) {
		return NULL;
	}
	/*
	 * We have to do the hash-walk again, to avoid races
	 * with another CPU:
	 */
792
	hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
P
Peter Zijlstra 已提交
793 794
		if (class->key == key)
			goto out_unlock_set;
795 796
	}

P
Peter Zijlstra 已提交
797 798 799 800 801 802 803 804 805
	/*
	 * Allocate a new key from the static array, and add it to
	 * the hash:
	 */
	if (nr_lock_classes >= MAX_LOCKDEP_KEYS) {
		if (!debug_locks_off_graph_unlock()) {
			return NULL;
		}

806
		print_lockdep_off("BUG: MAX_LOCKDEP_KEYS too low!");
807
		dump_stack();
P
Peter Zijlstra 已提交
808 809 810
		return NULL;
	}
	class = lock_classes + nr_lock_classes++;
811
	debug_atomic_inc(nr_unused_locks);
P
Peter Zijlstra 已提交
812 813 814 815 816 817 818 819 820 821 822
	class->key = key;
	class->name = lock->name;
	class->subclass = subclass;
	INIT_LIST_HEAD(&class->lock_entry);
	INIT_LIST_HEAD(&class->locks_before);
	INIT_LIST_HEAD(&class->locks_after);
	class->name_version = count_matching_names(class);
	/*
	 * We use RCU's safe list-add method to make
	 * parallel walking of the hash-list safe:
	 */
823
	hlist_add_head_rcu(&class->hash_entry, hash_head);
824 825 826 827
	/*
	 * Add it to the global list of classes:
	 */
	list_add_tail_rcu(&class->lock_entry, &all_lock_classes);
P
Peter Zijlstra 已提交
828 829 830 831 832 833

	if (verbose(class)) {
		graph_unlock();

		printk("\nnew class %p: %s", class->key, class->name);
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
834 835
			printk(KERN_CONT "#%d", class->name_version);
		printk(KERN_CONT "\n");
P
Peter Zijlstra 已提交
836 837 838 839 840 841 842 843 844
		dump_stack();

		if (!graph_lock()) {
			return NULL;
		}
	}
out_unlock_set:
	graph_unlock();

845
out_set_class_cache:
P
Peter Zijlstra 已提交
846
	if (!subclass || force)
847 848 849
		lock->class_cache[0] = class;
	else if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		lock->class_cache[subclass] = class;
P
Peter Zijlstra 已提交
850

P
Peter Zijlstra 已提交
851 852 853 854
	/*
	 * Hash collision, did we smoke some? We found a class with a matching
	 * hash but the subclass -- which is hashed in -- didn't match.
	 */
P
Peter Zijlstra 已提交
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871
	if (DEBUG_LOCKS_WARN_ON(class->subclass != subclass))
		return NULL;

	return class;
}

#ifdef CONFIG_PROVE_LOCKING
/*
 * Allocate a lockdep entry. (assumes the graph_lock held, returns
 * with NULL on failure)
 */
static struct lock_list *alloc_list_entry(void)
{
	if (nr_list_entries >= MAX_LOCKDEP_ENTRIES) {
		if (!debug_locks_off_graph_unlock())
			return NULL;

872
		print_lockdep_off("BUG: MAX_LOCKDEP_ENTRIES too low!");
873
		dump_stack();
P
Peter Zijlstra 已提交
874 875 876 877 878 879 880 881
		return NULL;
	}
	return list_entries + nr_list_entries++;
}

/*
 * Add a new dependency to the head of the list:
 */
882 883 884
static int add_lock_to_list(struct lock_class *this, struct list_head *head,
			    unsigned long ip, int distance,
			    struct stack_trace *trace)
P
Peter Zijlstra 已提交
885 886 887 888 889 890 891 892 893 894
{
	struct lock_list *entry;
	/*
	 * Lock not present yet - get a new dependency struct and
	 * add it to the list:
	 */
	entry = alloc_list_entry();
	if (!entry)
		return 0;

895 896
	entry->class = this;
	entry->distance = distance;
Y
Yong Zhang 已提交
897
	entry->trace = *trace;
P
Peter Zijlstra 已提交
898
	/*
899 900 901
	 * Both allocation and removal are done under the graph lock; but
	 * iteration is under RCU-sched; see look_up_lock_class() and
	 * lockdep_free_key_range().
P
Peter Zijlstra 已提交
902 903 904 905 906 907
	 */
	list_add_tail_rcu(&entry->entry, head);

	return 1;
}

P
Peter Zijlstra 已提交
908 909 910
/*
 * For good efficiency of modular, we use power of 2
 */
P
Peter Zijlstra 已提交
911 912 913
#define MAX_CIRCULAR_QUEUE_SIZE		4096UL
#define CQ_MASK				(MAX_CIRCULAR_QUEUE_SIZE-1)

P
Peter Zijlstra 已提交
914 915
/*
 * The circular_queue and helpers is used to implement the
P
Peter Zijlstra 已提交
916 917 918
 * breadth-first search(BFS)algorithem, by which we can build
 * the shortest path from the next lock to be acquired to the
 * previous held lock if there is a circular between them.
P
Peter Zijlstra 已提交
919
 */
P
Peter Zijlstra 已提交
920 921 922 923 924 925 926
struct circular_queue {
	unsigned long element[MAX_CIRCULAR_QUEUE_SIZE];
	unsigned int  front, rear;
};

static struct circular_queue lock_cq;

927
unsigned int max_bfs_queue_depth;
P
Peter Zijlstra 已提交
928

929 930
static unsigned int lockdep_dependency_gen_id;

P
Peter Zijlstra 已提交
931 932 933
static inline void __cq_init(struct circular_queue *cq)
{
	cq->front = cq->rear = 0;
934
	lockdep_dependency_gen_id++;
P
Peter Zijlstra 已提交
935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
}

static inline int __cq_empty(struct circular_queue *cq)
{
	return (cq->front == cq->rear);
}

static inline int __cq_full(struct circular_queue *cq)
{
	return ((cq->rear + 1) & CQ_MASK) == cq->front;
}

static inline int __cq_enqueue(struct circular_queue *cq, unsigned long elem)
{
	if (__cq_full(cq))
		return -1;

	cq->element[cq->rear] = elem;
	cq->rear = (cq->rear + 1) & CQ_MASK;
	return 0;
}

static inline int __cq_dequeue(struct circular_queue *cq, unsigned long *elem)
{
	if (__cq_empty(cq))
		return -1;

	*elem = cq->element[cq->front];
	cq->front = (cq->front + 1) & CQ_MASK;
	return 0;
}

static inline unsigned int  __cq_get_elem_count(struct circular_queue *cq)
{
	return (cq->rear - cq->front) & CQ_MASK;
}

static inline void mark_lock_accessed(struct lock_list *lock,
					struct lock_list *parent)
{
	unsigned long nr;
P
Peter Zijlstra 已提交
976

P
Peter Zijlstra 已提交
977
	nr = lock - list_entries;
P
Peter Zijlstra 已提交
978
	WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
P
Peter Zijlstra 已提交
979
	lock->parent = parent;
980
	lock->class->dep_gen_id = lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
981 982 983 984 985
}

static inline unsigned long lock_accessed(struct lock_list *lock)
{
	unsigned long nr;
P
Peter Zijlstra 已提交
986

P
Peter Zijlstra 已提交
987
	nr = lock - list_entries;
P
Peter Zijlstra 已提交
988
	WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
989
	return lock->class->dep_gen_id == lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
}

static inline struct lock_list *get_lock_parent(struct lock_list *child)
{
	return child->parent;
}

static inline int get_lock_depth(struct lock_list *child)
{
	int depth = 0;
	struct lock_list *parent;

	while ((parent = get_lock_parent(child))) {
		child = parent;
		depth++;
	}
	return depth;
}

1009
static int __bfs(struct lock_list *source_entry,
P
Peter Zijlstra 已提交
1010 1011 1012 1013
		 void *data,
		 int (*match)(struct lock_list *entry, void *data),
		 struct lock_list **target_entry,
		 int forward)
1014 1015
{
	struct lock_list *entry;
1016
	struct list_head *head;
1017 1018 1019
	struct circular_queue *cq = &lock_cq;
	int ret = 1;

1020
	if (match(source_entry, data)) {
1021 1022 1023 1024 1025
		*target_entry = source_entry;
		ret = 0;
		goto exit;
	}

1026 1027 1028 1029 1030 1031 1032 1033 1034
	if (forward)
		head = &source_entry->class->locks_after;
	else
		head = &source_entry->class->locks_before;

	if (list_empty(head))
		goto exit;

	__cq_init(cq);
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
	__cq_enqueue(cq, (unsigned long)source_entry);

	while (!__cq_empty(cq)) {
		struct lock_list *lock;

		__cq_dequeue(cq, (unsigned long *)&lock);

		if (!lock->class) {
			ret = -2;
			goto exit;
		}

		if (forward)
			head = &lock->class->locks_after;
		else
			head = &lock->class->locks_before;

1052 1053 1054
		DEBUG_LOCKS_WARN_ON(!irqs_disabled());

		list_for_each_entry_rcu(entry, head, entry) {
1055
			if (!lock_accessed(entry)) {
1056
				unsigned int cq_depth;
1057
				mark_lock_accessed(entry, lock);
1058
				if (match(entry, data)) {
1059 1060 1061 1062 1063 1064 1065 1066 1067
					*target_entry = entry;
					ret = 0;
					goto exit;
				}

				if (__cq_enqueue(cq, (unsigned long)entry)) {
					ret = -1;
					goto exit;
				}
1068 1069 1070
				cq_depth = __cq_get_elem_count(cq);
				if (max_bfs_queue_depth < cq_depth)
					max_bfs_queue_depth = cq_depth;
1071 1072 1073 1074 1075 1076 1077
			}
		}
	}
exit:
	return ret;
}

1078
static inline int __bfs_forwards(struct lock_list *src_entry,
1079 1080 1081
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1082
{
1083
	return __bfs(src_entry, data, match, target_entry, 1);
1084 1085 1086

}

1087
static inline int __bfs_backwards(struct lock_list *src_entry,
1088 1089 1090
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1091
{
1092
	return __bfs(src_entry, data, match, target_entry, 0);
1093 1094 1095

}

P
Peter Zijlstra 已提交
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
/*
 * Recursive, forwards-direction lock-dependency checking, used for
 * both noncyclic checking and for hardirq-unsafe/softirq-unsafe
 * checking.
 */

/*
 * Print a dependency chain entry (this is only done when a deadlock
 * has been detected):
 */
static noinline int
1107
print_circular_bug_entry(struct lock_list *target, int depth)
P
Peter Zijlstra 已提交
1108 1109 1110 1111 1112
{
	if (debug_locks_silent)
		return 0;
	printk("\n-> #%u", depth);
	print_lock_name(target->class);
D
Dmitry Vyukov 已提交
1113
	printk(KERN_CONT ":\n");
P
Peter Zijlstra 已提交
1114 1115 1116 1117 1118
	print_stack_trace(&target->trace, 6);

	return 0;
}

1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143
static void
print_circular_lock_scenario(struct held_lock *src,
			     struct held_lock *tgt,
			     struct lock_list *prt)
{
	struct lock_class *source = hlock_class(src);
	struct lock_class *target = hlock_class(tgt);
	struct lock_class *parent = prt->class;

	/*
	 * A direct locking problem where unsafe_class lock is taken
	 * directly by safe_class lock, then all we need to show
	 * is the deadlock scenario, as it is obvious that the
	 * unsafe lock is taken under the safe lock.
	 *
	 * But if there is a chain instead, where the safe lock takes
	 * an intermediate lock (middle_class) where this lock is
	 * not the same as the safe lock, then the lock chain is
	 * used to describe the problem. Otherwise we would need
	 * to show a different CPU case for each link in the chain
	 * from the safe_class lock to the unsafe_class lock.
	 */
	if (parent != source) {
		printk("Chain exists of:\n  ");
		__print_lock_name(source);
D
Dmitry Vyukov 已提交
1144
		printk(KERN_CONT " --> ");
1145
		__print_lock_name(parent);
D
Dmitry Vyukov 已提交
1146
		printk(KERN_CONT " --> ");
1147
		__print_lock_name(target);
D
Dmitry Vyukov 已提交
1148
		printk(KERN_CONT "\n\n");
1149 1150
	}

1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185
	if (cross_lock(tgt->instance)) {
		printk(" Possible unsafe locking scenario by crosslock:\n\n");
		printk("       CPU0                    CPU1\n");
		printk("       ----                    ----\n");
		printk("  lock(");
		__print_lock_name(parent);
		printk(KERN_CONT ");\n");
		printk("  lock(");
		__print_lock_name(target);
		printk(KERN_CONT ");\n");
		printk("                               lock(");
		__print_lock_name(source);
		printk(KERN_CONT ");\n");
		printk("                               unlock(");
		__print_lock_name(target);
		printk(KERN_CONT ");\n");
		printk("\n *** DEADLOCK ***\n\n");
	} else {
		printk(" Possible unsafe locking scenario:\n\n");
		printk("       CPU0                    CPU1\n");
		printk("       ----                    ----\n");
		printk("  lock(");
		__print_lock_name(target);
		printk(KERN_CONT ");\n");
		printk("                               lock(");
		__print_lock_name(parent);
		printk(KERN_CONT ");\n");
		printk("                               lock(");
		__print_lock_name(target);
		printk(KERN_CONT ");\n");
		printk("  lock(");
		__print_lock_name(source);
		printk(KERN_CONT ");\n");
		printk("\n *** DEADLOCK ***\n\n");
	}
1186 1187
}

P
Peter Zijlstra 已提交
1188 1189 1190 1191 1192
/*
 * When a circular dependency is detected, print the
 * header first:
 */
static noinline int
1193 1194 1195
print_circular_bug_header(struct lock_list *entry, unsigned int depth,
			struct held_lock *check_src,
			struct held_lock *check_tgt)
P
Peter Zijlstra 已提交
1196 1197 1198
{
	struct task_struct *curr = current;

1199
	if (debug_locks_silent)
P
Peter Zijlstra 已提交
1200 1201
		return 0;

1202
	pr_warn("\n");
1203 1204
	pr_warn("======================================================\n");
	pr_warn("WARNING: possible circular locking dependency detected\n");
1205
	print_kernel_ident();
1206
	pr_warn("------------------------------------------------------\n");
1207
	pr_warn("%s/%d is trying to acquire lock:\n",
1208
		curr->comm, task_pid_nr(curr));
1209
	print_lock(check_src);
1210 1211 1212 1213 1214 1215

	if (cross_lock(check_tgt->instance))
		pr_warn("\nbut now in release context of a crosslock acquired at the following:\n");
	else
		pr_warn("\nbut task is already holding lock:\n");

1216
	print_lock(check_tgt);
1217 1218
	pr_warn("\nwhich lock already depends on the new lock.\n\n");
	pr_warn("\nthe existing dependency chain (in reverse order) is:\n");
P
Peter Zijlstra 已提交
1219 1220 1221 1222 1223 1224

	print_circular_bug_entry(entry, depth);

	return 0;
}

1225 1226 1227 1228 1229
static inline int class_equal(struct lock_list *entry, void *data)
{
	return entry->class == data;
}

1230 1231 1232
static noinline int print_circular_bug(struct lock_list *this,
				struct lock_list *target,
				struct held_lock *check_src,
1233 1234
				struct held_lock *check_tgt,
				struct stack_trace *trace)
P
Peter Zijlstra 已提交
1235 1236
{
	struct task_struct *curr = current;
1237
	struct lock_list *parent;
1238
	struct lock_list *first_parent;
1239
	int depth;
P
Peter Zijlstra 已提交
1240

1241
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
P
Peter Zijlstra 已提交
1242 1243
		return 0;

1244 1245 1246
	if (cross_lock(check_tgt->instance))
		this->trace = *trace;
	else if (!save_trace(&this->trace))
P
Peter Zijlstra 已提交
1247 1248
		return 0;

1249 1250
	depth = get_lock_depth(target);

1251
	print_circular_bug_header(target, depth, check_src, check_tgt);
1252 1253

	parent = get_lock_parent(target);
1254
	first_parent = parent;
1255 1256 1257 1258 1259

	while (parent) {
		print_circular_bug_entry(parent, --depth);
		parent = get_lock_parent(parent);
	}
P
Peter Zijlstra 已提交
1260 1261

	printk("\nother info that might help us debug this:\n\n");
1262 1263 1264
	print_circular_lock_scenario(check_src, check_tgt,
				     first_parent);

P
Peter Zijlstra 已提交
1265 1266 1267 1268 1269 1270 1271 1272
	lockdep_print_held_locks(curr);

	printk("\nstack backtrace:\n");
	dump_stack();

	return 0;
}

1273 1274 1275 1276 1277
static noinline int print_bfs_bug(int ret)
{
	if (!debug_locks_off_graph_unlock())
		return 0;

P
Peter Zijlstra 已提交
1278 1279 1280
	/*
	 * Breadth-first-search failed, graph got corrupted?
	 */
1281 1282 1283 1284 1285
	WARN(1, "lockdep bfs error:%d\n", ret);

	return 0;
}

1286
static int noop_count(struct lock_list *entry, void *data)
1287
{
1288 1289 1290
	(*(unsigned long *)data)++;
	return 0;
}
1291

1292
static unsigned long __lockdep_count_forward_deps(struct lock_list *this)
1293 1294 1295
{
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1296

1297
	__bfs_forwards(this, (void *)&count, noop_count, &target_entry);
1298

1299
	return count;
1300 1301 1302 1303
}
unsigned long lockdep_count_forward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1304 1305 1306 1307
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1308 1309

	local_irq_save(flags);
1310
	arch_spin_lock(&lockdep_lock);
1311
	ret = __lockdep_count_forward_deps(&this);
1312
	arch_spin_unlock(&lockdep_lock);
1313 1314 1315 1316 1317
	local_irq_restore(flags);

	return ret;
}

1318
static unsigned long __lockdep_count_backward_deps(struct lock_list *this)
1319
{
1320 1321
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1322

1323
	__bfs_backwards(this, (void *)&count, noop_count, &target_entry);
1324

1325
	return count;
1326 1327 1328 1329 1330
}

unsigned long lockdep_count_backward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1331 1332 1333 1334
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1335 1336

	local_irq_save(flags);
1337
	arch_spin_lock(&lockdep_lock);
1338
	ret = __lockdep_count_backward_deps(&this);
1339
	arch_spin_unlock(&lockdep_lock);
1340 1341 1342 1343 1344
	local_irq_restore(flags);

	return ret;
}

P
Peter Zijlstra 已提交
1345 1346 1347 1348 1349
/*
 * Prove that the dependency graph starting at <entry> can not
 * lead to <target>. Print an error and return 0 if it does.
 */
static noinline int
1350 1351
check_noncircular(struct lock_list *root, struct lock_class *target,
		struct lock_list **target_entry)
P
Peter Zijlstra 已提交
1352
{
1353
	int result;
P
Peter Zijlstra 已提交
1354

1355
	debug_atomic_inc(nr_cyclic_checks);
1356

1357
	result = __bfs_forwards(root, target, class_equal, target_entry);
I
Ingo Molnar 已提交
1358

1359 1360
	return result;
}
1361

1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
static noinline int
check_redundant(struct lock_list *root, struct lock_class *target,
		struct lock_list **target_entry)
{
	int result;

	debug_atomic_inc(nr_redundant_checks);

	result = __bfs_forwards(root, target, class_equal, target_entry);

	return result;
}

1375
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
1376 1377 1378 1379 1380 1381
/*
 * Forwards and backwards subgraph searching, for the purposes of
 * proving that two subgraphs can be connected by a new dependency
 * without creating any illegal irq-safe -> irq-unsafe lock dependency.
 */

1382 1383 1384 1385 1386 1387 1388
static inline int usage_match(struct lock_list *entry, void *bit)
{
	return entry->class->usage_mask & (1 << (enum lock_usage_bit)bit);
}



I
Ingo Molnar 已提交
1389 1390
/*
 * Find a node in the forwards-direction dependency sub-graph starting
1391
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1392
 *
1393 1394
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1395
 *
1396 1397
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1398
 */
1399 1400 1401
static int
find_usage_forwards(struct lock_list *root, enum lock_usage_bit bit,
			struct lock_list **target_entry)
I
Ingo Molnar 已提交
1402
{
1403
	int result;
I
Ingo Molnar 已提交
1404

1405
	debug_atomic_inc(nr_find_usage_forwards_checks);
I
Ingo Molnar 已提交
1406

1407 1408 1409
	result = __bfs_forwards(root, (void *)bit, usage_match, target_entry);

	return result;
I
Ingo Molnar 已提交
1410 1411 1412 1413
}

/*
 * Find a node in the backwards-direction dependency sub-graph starting
1414
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1415
 *
1416 1417
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1418
 *
1419 1420
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1421
 */
1422 1423 1424
static int
find_usage_backwards(struct lock_list *root, enum lock_usage_bit bit,
			struct lock_list **target_entry)
I
Ingo Molnar 已提交
1425
{
1426
	int result;
I
Ingo Molnar 已提交
1427

1428
	debug_atomic_inc(nr_find_usage_backwards_checks);
I
Ingo Molnar 已提交
1429

1430
	result = __bfs_backwards(root, (void *)bit, usage_match, target_entry);
D
Dave Jones 已提交
1431

1432
	return result;
I
Ingo Molnar 已提交
1433 1434
}

P
Peter Zijlstra 已提交
1435 1436 1437 1438 1439 1440
static void print_lock_class_header(struct lock_class *class, int depth)
{
	int bit;

	printk("%*s->", depth, "");
	print_lock_name(class);
D
Dmitry Vyukov 已提交
1441 1442
	printk(KERN_CONT " ops: %lu", class->ops);
	printk(KERN_CONT " {\n");
P
Peter Zijlstra 已提交
1443 1444 1445 1446 1447 1448

	for (bit = 0; bit < LOCK_USAGE_STATES; bit++) {
		if (class->usage_mask & (1 << bit)) {
			int len = depth;

			len += printk("%*s   %s", depth, "", usage_str[bit]);
D
Dmitry Vyukov 已提交
1449
			len += printk(KERN_CONT " at:\n");
P
Peter Zijlstra 已提交
1450 1451 1452 1453 1454
			print_stack_trace(class->usage_traces + bit, len);
		}
	}
	printk("%*s }\n", depth, "");

D
Dmitry Vyukov 已提交
1455 1456
	printk("%*s ... key      at: [<%p>] %pS\n",
		depth, "", class->key, class->key);
P
Peter Zijlstra 已提交
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478
}

/*
 * printk the shortest lock dependencies from @start to @end in reverse order:
 */
static void __used
print_shortest_lock_dependencies(struct lock_list *leaf,
				struct lock_list *root)
{
	struct lock_list *entry = leaf;
	int depth;

	/*compute depth from generated tree by BFS*/
	depth = get_lock_depth(leaf);

	do {
		print_lock_class_header(entry->class, depth);
		printk("%*s ... acquired at:\n", depth, "");
		print_stack_trace(&entry->trace, 2);
		printk("\n");

		if (depth == 0 && (entry != root)) {
1479
			printk("lockdep:%s bad path found in chain graph\n", __func__);
P
Peter Zijlstra 已提交
1480 1481 1482 1483 1484 1485 1486 1487 1488
			break;
		}

		entry = get_lock_parent(entry);
		depth--;
	} while (entry && (depth >= 0));

	return;
}
1489

1490 1491 1492
static void
print_irq_lock_scenario(struct lock_list *safe_entry,
			struct lock_list *unsafe_entry,
1493 1494
			struct lock_class *prev_class,
			struct lock_class *next_class)
1495 1496 1497
{
	struct lock_class *safe_class = safe_entry->class;
	struct lock_class *unsafe_class = unsafe_entry->class;
1498
	struct lock_class *middle_class = prev_class;
1499 1500

	if (middle_class == safe_class)
1501
		middle_class = next_class;
1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518

	/*
	 * A direct locking problem where unsafe_class lock is taken
	 * directly by safe_class lock, then all we need to show
	 * is the deadlock scenario, as it is obvious that the
	 * unsafe lock is taken under the safe lock.
	 *
	 * But if there is a chain instead, where the safe lock takes
	 * an intermediate lock (middle_class) where this lock is
	 * not the same as the safe lock, then the lock chain is
	 * used to describe the problem. Otherwise we would need
	 * to show a different CPU case for each link in the chain
	 * from the safe_class lock to the unsafe_class lock.
	 */
	if (middle_class != unsafe_class) {
		printk("Chain exists of:\n  ");
		__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1519
		printk(KERN_CONT " --> ");
1520
		__print_lock_name(middle_class);
D
Dmitry Vyukov 已提交
1521
		printk(KERN_CONT " --> ");
1522
		__print_lock_name(unsafe_class);
D
Dmitry Vyukov 已提交
1523
		printk(KERN_CONT "\n\n");
1524 1525 1526 1527 1528 1529 1530
	}

	printk(" Possible interrupt unsafe locking scenario:\n\n");
	printk("       CPU0                    CPU1\n");
	printk("       ----                    ----\n");
	printk("  lock(");
	__print_lock_name(unsafe_class);
D
Dmitry Vyukov 已提交
1531
	printk(KERN_CONT ");\n");
1532 1533 1534
	printk("                               local_irq_disable();\n");
	printk("                               lock(");
	__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1535
	printk(KERN_CONT ");\n");
1536 1537
	printk("                               lock(");
	__print_lock_name(middle_class);
D
Dmitry Vyukov 已提交
1538
	printk(KERN_CONT ");\n");
1539 1540 1541
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1542
	printk(KERN_CONT ");\n");
1543 1544 1545
	printk("\n *** DEADLOCK ***\n\n");
}

I
Ingo Molnar 已提交
1546 1547
static int
print_bad_irq_dependency(struct task_struct *curr,
1548 1549 1550 1551
			 struct lock_list *prev_root,
			 struct lock_list *next_root,
			 struct lock_list *backwards_entry,
			 struct lock_list *forwards_entry,
I
Ingo Molnar 已提交
1552 1553 1554 1555 1556 1557
			 struct held_lock *prev,
			 struct held_lock *next,
			 enum lock_usage_bit bit1,
			 enum lock_usage_bit bit2,
			 const char *irqclass)
{
1558
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1559 1560
		return 0;

1561
	pr_warn("\n");
1562 1563
	pr_warn("=====================================================\n");
	pr_warn("WARNING: %s-safe -> %s-unsafe lock order detected\n",
I
Ingo Molnar 已提交
1564
		irqclass, irqclass);
1565
	print_kernel_ident();
1566
	pr_warn("-----------------------------------------------------\n");
1567
	pr_warn("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] is trying to acquire:\n",
1568
		curr->comm, task_pid_nr(curr),
I
Ingo Molnar 已提交
1569 1570 1571 1572 1573 1574
		curr->hardirq_context, hardirq_count() >> HARDIRQ_SHIFT,
		curr->softirq_context, softirq_count() >> SOFTIRQ_SHIFT,
		curr->hardirqs_enabled,
		curr->softirqs_enabled);
	print_lock(next);

1575
	pr_warn("\nand this task is already holding:\n");
I
Ingo Molnar 已提交
1576
	print_lock(prev);
1577
	pr_warn("which would create a new lock dependency:\n");
D
Dave Jones 已提交
1578
	print_lock_name(hlock_class(prev));
1579
	pr_cont(" ->");
D
Dave Jones 已提交
1580
	print_lock_name(hlock_class(next));
1581
	pr_cont("\n");
I
Ingo Molnar 已提交
1582

1583
	pr_warn("\nbut this new dependency connects a %s-irq-safe lock:\n",
I
Ingo Molnar 已提交
1584
		irqclass);
1585
	print_lock_name(backwards_entry->class);
1586
	pr_warn("\n... which became %s-irq-safe at:\n", irqclass);
I
Ingo Molnar 已提交
1587

1588
	print_stack_trace(backwards_entry->class->usage_traces + bit1, 1);
I
Ingo Molnar 已提交
1589

1590
	pr_warn("\nto a %s-irq-unsafe lock:\n", irqclass);
1591
	print_lock_name(forwards_entry->class);
1592 1593
	pr_warn("\n... which became %s-irq-unsafe at:\n", irqclass);
	pr_warn("...");
I
Ingo Molnar 已提交
1594

1595
	print_stack_trace(forwards_entry->class->usage_traces + bit2, 1);
I
Ingo Molnar 已提交
1596

1597
	pr_warn("\nother info that might help us debug this:\n\n");
1598 1599
	print_irq_lock_scenario(backwards_entry, forwards_entry,
				hlock_class(prev), hlock_class(next));
1600

I
Ingo Molnar 已提交
1601 1602
	lockdep_print_held_locks(curr);

1603
	pr_warn("\nthe dependencies between %s-irq-safe lock and the holding lock:\n", irqclass);
1604 1605 1606
	if (!save_trace(&prev_root->trace))
		return 0;
	print_shortest_lock_dependencies(backwards_entry, prev_root);
I
Ingo Molnar 已提交
1607

1608 1609
	pr_warn("\nthe dependencies between the lock to be acquired");
	pr_warn(" and %s-irq-unsafe lock:\n", irqclass);
1610 1611 1612
	if (!save_trace(&next_root->trace))
		return 0;
	print_shortest_lock_dependencies(forwards_entry, next_root);
I
Ingo Molnar 已提交
1613

1614
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625
	dump_stack();

	return 0;
}

static int
check_usage(struct task_struct *curr, struct held_lock *prev,
	    struct held_lock *next, enum lock_usage_bit bit_backwards,
	    enum lock_usage_bit bit_forwards, const char *irqclass)
{
	int ret;
1626
	struct lock_list this, that;
1627
	struct lock_list *uninitialized_var(target_entry);
1628
	struct lock_list *uninitialized_var(target_entry1);
1629 1630 1631 1632 1633

	this.parent = NULL;

	this.class = hlock_class(prev);
	ret = find_usage_backwards(&this, bit_backwards, &target_entry);
P
Peter Zijlstra 已提交
1634 1635 1636 1637
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
1638

1639 1640 1641
	that.parent = NULL;
	that.class = hlock_class(next);
	ret = find_usage_forwards(&that, bit_forwards, &target_entry1);
P
Peter Zijlstra 已提交
1642 1643 1644 1645
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
1646

1647 1648 1649
	return print_bad_irq_dependency(curr, &this, &that,
			target_entry, target_entry1,
			prev, next,
I
Ingo Molnar 已提交
1650 1651 1652
			bit_backwards, bit_forwards, irqclass);
}

1653 1654
static const char *state_names[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1655
	__stringify(__STATE),
1656 1657 1658 1659 1660 1661
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static const char *state_rnames[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1662
	__stringify(__STATE)"-READ",
1663 1664 1665 1666 1667
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static inline const char *state_name(enum lock_usage_bit bit)
P
Peter Zijlstra 已提交
1668
{
1669 1670
	return (bit & 1) ? state_rnames[bit >> 2] : state_names[bit >> 2];
}
P
Peter Zijlstra 已提交
1671

1672 1673
static int exclusive_bit(int new_bit)
{
P
Peter Zijlstra 已提交
1674
	/*
1675 1676 1677 1678 1679 1680 1681 1682
	 * USED_IN
	 * USED_IN_READ
	 * ENABLED
	 * ENABLED_READ
	 *
	 * bit 0 - write/read
	 * bit 1 - used_in/enabled
	 * bit 2+  state
P
Peter Zijlstra 已提交
1683
	 */
1684 1685 1686

	int state = new_bit & ~3;
	int dir = new_bit & 2;
P
Peter Zijlstra 已提交
1687 1688

	/*
1689
	 * keep state, bit flip the direction and strip read.
P
Peter Zijlstra 已提交
1690
	 */
1691 1692 1693 1694 1695 1696
	return state | (dir ^ 2);
}

static int check_irq_usage(struct task_struct *curr, struct held_lock *prev,
			   struct held_lock *next, enum lock_usage_bit bit)
{
P
Peter Zijlstra 已提交
1697
	/*
1698 1699
	 * Prove that the new dependency does not connect a hardirq-safe
	 * lock with a hardirq-unsafe lock - to achieve this we search
P
Peter Zijlstra 已提交
1700 1701 1702
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1703 1704
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
P
Peter Zijlstra 已提交
1705 1706
		return 0;

1707 1708
	bit++; /* _READ */

1709
	/*
1710 1711
	 * Prove that the new dependency does not connect a hardirq-safe-read
	 * lock with a hardirq-unsafe lock - to achieve this we search
1712 1713 1714
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1715 1716
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
1717 1718
		return 0;

1719 1720 1721 1722 1723 1724 1725 1726 1727
	return 1;
}

static int
check_prev_add_irq(struct task_struct *curr, struct held_lock *prev,
		struct held_lock *next)
{
#define LOCKDEP_STATE(__STATE)						\
	if (!check_irq_usage(curr, prev, next, LOCK_USED_IN_##__STATE))	\
1728
		return 0;
1729 1730
#include "lockdep_states.h"
#undef LOCKDEP_STATE
1731

P
Peter Zijlstra 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760
	return 1;
}

static void inc_chains(void)
{
	if (current->hardirq_context)
		nr_hardirq_chains++;
	else {
		if (current->softirq_context)
			nr_softirq_chains++;
		else
			nr_process_chains++;
	}
}

#else

static inline int
check_prev_add_irq(struct task_struct *curr, struct held_lock *prev,
		struct held_lock *next)
{
	return 1;
}

static inline void inc_chains(void)
{
	nr_process_chains++;
}

I
Ingo Molnar 已提交
1761 1762
#endif

1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774
static void
print_deadlock_scenario(struct held_lock *nxt,
			     struct held_lock *prv)
{
	struct lock_class *next = hlock_class(nxt);
	struct lock_class *prev = hlock_class(prv);

	printk(" Possible unsafe locking scenario:\n\n");
	printk("       CPU0\n");
	printk("       ----\n");
	printk("  lock(");
	__print_lock_name(prev);
D
Dmitry Vyukov 已提交
1775
	printk(KERN_CONT ");\n");
1776 1777
	printk("  lock(");
	__print_lock_name(next);
D
Dmitry Vyukov 已提交
1778
	printk(KERN_CONT ");\n");
1779 1780 1781 1782
	printk("\n *** DEADLOCK ***\n\n");
	printk(" May be due to missing lock nesting notation\n\n");
}

I
Ingo Molnar 已提交
1783 1784 1785 1786
static int
print_deadlock_bug(struct task_struct *curr, struct held_lock *prev,
		   struct held_lock *next)
{
1787
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1788 1789
		return 0;

1790
	pr_warn("\n");
1791 1792
	pr_warn("============================================\n");
	pr_warn("WARNING: possible recursive locking detected\n");
1793
	print_kernel_ident();
1794
	pr_warn("--------------------------------------------\n");
1795
	pr_warn("%s/%d is trying to acquire lock:\n",
1796
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
1797
	print_lock(next);
1798
	pr_warn("\nbut task is already holding lock:\n");
I
Ingo Molnar 已提交
1799 1800
	print_lock(prev);

1801
	pr_warn("\nother info that might help us debug this:\n");
1802
	print_deadlock_scenario(next, prev);
I
Ingo Molnar 已提交
1803 1804
	lockdep_print_held_locks(curr);

1805
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
	dump_stack();

	return 0;
}

/*
 * Check whether we are holding such a class already.
 *
 * (Note that this has to be done separately, because the graph cannot
 * detect such classes of deadlocks.)
 *
 * Returns: 0 on deadlock detected, 1 on OK, 2 on recursive read
 */
static int
check_deadlock(struct task_struct *curr, struct held_lock *next,
	       struct lockdep_map *next_instance, int read)
{
	struct held_lock *prev;
P
Peter Zijlstra 已提交
1824
	struct held_lock *nest = NULL;
I
Ingo Molnar 已提交
1825 1826 1827 1828
	int i;

	for (i = 0; i < curr->lockdep_depth; i++) {
		prev = curr->held_locks + i;
P
Peter Zijlstra 已提交
1829 1830 1831 1832

		if (prev->instance == next->nest_lock)
			nest = prev;

D
Dave Jones 已提交
1833
		if (hlock_class(prev) != hlock_class(next))
I
Ingo Molnar 已提交
1834
			continue;
P
Peter Zijlstra 已提交
1835

I
Ingo Molnar 已提交
1836 1837
		/*
		 * Allow read-after-read recursion of the same
1838
		 * lock class (i.e. read_lock(lock)+read_lock(lock)):
I
Ingo Molnar 已提交
1839
		 */
1840
		if ((read == 2) && prev->read)
I
Ingo Molnar 已提交
1841
			return 2;
P
Peter Zijlstra 已提交
1842 1843 1844 1845 1846 1847 1848 1849

		/*
		 * We're holding the nest_lock, which serializes this lock's
		 * nesting behaviour.
		 */
		if (nest)
			return 2;

1850 1851 1852
		if (cross_lock(prev->instance))
			continue;

I
Ingo Molnar 已提交
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
		return print_deadlock_bug(curr, prev, next);
	}
	return 1;
}

/*
 * There was a chain-cache miss, and we are about to add a new dependency
 * to a previous lock. We recursively validate the following rules:
 *
 *  - would the adding of the <prev> -> <next> dependency create a
 *    circular dependency in the graph? [== circular deadlock]
 *
 *  - does the new prev->next dependency connect any hardirq-safe lock
 *    (in the full backwards-subgraph starting at <prev>) with any
 *    hardirq-unsafe lock (in the full forwards-subgraph starting at
 *    <next>)? [== illegal lock inversion with hardirq contexts]
 *
 *  - does the new prev->next dependency connect any softirq-safe lock
 *    (in the full backwards-subgraph starting at <prev>) with any
 *    softirq-unsafe lock (in the full forwards-subgraph starting at
 *    <next>)? [== illegal lock inversion with softirq contexts]
 *
 * any of these scenarios could lead to a deadlock.
 *
 * Then if all the validations pass, we add the forwards and backwards
 * dependency.
 */
static int
check_prev_add(struct task_struct *curr, struct held_lock *prev,
1882 1883
	       struct held_lock *next, int distance, struct stack_trace *trace,
	       int (*save)(struct stack_trace *trace))
I
Ingo Molnar 已提交
1884
{
1885
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
1886
	struct lock_list *entry;
1887
	struct lock_list this;
1888
	int ret;
I
Ingo Molnar 已提交
1889 1890 1891 1892 1893 1894 1895 1896 1897 1898

	/*
	 * Prove that the new <prev> -> <next> dependency would not
	 * create a circular dependency in the graph. (We do this by
	 * forward-recursing into the graph starting at <next>, and
	 * checking whether we can reach <prev>.)
	 *
	 * We are using global variables to control the recursion, to
	 * keep the stackframe size of the recursive functions low:
	 */
1899 1900 1901
	this.class = hlock_class(next);
	this.parent = NULL;
	ret = check_noncircular(&this, hlock_class(prev), &target_entry);
1902 1903 1904 1905 1906 1907 1908 1909 1910
	if (unlikely(!ret)) {
		if (!trace->entries) {
			/*
			 * If @save fails here, the printing might trigger
			 * a WARN but because of the !nr_entries it should
			 * not do bad things.
			 */
			save(trace);
		}
1911
		return print_circular_bug(&this, target_entry, next, prev, trace);
1912
	}
1913 1914
	else if (unlikely(ret < 0))
		return print_bfs_bug(ret);
1915

P
Peter Zijlstra 已提交
1916
	if (!check_prev_add_irq(curr, prev, next))
I
Ingo Molnar 已提交
1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
		return 0;

	/*
	 * For recursive read-locks we do all the dependency checks,
	 * but we dont store read-triggered dependencies (only
	 * write-triggered dependencies). This ensures that only the
	 * write-side dependencies matter, and that if for example a
	 * write-lock never takes any other locks, then the reads are
	 * equivalent to a NOP.
	 */
	if (next->read == 2 || prev->read == 2)
		return 1;
	/*
	 * Is the <prev> -> <next> dependency already present?
	 *
	 * (this may occur even though this is a new chain: consider
	 *  e.g. the L1 -> L2 -> L3 -> L4 and the L5 -> L1 -> L2 -> L3
	 *  chains - the second one will be new, but L1 already has
	 *  L2 added to its dependency list, due to the first chain.)
	 */
D
Dave Jones 已提交
1937 1938
	list_for_each_entry(entry, &hlock_class(prev)->locks_after, entry) {
		if (entry->class == hlock_class(next)) {
1939 1940
			if (distance == 1)
				entry->distance = 1;
1941
			return 1;
1942
		}
I
Ingo Molnar 已提交
1943 1944
	}

1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
	/*
	 * Is the <prev> -> <next> link redundant?
	 */
	this.class = hlock_class(prev);
	this.parent = NULL;
	ret = check_redundant(&this, hlock_class(next), &target_entry);
	if (!ret) {
		debug_atomic_inc(nr_redundant);
		return 2;
	}
	if (ret < 0)
		return print_bfs_bug(ret);


1959
	if (!trace->entries && !save(trace))
1960
		return 0;
Y
Yong Zhang 已提交
1961

I
Ingo Molnar 已提交
1962 1963 1964 1965
	/*
	 * Ok, all validations passed, add the new lock
	 * to the previous lock's dependency list:
	 */
1966
	ret = add_lock_to_list(hlock_class(next),
D
Dave Jones 已提交
1967
			       &hlock_class(prev)->locks_after,
1968
			       next->acquire_ip, distance, trace);
1969

I
Ingo Molnar 已提交
1970 1971
	if (!ret)
		return 0;
1972

1973
	ret = add_lock_to_list(hlock_class(prev),
D
Dave Jones 已提交
1974
			       &hlock_class(next)->locks_before,
1975
			       next->acquire_ip, distance, trace);
1976 1977
	if (!ret)
		return 0;
I
Ingo Molnar 已提交
1978

1979
	return 2;
P
Peter Zijlstra 已提交
1980
}
I
Ingo Molnar 已提交
1981

P
Peter Zijlstra 已提交
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992
/*
 * Add the dependency to all directly-previous locks that are 'relevant'.
 * The ones that are relevant are (in increasing distance from curr):
 * all consecutive trylock entries and the final non-trylock entry - or
 * the end of this context's lock-chain - whichever comes first.
 */
static int
check_prevs_add(struct task_struct *curr, struct held_lock *next)
{
	int depth = curr->lockdep_depth;
	struct held_lock *hlock;
1993 1994 1995 1996 1997 1998
	struct stack_trace trace = {
		.nr_entries = 0,
		.max_entries = 0,
		.entries = NULL,
		.skip = 0,
	};
1999

I
Ingo Molnar 已提交
2000
	/*
P
Peter Zijlstra 已提交
2001 2002 2003
	 * Debugging checks.
	 *
	 * Depth must not be zero for a non-head lock:
I
Ingo Molnar 已提交
2004
	 */
P
Peter Zijlstra 已提交
2005 2006
	if (!depth)
		goto out_bug;
I
Ingo Molnar 已提交
2007
	/*
P
Peter Zijlstra 已提交
2008 2009
	 * At least two relevant locks must exist for this
	 * to be a head:
I
Ingo Molnar 已提交
2010
	 */
P
Peter Zijlstra 已提交
2011 2012 2013
	if (curr->held_locks[depth].irq_context !=
			curr->held_locks[depth-1].irq_context)
		goto out_bug;
2014

P
Peter Zijlstra 已提交
2015 2016
	for (;;) {
		int distance = curr->lockdep_depth - depth + 1;
2017
		hlock = curr->held_locks + depth - 1;
P
Peter Zijlstra 已提交
2018
		/*
2019 2020
		 * Only non-crosslock entries get new dependencies added.
		 * Crosslock entries will be added by commit later:
P
Peter Zijlstra 已提交
2021
		 */
2022
		if (!cross_lock(hlock->instance)) {
2023
			/*
2024 2025
			 * Only non-recursive-read entries get new dependencies
			 * added:
2026
			 */
2027 2028
			if (hlock->read != 2 && hlock->check) {
				int ret = check_prev_add(curr, hlock, next,
2029
							 distance, &trace, save_trace);
2030 2031
				if (!ret)
					return 0;
2032

2033 2034 2035 2036 2037 2038 2039 2040 2041
				/*
				 * Stop after the first non-trylock entry,
				 * as non-trylock entries have added their
				 * own direct dependencies already, so this
				 * lock is connected to them indirectly:
				 */
				if (!hlock->trylock)
					break;
			}
2042
		}
P
Peter Zijlstra 已提交
2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054
		depth--;
		/*
		 * End of lock-stack?
		 */
		if (!depth)
			break;
		/*
		 * Stop the search if we cross into another context:
		 */
		if (curr->held_locks[depth].irq_context !=
				curr->held_locks[depth-1].irq_context)
			break;
I
Ingo Molnar 已提交
2055
	}
P
Peter Zijlstra 已提交
2056 2057 2058 2059
	return 1;
out_bug:
	if (!debug_locks_off_graph_unlock())
		return 0;
I
Ingo Molnar 已提交
2060

P
Peter Zijlstra 已提交
2061 2062 2063 2064 2065
	/*
	 * Clearly we all shouldn't be here, but since we made it we
	 * can reliable say we messed up our state. See the above two
	 * gotos for reasons why we could possibly end up here.
	 */
P
Peter Zijlstra 已提交
2066
	WARN_ON(1);
I
Ingo Molnar 已提交
2067

P
Peter Zijlstra 已提交
2068
	return 0;
I
Ingo Molnar 已提交
2069 2070
}

P
Peter Zijlstra 已提交
2071
unsigned long nr_lock_chains;
2072
struct lock_chain lock_chains[MAX_LOCKDEP_CHAINS];
2073
int nr_chain_hlocks;
2074 2075 2076 2077 2078 2079
static u16 chain_hlocks[MAX_LOCKDEP_CHAIN_HLOCKS];

struct lock_class *lock_chain_get_class(struct lock_chain *chain, int i)
{
	return lock_classes + chain_hlocks[chain->base + i];
}
P
Peter Zijlstra 已提交
2080

2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099
/*
 * Returns the index of the first held_lock of the current chain
 */
static inline int get_first_held_lock(struct task_struct *curr,
					struct held_lock *hlock)
{
	int i;
	struct held_lock *hlock_curr;

	for (i = curr->lockdep_depth - 1; i >= 0; i--) {
		hlock_curr = curr->held_locks + i;
		if (hlock_curr->irq_context != hlock->irq_context)
			break;

	}

	return ++i;
}

2100
#ifdef CONFIG_DEBUG_LOCKDEP
2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153
/*
 * Returns the next chain_key iteration
 */
static u64 print_chain_key_iteration(int class_idx, u64 chain_key)
{
	u64 new_chain_key = iterate_chain_key(chain_key, class_idx);

	printk(" class_idx:%d -> chain_key:%016Lx",
		class_idx,
		(unsigned long long)new_chain_key);
	return new_chain_key;
}

static void
print_chain_keys_held_locks(struct task_struct *curr, struct held_lock *hlock_next)
{
	struct held_lock *hlock;
	u64 chain_key = 0;
	int depth = curr->lockdep_depth;
	int i;

	printk("depth: %u\n", depth + 1);
	for (i = get_first_held_lock(curr, hlock_next); i < depth; i++) {
		hlock = curr->held_locks + i;
		chain_key = print_chain_key_iteration(hlock->class_idx, chain_key);

		print_lock(hlock);
	}

	print_chain_key_iteration(hlock_next->class_idx, chain_key);
	print_lock(hlock_next);
}

static void print_chain_keys_chain(struct lock_chain *chain)
{
	int i;
	u64 chain_key = 0;
	int class_id;

	printk("depth: %u\n", chain->depth);
	for (i = 0; i < chain->depth; i++) {
		class_id = chain_hlocks[chain->base + i];
		chain_key = print_chain_key_iteration(class_id + 1, chain_key);

		print_lock_name(lock_classes + class_id);
		printk("\n");
	}
}

static void print_collision(struct task_struct *curr,
			struct held_lock *hlock_next,
			struct lock_chain *chain)
{
2154
	pr_warn("\n");
2155 2156
	pr_warn("============================\n");
	pr_warn("WARNING: chain_key collision\n");
2157
	print_kernel_ident();
2158
	pr_warn("----------------------------\n");
2159 2160
	pr_warn("%s/%d: ", current->comm, task_pid_nr(current));
	pr_warn("Hash chain already cached but the contents don't match!\n");
2161

2162
	pr_warn("Held locks:");
2163 2164
	print_chain_keys_held_locks(curr, hlock_next);

2165
	pr_warn("Locks in cached chain:");
2166 2167
	print_chain_keys_chain(chain);

2168
	pr_warn("\nstack backtrace:\n");
2169 2170
	dump_stack();
}
2171
#endif
2172

2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187
/*
 * Checks whether the chain and the current held locks are consistent
 * in depth and also in content. If they are not it most likely means
 * that there was a collision during the calculation of the chain_key.
 * Returns: 0 not passed, 1 passed
 */
static int check_no_collision(struct task_struct *curr,
			struct held_lock *hlock,
			struct lock_chain *chain)
{
#ifdef CONFIG_DEBUG_LOCKDEP
	int i, j, id;

	i = get_first_held_lock(curr, hlock);

2188 2189
	if (DEBUG_LOCKS_WARN_ON(chain->depth != curr->lockdep_depth - (i - 1))) {
		print_collision(curr, hlock, chain);
2190
		return 0;
2191
	}
2192 2193 2194 2195

	for (j = 0; j < chain->depth - 1; j++, i++) {
		id = curr->held_locks[i].class_idx - 1;

2196 2197
		if (DEBUG_LOCKS_WARN_ON(chain_hlocks[chain->base + j] != id)) {
			print_collision(curr, hlock, chain);
2198
			return 0;
2199
		}
2200 2201 2202 2203 2204
	}
#endif
	return 1;
}

2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274
/*
 * This is for building a chain between just two different classes,
 * instead of adding a new hlock upon current, which is done by
 * add_chain_cache().
 *
 * This can be called in any context with two classes, while
 * add_chain_cache() must be done within the lock owener's context
 * since it uses hlock which might be racy in another context.
 */
static inline int add_chain_cache_classes(unsigned int prev,
					  unsigned int next,
					  unsigned int irq_context,
					  u64 chain_key)
{
	struct hlist_head *hash_head = chainhashentry(chain_key);
	struct lock_chain *chain;

	/*
	 * Allocate a new chain entry from the static array, and add
	 * it to the hash:
	 */

	/*
	 * We might need to take the graph lock, ensure we've got IRQs
	 * disabled to make this an IRQ-safe lock.. for recursion reasons
	 * lockdep won't complain about its own locking errors.
	 */
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;

	if (unlikely(nr_lock_chains >= MAX_LOCKDEP_CHAINS)) {
		if (!debug_locks_off_graph_unlock())
			return 0;

		print_lockdep_off("BUG: MAX_LOCKDEP_CHAINS too low!");
		dump_stack();
		return 0;
	}

	chain = lock_chains + nr_lock_chains++;
	chain->chain_key = chain_key;
	chain->irq_context = irq_context;
	chain->depth = 2;
	if (likely(nr_chain_hlocks + chain->depth <= MAX_LOCKDEP_CHAIN_HLOCKS)) {
		chain->base = nr_chain_hlocks;
		nr_chain_hlocks += chain->depth;
		chain_hlocks[chain->base] = prev - 1;
		chain_hlocks[chain->base + 1] = next -1;
	}
#ifdef CONFIG_DEBUG_LOCKDEP
	/*
	 * Important for check_no_collision().
	 */
	else {
		if (!debug_locks_off_graph_unlock())
			return 0;

		print_lockdep_off("BUG: MAX_LOCKDEP_CHAIN_HLOCKS too low!");
		dump_stack();
		return 0;
	}
#endif

	hlist_add_head_rcu(&chain->entry, hash_head);
	debug_atomic_inc(chain_lookup_misses);
	inc_chains();

	return 1;
}

I
Ingo Molnar 已提交
2275
/*
2276 2277 2278 2279 2280
 * Adds a dependency chain into chain hashtable. And must be called with
 * graph_lock held.
 *
 * Return 0 if fail, and graph_lock is released.
 * Return 1 if succeed, with graph_lock held.
I
Ingo Molnar 已提交
2281
 */
2282 2283 2284
static inline int add_chain_cache(struct task_struct *curr,
				  struct held_lock *hlock,
				  u64 chain_key)
I
Ingo Molnar 已提交
2285
{
D
Dave Jones 已提交
2286
	struct lock_class *class = hlock_class(hlock);
2287
	struct hlist_head *hash_head = chainhashentry(chain_key);
I
Ingo Molnar 已提交
2288
	struct lock_chain *chain;
2289
	int i, j;
I
Ingo Molnar 已提交
2290

2291 2292 2293 2294 2295
	/*
	 * Allocate a new chain entry from the static array, and add
	 * it to the hash:
	 */

P
Peter Zijlstra 已提交
2296 2297 2298 2299 2300
	/*
	 * We might need to take the graph lock, ensure we've got IRQs
	 * disabled to make this an IRQ-safe lock.. for recursion reasons
	 * lockdep won't complain about its own locking errors.
	 */
2301 2302
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;
2303

I
Ingo Molnar 已提交
2304
	if (unlikely(nr_lock_chains >= MAX_LOCKDEP_CHAINS)) {
2305 2306 2307
		if (!debug_locks_off_graph_unlock())
			return 0;

2308
		print_lockdep_off("BUG: MAX_LOCKDEP_CHAINS too low!");
2309
		dump_stack();
I
Ingo Molnar 已提交
2310 2311 2312 2313
		return 0;
	}
	chain = lock_chains + nr_lock_chains++;
	chain->chain_key = chain_key;
2314
	chain->irq_context = hlock->irq_context;
2315
	i = get_first_held_lock(curr, hlock);
2316
	chain->depth = curr->lockdep_depth + 1 - i;
2317 2318 2319 2320 2321

	BUILD_BUG_ON((1UL << 24) <= ARRAY_SIZE(chain_hlocks));
	BUILD_BUG_ON((1UL << 6)  <= ARRAY_SIZE(curr->held_locks));
	BUILD_BUG_ON((1UL << 8*sizeof(chain_hlocks[0])) <= ARRAY_SIZE(lock_classes));

2322 2323
	if (likely(nr_chain_hlocks + chain->depth <= MAX_LOCKDEP_CHAIN_HLOCKS)) {
		chain->base = nr_chain_hlocks;
2324
		for (j = 0; j < chain->depth - 1; j++, i++) {
D
Dave Jones 已提交
2325
			int lock_id = curr->held_locks[i].class_idx - 1;
2326 2327 2328 2329
			chain_hlocks[chain->base + j] = lock_id;
		}
		chain_hlocks[chain->base + j] = class - lock_classes;
	}
2330 2331 2332 2333 2334 2335 2336 2337 2338

	if (nr_chain_hlocks < MAX_LOCKDEP_CHAIN_HLOCKS)
		nr_chain_hlocks += chain->depth;

#ifdef CONFIG_DEBUG_LOCKDEP
	/*
	 * Important for check_no_collision().
	 */
	if (unlikely(nr_chain_hlocks > MAX_LOCKDEP_CHAIN_HLOCKS)) {
2339
		if (!debug_locks_off_graph_unlock())
2340 2341 2342 2343 2344 2345 2346 2347
			return 0;

		print_lockdep_off("BUG: MAX_LOCKDEP_CHAIN_HLOCKS too low!");
		dump_stack();
		return 0;
	}
#endif

2348
	hlist_add_head_rcu(&chain->entry, hash_head);
2349
	debug_atomic_inc(chain_lookup_misses);
P
Peter Zijlstra 已提交
2350 2351 2352 2353 2354
	inc_chains();

	return 1;
}

2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426
/*
 * Look up a dependency chain.
 */
static inline struct lock_chain *lookup_chain_cache(u64 chain_key)
{
	struct hlist_head *hash_head = chainhashentry(chain_key);
	struct lock_chain *chain;

	/*
	 * We can walk it lock-free, because entries only get added
	 * to the hash:
	 */
	hlist_for_each_entry_rcu(chain, hash_head, entry) {
		if (chain->chain_key == chain_key) {
			debug_atomic_inc(chain_lookup_hits);
			return chain;
		}
	}
	return NULL;
}

/*
 * If the key is not present yet in dependency chain cache then
 * add it and return 1 - in this case the new dependency chain is
 * validated. If the key is already hashed, return 0.
 * (On return with 1 graph_lock is held.)
 */
static inline int lookup_chain_cache_add(struct task_struct *curr,
					 struct held_lock *hlock,
					 u64 chain_key)
{
	struct lock_class *class = hlock_class(hlock);
	struct lock_chain *chain = lookup_chain_cache(chain_key);

	if (chain) {
cache_hit:
		if (!check_no_collision(curr, hlock, chain))
			return 0;

		if (very_verbose(class)) {
			printk("\nhash chain already cached, key: "
					"%016Lx tail class: [%p] %s\n",
					(unsigned long long)chain_key,
					class->key, class->name);
		}

		return 0;
	}

	if (very_verbose(class)) {
		printk("\nnew hash chain, key: %016Lx tail class: [%p] %s\n",
			(unsigned long long)chain_key, class->key, class->name);
	}

	if (!graph_lock())
		return 0;

	/*
	 * We have to walk the chain again locked - to avoid duplicates:
	 */
	chain = lookup_chain_cache(chain_key);
	if (chain) {
		graph_unlock();
		goto cache_hit;
	}

	if (!add_chain_cache(curr, hlock, chain_key))
		return 0;

	return 1;
}

P
Peter Zijlstra 已提交
2427
static int validate_chain(struct task_struct *curr, struct lockdep_map *lock,
2428
		struct held_lock *hlock, int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2429 2430 2431 2432 2433 2434 2435 2436
{
	/*
	 * Trylock needs to maintain the stack of held locks, but it
	 * does not add new dependencies, because trylock can be done
	 * in any order.
	 *
	 * We look up the chain_key and do the O(N^2) check and update of
	 * the dependencies only if this is a new dependency chain.
2437
	 * (If lookup_chain_cache_add() return with 1 it acquires
P
Peter Zijlstra 已提交
2438 2439
	 * graph_lock for us)
	 */
2440
	if (!hlock->trylock && hlock->check &&
2441
	    lookup_chain_cache_add(curr, hlock, chain_key)) {
P
Peter Zijlstra 已提交
2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468
		/*
		 * Check whether last held lock:
		 *
		 * - is irq-safe, if this lock is irq-unsafe
		 * - is softirq-safe, if this lock is hardirq-unsafe
		 *
		 * And check whether the new lock's dependency graph
		 * could lead back to the previous lock.
		 *
		 * any of these scenarios could lead to a deadlock. If
		 * All validations
		 */
		int ret = check_deadlock(curr, hlock, lock, hlock->read);

		if (!ret)
			return 0;
		/*
		 * Mark recursive read, as we jump over it when
		 * building dependencies (just like we jump over
		 * trylock entries):
		 */
		if (ret == 2)
			hlock->read = 2;
		/*
		 * Add dependency only if this lock is not the head
		 * of the chain, and if it's not a secondary read-lock:
		 */
2469
		if (!chain_head && ret != 2) {
P
Peter Zijlstra 已提交
2470 2471
			if (!check_prevs_add(curr, hlock))
				return 0;
2472 2473
		}

P
Peter Zijlstra 已提交
2474
		graph_unlock();
2475 2476
	} else {
		/* after lookup_chain_cache_add(): */
P
Peter Zijlstra 已提交
2477 2478
		if (unlikely(!debug_locks))
			return 0;
2479
	}
I
Ingo Molnar 已提交
2480 2481 2482

	return 1;
}
P
Peter Zijlstra 已提交
2483 2484 2485
#else
static inline int validate_chain(struct task_struct *curr,
	       	struct lockdep_map *lock, struct held_lock *hlock,
2486
		int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2487 2488 2489
{
	return 1;
}
2490
#endif
I
Ingo Molnar 已提交
2491 2492 2493 2494 2495

/*
 * We are building curr_chain_key incrementally, so double-check
 * it from scratch, to make sure that it's done correctly:
 */
2496
static void check_chain_key(struct task_struct *curr)
I
Ingo Molnar 已提交
2497 2498 2499
{
#ifdef CONFIG_DEBUG_LOCKDEP
	struct held_lock *hlock, *prev_hlock = NULL;
2500
	unsigned int i;
I
Ingo Molnar 已提交
2501 2502 2503 2504 2505 2506
	u64 chain_key = 0;

	for (i = 0; i < curr->lockdep_depth; i++) {
		hlock = curr->held_locks + i;
		if (chain_key != hlock->prev_chain_key) {
			debug_locks_off();
P
Peter Zijlstra 已提交
2507 2508 2509 2510
			/*
			 * We got mighty confused, our chain keys don't match
			 * with what we expect, someone trample on our task state?
			 */
2511
			WARN(1, "hm#1, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2512 2513 2514 2515 2516
				curr->lockdep_depth, i,
				(unsigned long long)chain_key,
				(unsigned long long)hlock->prev_chain_key);
			return;
		}
P
Peter Zijlstra 已提交
2517 2518 2519
		/*
		 * Whoops ran out of static storage again?
		 */
2520
		if (DEBUG_LOCKS_WARN_ON(hlock->class_idx > MAX_LOCKDEP_KEYS))
2521 2522
			return;

I
Ingo Molnar 已提交
2523 2524 2525
		if (prev_hlock && (prev_hlock->irq_context !=
							hlock->irq_context))
			chain_key = 0;
2526
		chain_key = iterate_chain_key(chain_key, hlock->class_idx);
I
Ingo Molnar 已提交
2527 2528 2529 2530
		prev_hlock = hlock;
	}
	if (chain_key != curr->curr_chain_key) {
		debug_locks_off();
P
Peter Zijlstra 已提交
2531 2532 2533 2534
		/*
		 * More smoking hash instead of calculating it, damn see these
		 * numbers float.. I bet that a pink elephant stepped on my memory.
		 */
2535
		WARN(1, "hm#2, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2536 2537 2538 2539 2540 2541 2542
			curr->lockdep_depth, i,
			(unsigned long long)chain_key,
			(unsigned long long)curr->curr_chain_key);
	}
#endif
}

2543 2544 2545 2546 2547 2548 2549 2550 2551 2552
static void
print_usage_bug_scenario(struct held_lock *lock)
{
	struct lock_class *class = hlock_class(lock);

	printk(" Possible unsafe locking scenario:\n\n");
	printk("       CPU0\n");
	printk("       ----\n");
	printk("  lock(");
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
2553
	printk(KERN_CONT ");\n");
2554 2555 2556
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
2557
	printk(KERN_CONT ");\n");
2558 2559 2560
	printk("\n *** DEADLOCK ***\n\n");
}

P
Peter Zijlstra 已提交
2561 2562 2563 2564 2565 2566 2567
static int
print_usage_bug(struct task_struct *curr, struct held_lock *this,
		enum lock_usage_bit prev_bit, enum lock_usage_bit new_bit)
{
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
		return 0;

2568
	pr_warn("\n");
2569 2570
	pr_warn("================================\n");
	pr_warn("WARNING: inconsistent lock state\n");
2571
	print_kernel_ident();
2572
	pr_warn("--------------------------------\n");
P
Peter Zijlstra 已提交
2573

2574
	pr_warn("inconsistent {%s} -> {%s} usage.\n",
P
Peter Zijlstra 已提交
2575 2576
		usage_str[prev_bit], usage_str[new_bit]);

2577
	pr_warn("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] takes:\n",
2578
		curr->comm, task_pid_nr(curr),
P
Peter Zijlstra 已提交
2579 2580 2581 2582 2583 2584
		trace_hardirq_context(curr), hardirq_count() >> HARDIRQ_SHIFT,
		trace_softirq_context(curr), softirq_count() >> SOFTIRQ_SHIFT,
		trace_hardirqs_enabled(curr),
		trace_softirqs_enabled(curr));
	print_lock(this);

2585
	pr_warn("{%s} state was registered at:\n", usage_str[prev_bit]);
D
Dave Jones 已提交
2586
	print_stack_trace(hlock_class(this)->usage_traces + prev_bit, 1);
P
Peter Zijlstra 已提交
2587 2588

	print_irqtrace_events(curr);
2589
	pr_warn("\nother info that might help us debug this:\n");
2590 2591
	print_usage_bug_scenario(this);

P
Peter Zijlstra 已提交
2592 2593
	lockdep_print_held_locks(curr);

2594
	pr_warn("\nstack backtrace:\n");
P
Peter Zijlstra 已提交
2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
	dump_stack();

	return 0;
}

/*
 * Print out an error if an invalid bit is set:
 */
static inline int
valid_state(struct task_struct *curr, struct held_lock *this,
	    enum lock_usage_bit new_bit, enum lock_usage_bit bad_bit)
{
D
Dave Jones 已提交
2607
	if (unlikely(hlock_class(this)->usage_mask & (1 << bad_bit)))
P
Peter Zijlstra 已提交
2608 2609 2610 2611 2612 2613 2614
		return print_usage_bug(curr, this, bad_bit, new_bit);
	return 1;
}

static int mark_lock(struct task_struct *curr, struct held_lock *this,
		     enum lock_usage_bit new_bit);

2615
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
2616 2617 2618 2619 2620

/*
 * print irq inversion bug:
 */
static int
2621 2622
print_irq_inversion_bug(struct task_struct *curr,
			struct lock_list *root, struct lock_list *other,
I
Ingo Molnar 已提交
2623 2624 2625
			struct held_lock *this, int forwards,
			const char *irqclass)
{
2626 2627 2628 2629
	struct lock_list *entry = other;
	struct lock_list *middle = NULL;
	int depth;

2630
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
2631 2632
		return 0;

2633
	pr_warn("\n");
2634 2635
	pr_warn("========================================================\n");
	pr_warn("WARNING: possible irq lock inversion dependency detected\n");
2636
	print_kernel_ident();
2637
	pr_warn("--------------------------------------------------------\n");
2638
	pr_warn("%s/%d just changed the state of lock:\n",
2639
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
2640 2641
	print_lock(this);
	if (forwards)
2642
		pr_warn("but this lock took another, %s-unsafe lock in the past:\n", irqclass);
I
Ingo Molnar 已提交
2643
	else
2644
		pr_warn("but this lock was taken by another, %s-safe lock in the past:\n", irqclass);
2645
	print_lock_name(other->class);
2646
	pr_warn("\n\nand interrupts could create inverse lock ordering between them.\n\n");
I
Ingo Molnar 已提交
2647

2648
	pr_warn("\nother info that might help us debug this:\n");
2649 2650 2651 2652 2653

	/* Find a middle lock (if one exists) */
	depth = get_lock_depth(other);
	do {
		if (depth == 0 && (entry != root)) {
2654
			pr_warn("lockdep:%s bad path found in chain graph\n", __func__);
2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667
			break;
		}
		middle = entry;
		entry = get_lock_parent(entry);
		depth--;
	} while (entry && entry != root && (depth >= 0));
	if (forwards)
		print_irq_lock_scenario(root, other,
			middle ? middle->class : root->class, other->class);
	else
		print_irq_lock_scenario(other, root,
			middle ? middle->class : other->class, root->class);

I
Ingo Molnar 已提交
2668 2669
	lockdep_print_held_locks(curr);

2670
	pr_warn("\nthe shortest dependencies between 2nd lock and 1st lock:\n");
2671 2672 2673
	if (!save_trace(&root->trace))
		return 0;
	print_shortest_lock_dependencies(other, root);
I
Ingo Molnar 已提交
2674

2675
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689
	dump_stack();

	return 0;
}

/*
 * Prove that in the forwards-direction subgraph starting at <this>
 * there is no lock matching <mask>:
 */
static int
check_usage_forwards(struct task_struct *curr, struct held_lock *this,
		     enum lock_usage_bit bit, const char *irqclass)
{
	int ret;
2690 2691
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2692

2693 2694 2695
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_forwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2696 2697 2698 2699
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2700

2701
	return print_irq_inversion_bug(curr, &root, target_entry,
2702
					this, 1, irqclass);
I
Ingo Molnar 已提交
2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713
}

/*
 * Prove that in the backwards-direction subgraph starting at <this>
 * there is no lock matching <mask>:
 */
static int
check_usage_backwards(struct task_struct *curr, struct held_lock *this,
		      enum lock_usage_bit bit, const char *irqclass)
{
	int ret;
2714 2715
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2716

2717 2718 2719
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_backwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2720 2721 2722 2723
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2724

2725
	return print_irq_inversion_bug(curr, &root, target_entry,
2726
					this, 0, irqclass);
I
Ingo Molnar 已提交
2727 2728
}

2729
void print_irqtrace_events(struct task_struct *curr)
I
Ingo Molnar 已提交
2730 2731
{
	printk("irq event stamp: %u\n", curr->irq_events);
D
Dmitry Vyukov 已提交
2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743
	printk("hardirqs last  enabled at (%u): [<%p>] %pS\n",
		curr->hardirq_enable_event, (void *)curr->hardirq_enable_ip,
		(void *)curr->hardirq_enable_ip);
	printk("hardirqs last disabled at (%u): [<%p>] %pS\n",
		curr->hardirq_disable_event, (void *)curr->hardirq_disable_ip,
		(void *)curr->hardirq_disable_ip);
	printk("softirqs last  enabled at (%u): [<%p>] %pS\n",
		curr->softirq_enable_event, (void *)curr->softirq_enable_ip,
		(void *)curr->softirq_enable_ip);
	printk("softirqs last disabled at (%u): [<%p>] %pS\n",
		curr->softirq_disable_event, (void *)curr->softirq_disable_ip,
		(void *)curr->softirq_disable_ip);
I
Ingo Molnar 已提交
2744 2745
}

2746
static int HARDIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2747
{
P
Peter Zijlstra 已提交
2748 2749 2750
#if HARDIRQ_VERBOSE
	return class_filter(class);
#endif
I
Ingo Molnar 已提交
2751 2752 2753
	return 0;
}

2754
static int SOFTIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2755
{
P
Peter Zijlstra 已提交
2756 2757 2758 2759
#if SOFTIRQ_VERBOSE
	return class_filter(class);
#endif
	return 0;
I
Ingo Molnar 已提交
2760 2761 2762 2763
}

#define STRICT_READ_CHECKS	1

2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
static int (*state_verbose_f[])(struct lock_class *class) = {
#define LOCKDEP_STATE(__STATE) \
	__STATE##_verbose,
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static inline int state_verbose(enum lock_usage_bit bit,
				struct lock_class *class)
{
	return state_verbose_f[bit >> 2](class);
}

2777 2778 2779
typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
			     enum lock_usage_bit bit, const char *name);

2780
static int
2781 2782
mark_lock_irq(struct task_struct *curr, struct held_lock *this,
		enum lock_usage_bit new_bit)
2783
{
2784
	int excl_bit = exclusive_bit(new_bit);
2785
	int read = new_bit & 1;
2786 2787
	int dir = new_bit & 2;

2788 2789 2790 2791 2792 2793 2794
	/*
	 * mark USED_IN has to look forwards -- to ensure no dependency
	 * has ENABLED state, which would allow recursion deadlocks.
	 *
	 * mark ENABLED has to look backwards -- to ensure no dependee
	 * has USED_IN state, which, again, would allow  recursion deadlocks.
	 */
2795 2796
	check_usage_f usage = dir ?
		check_usage_backwards : check_usage_forwards;
2797

2798 2799 2800 2801
	/*
	 * Validate that this particular lock does not have conflicting
	 * usage states.
	 */
2802 2803
	if (!valid_state(curr, this, new_bit, excl_bit))
		return 0;
2804

2805 2806 2807 2808 2809
	/*
	 * Validate that the lock dependencies don't have conflicting usage
	 * states.
	 */
	if ((!read || !dir || STRICT_READ_CHECKS) &&
2810
			!usage(curr, this, excl_bit, state_name(new_bit & ~1)))
2811
		return 0;
2812

2813 2814 2815 2816 2817 2818 2819 2820
	/*
	 * Check for read in write conflicts
	 */
	if (!read) {
		if (!valid_state(curr, this, new_bit, excl_bit + 1))
			return 0;

		if (STRICT_READ_CHECKS &&
2821 2822
			!usage(curr, this, excl_bit + 1,
				state_name(new_bit + 1)))
2823 2824
			return 0;
	}
2825

2826
	if (state_verbose(new_bit, hlock_class(this)))
2827 2828 2829 2830 2831
		return 2;

	return 1;
}

2832
enum mark_type {
2833 2834 2835
#define LOCKDEP_STATE(__STATE)	__STATE,
#include "lockdep_states.h"
#undef LOCKDEP_STATE
2836 2837
};

I
Ingo Molnar 已提交
2838 2839 2840
/*
 * Mark all held locks with a usage bit:
 */
2841
static int
2842
mark_held_locks(struct task_struct *curr, enum mark_type mark)
I
Ingo Molnar 已提交
2843 2844 2845 2846 2847 2848 2849 2850
{
	enum lock_usage_bit usage_bit;
	struct held_lock *hlock;
	int i;

	for (i = 0; i < curr->lockdep_depth; i++) {
		hlock = curr->held_locks + i;

2851 2852 2853 2854 2855
		usage_bit = 2 + (mark << 2); /* ENABLED */
		if (hlock->read)
			usage_bit += 1; /* READ */

		BUG_ON(usage_bit >= LOCK_USAGE_STATES);
2856

2857
		if (!hlock->check)
2858 2859
			continue;

2860
		if (!mark_lock(curr, hlock, usage_bit))
I
Ingo Molnar 已提交
2861 2862 2863 2864 2865 2866 2867 2868 2869
			return 0;
	}

	return 1;
}

/*
 * Hardirqs will be enabled:
 */
2870
static void __trace_hardirqs_on_caller(unsigned long ip)
I
Ingo Molnar 已提交
2871 2872 2873 2874 2875 2876 2877 2878 2879 2880
{
	struct task_struct *curr = current;

	/* we'll do an OFF -> ON transition: */
	curr->hardirqs_enabled = 1;

	/*
	 * We are going to turn hardirqs on, so set the
	 * usage bit for all held locks:
	 */
2881
	if (!mark_held_locks(curr, HARDIRQ))
I
Ingo Molnar 已提交
2882 2883 2884 2885 2886 2887 2888
		return;
	/*
	 * If we have softirqs enabled, then set the usage
	 * bit for all held locks. (disabled hardirqs prevented
	 * this bit from being set before)
	 */
	if (curr->softirqs_enabled)
2889
		if (!mark_held_locks(curr, SOFTIRQ))
I
Ingo Molnar 已提交
2890 2891
			return;

P
Peter Zijlstra 已提交
2892 2893
	curr->hardirq_enable_ip = ip;
	curr->hardirq_enable_event = ++curr->irq_events;
2894
	debug_atomic_inc(hardirqs_on_events);
P
Peter Zijlstra 已提交
2895
}
2896

2897
__visible void trace_hardirqs_on_caller(unsigned long ip)
2898 2899 2900 2901 2902 2903
{
	time_hardirqs_on(CALLER_ADDR0, ip);

	if (unlikely(!debug_locks || current->lockdep_recursion))
		return;

2904 2905 2906 2907 2908 2909 2910 2911 2912 2913
	if (unlikely(current->hardirqs_enabled)) {
		/*
		 * Neither irq nor preemption are disabled here
		 * so this is racy by nature but losing one hit
		 * in a stat is not a big deal.
		 */
		__debug_atomic_inc(redundant_hardirqs_on);
		return;
	}

P
Peter Zijlstra 已提交
2914 2915 2916 2917 2918
	/*
	 * We're enabling irqs and according to our state above irqs weren't
	 * already enabled, yet we find the hardware thinks they are in fact
	 * enabled.. someone messed up their IRQ state tracing.
	 */
2919 2920 2921
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

P
Peter Zijlstra 已提交
2922 2923 2924
	/*
	 * See the fine text that goes along with this variable definition.
	 */
2925 2926 2927
	if (DEBUG_LOCKS_WARN_ON(unlikely(early_boot_irqs_disabled)))
		return;

P
Peter Zijlstra 已提交
2928 2929 2930 2931
	/*
	 * Can't allow enabling interrupts while in an interrupt handler,
	 * that's general bad form and such. Recursion, limited stack etc..
	 */
2932 2933 2934
	if (DEBUG_LOCKS_WARN_ON(current->hardirq_context))
		return;

2935 2936 2937 2938
	current->lockdep_recursion = 1;
	__trace_hardirqs_on_caller(ip);
	current->lockdep_recursion = 0;
}
2939
EXPORT_SYMBOL(trace_hardirqs_on_caller);
P
Peter Zijlstra 已提交
2940

2941
void trace_hardirqs_on(void)
2942 2943 2944
{
	trace_hardirqs_on_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2945 2946 2947 2948 2949
EXPORT_SYMBOL(trace_hardirqs_on);

/*
 * Hardirqs were disabled:
 */
2950
__visible void trace_hardirqs_off_caller(unsigned long ip)
P
Peter Zijlstra 已提交
2951 2952 2953
{
	struct task_struct *curr = current;

2954
	time_hardirqs_off(CALLER_ADDR0, ip);
2955

P
Peter Zijlstra 已提交
2956 2957 2958
	if (unlikely(!debug_locks || current->lockdep_recursion))
		return;

P
Peter Zijlstra 已提交
2959 2960 2961 2962
	/*
	 * So we're supposed to get called after you mask local IRQs, but for
	 * some reason the hardware doesn't quite think you did a proper job.
	 */
P
Peter Zijlstra 已提交
2963 2964 2965 2966 2967 2968 2969 2970
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->hardirqs_enabled) {
		/*
		 * We have done an ON -> OFF transition:
		 */
		curr->hardirqs_enabled = 0;
2971
		curr->hardirq_disable_ip = ip;
P
Peter Zijlstra 已提交
2972
		curr->hardirq_disable_event = ++curr->irq_events;
2973
		debug_atomic_inc(hardirqs_off_events);
P
Peter Zijlstra 已提交
2974
	} else
2975
		debug_atomic_inc(redundant_hardirqs_off);
P
Peter Zijlstra 已提交
2976
}
2977
EXPORT_SYMBOL(trace_hardirqs_off_caller);
P
Peter Zijlstra 已提交
2978

2979
void trace_hardirqs_off(void)
2980 2981 2982
{
	trace_hardirqs_off_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2983 2984 2985 2986 2987 2988 2989 2990 2991
EXPORT_SYMBOL(trace_hardirqs_off);

/*
 * Softirqs will be enabled:
 */
void trace_softirqs_on(unsigned long ip)
{
	struct task_struct *curr = current;

2992
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
2993 2994
		return;

P
Peter Zijlstra 已提交
2995 2996 2997 2998
	/*
	 * We fancy IRQs being disabled here, see softirq.c, avoids
	 * funny state and nesting things.
	 */
P
Peter Zijlstra 已提交
2999 3000 3001 3002
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->softirqs_enabled) {
3003
		debug_atomic_inc(redundant_softirqs_on);
P
Peter Zijlstra 已提交
3004 3005 3006
		return;
	}

3007
	current->lockdep_recursion = 1;
P
Peter Zijlstra 已提交
3008 3009 3010 3011 3012 3013
	/*
	 * We'll do an OFF -> ON transition:
	 */
	curr->softirqs_enabled = 1;
	curr->softirq_enable_ip = ip;
	curr->softirq_enable_event = ++curr->irq_events;
3014
	debug_atomic_inc(softirqs_on_events);
P
Peter Zijlstra 已提交
3015 3016 3017 3018 3019 3020
	/*
	 * We are going to turn softirqs on, so set the
	 * usage bit for all held locks, if hardirqs are
	 * enabled too:
	 */
	if (curr->hardirqs_enabled)
3021
		mark_held_locks(curr, SOFTIRQ);
3022
	current->lockdep_recursion = 0;
P
Peter Zijlstra 已提交
3023 3024 3025 3026 3027 3028 3029 3030 3031
}

/*
 * Softirqs were disabled:
 */
void trace_softirqs_off(unsigned long ip)
{
	struct task_struct *curr = current;

3032
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
3033 3034
		return;

P
Peter Zijlstra 已提交
3035 3036 3037
	/*
	 * We fancy IRQs being disabled here, see softirq.c
	 */
P
Peter Zijlstra 已提交
3038 3039 3040 3041 3042 3043 3044 3045 3046 3047
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->softirqs_enabled) {
		/*
		 * We have done an ON -> OFF transition:
		 */
		curr->softirqs_enabled = 0;
		curr->softirq_disable_ip = ip;
		curr->softirq_disable_event = ++curr->irq_events;
3048
		debug_atomic_inc(softirqs_off_events);
P
Peter Zijlstra 已提交
3049 3050 3051
		/*
		 * Whoops, we wanted softirqs off, so why aren't they?
		 */
P
Peter Zijlstra 已提交
3052 3053
		DEBUG_LOCKS_WARN_ON(!softirq_count());
	} else
3054
		debug_atomic_inc(redundant_softirqs_off);
P
Peter Zijlstra 已提交
3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084
}

static int mark_irqflags(struct task_struct *curr, struct held_lock *hlock)
{
	/*
	 * If non-trylock use in a hardirq or softirq context, then
	 * mark the lock as used in these contexts:
	 */
	if (!hlock->trylock) {
		if (hlock->read) {
			if (curr->hardirq_context)
				if (!mark_lock(curr, hlock,
						LOCK_USED_IN_HARDIRQ_READ))
					return 0;
			if (curr->softirq_context)
				if (!mark_lock(curr, hlock,
						LOCK_USED_IN_SOFTIRQ_READ))
					return 0;
		} else {
			if (curr->hardirq_context)
				if (!mark_lock(curr, hlock, LOCK_USED_IN_HARDIRQ))
					return 0;
			if (curr->softirq_context)
				if (!mark_lock(curr, hlock, LOCK_USED_IN_SOFTIRQ))
					return 0;
		}
	}
	if (!hlock->hardirqs_off) {
		if (hlock->read) {
			if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
3085
					LOCK_ENABLED_HARDIRQ_READ))
P
Peter Zijlstra 已提交
3086 3087 3088
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
3089
						LOCK_ENABLED_SOFTIRQ_READ))
P
Peter Zijlstra 已提交
3090 3091 3092
					return 0;
		} else {
			if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
3093
					LOCK_ENABLED_HARDIRQ))
P
Peter Zijlstra 已提交
3094 3095 3096
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
3097
						LOCK_ENABLED_SOFTIRQ))
P
Peter Zijlstra 已提交
3098 3099 3100 3101 3102 3103 3104
					return 0;
		}
	}

	return 1;
}

3105 3106 3107 3108 3109
static inline unsigned int task_irq_context(struct task_struct *task)
{
	return 2 * !!task->hardirq_context + !!task->softirq_context;
}

P
Peter Zijlstra 已提交
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130
static int separate_irq_context(struct task_struct *curr,
		struct held_lock *hlock)
{
	unsigned int depth = curr->lockdep_depth;

	/*
	 * Keep track of points where we cross into an interrupt context:
	 */
	if (depth) {
		struct held_lock *prev_hlock;

		prev_hlock = curr->held_locks + depth-1;
		/*
		 * If we cross into another context, reset the
		 * hash key (this also prevents the checking and the
		 * adding of the dependency to 'prev'):
		 */
		if (prev_hlock->irq_context != hlock->irq_context)
			return 1;
	}
	return 0;
I
Ingo Molnar 已提交
3131 3132
}

P
Peter Zijlstra 已提交
3133
#else /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
I
Ingo Molnar 已提交
3134

P
Peter Zijlstra 已提交
3135 3136 3137
static inline
int mark_lock_irq(struct task_struct *curr, struct held_lock *this,
		enum lock_usage_bit new_bit)
I
Ingo Molnar 已提交
3138
{
P
Peter Zijlstra 已提交
3139
	WARN_ON(1); /* Impossible innit? when we don't have TRACE_IRQFLAG */
P
Peter Zijlstra 已提交
3140 3141
	return 1;
}
I
Ingo Molnar 已提交
3142

P
Peter Zijlstra 已提交
3143 3144 3145 3146 3147
static inline int mark_irqflags(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 1;
}
I
Ingo Molnar 已提交
3148

3149 3150 3151 3152 3153
static inline unsigned int task_irq_context(struct task_struct *task)
{
	return 0;
}

P
Peter Zijlstra 已提交
3154 3155 3156 3157
static inline int separate_irq_context(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 0;
I
Ingo Molnar 已提交
3158 3159
}

P
Peter Zijlstra 已提交
3160
#endif /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
I
Ingo Molnar 已提交
3161 3162

/*
P
Peter Zijlstra 已提交
3163
 * Mark a lock with a usage bit, and validate the state transition:
I
Ingo Molnar 已提交
3164
 */
3165
static int mark_lock(struct task_struct *curr, struct held_lock *this,
3166
			     enum lock_usage_bit new_bit)
I
Ingo Molnar 已提交
3167
{
P
Peter Zijlstra 已提交
3168
	unsigned int new_mask = 1 << new_bit, ret = 1;
I
Ingo Molnar 已提交
3169 3170

	/*
P
Peter Zijlstra 已提交
3171 3172
	 * If already set then do not dirty the cacheline,
	 * nor do any checks:
I
Ingo Molnar 已提交
3173
	 */
D
Dave Jones 已提交
3174
	if (likely(hlock_class(this)->usage_mask & new_mask))
P
Peter Zijlstra 已提交
3175 3176 3177 3178
		return 1;

	if (!graph_lock())
		return 0;
I
Ingo Molnar 已提交
3179
	/*
L
Lucas De Marchi 已提交
3180
	 * Make sure we didn't race:
I
Ingo Molnar 已提交
3181
	 */
D
Dave Jones 已提交
3182
	if (unlikely(hlock_class(this)->usage_mask & new_mask)) {
P
Peter Zijlstra 已提交
3183 3184 3185
		graph_unlock();
		return 1;
	}
I
Ingo Molnar 已提交
3186

D
Dave Jones 已提交
3187
	hlock_class(this)->usage_mask |= new_mask;
I
Ingo Molnar 已提交
3188

D
Dave Jones 已提交
3189
	if (!save_trace(hlock_class(this)->usage_traces + new_bit))
P
Peter Zijlstra 已提交
3190
		return 0;
I
Ingo Molnar 已提交
3191

P
Peter Zijlstra 已提交
3192
	switch (new_bit) {
P
Peter Zijlstra 已提交
3193 3194 3195 3196 3197 3198 3199
#define LOCKDEP_STATE(__STATE)			\
	case LOCK_USED_IN_##__STATE:		\
	case LOCK_USED_IN_##__STATE##_READ:	\
	case LOCK_ENABLED_##__STATE:		\
	case LOCK_ENABLED_##__STATE##_READ:
#include "lockdep_states.h"
#undef LOCKDEP_STATE
P
Peter Zijlstra 已提交
3200 3201 3202 3203 3204
		ret = mark_lock_irq(curr, this, new_bit);
		if (!ret)
			return 0;
		break;
	case LOCK_USED:
3205
		debug_atomic_dec(nr_unused_locks);
P
Peter Zijlstra 已提交
3206 3207 3208 3209 3210 3211 3212
		break;
	default:
		if (!debug_locks_off_graph_unlock())
			return 0;
		WARN_ON(1);
		return 0;
	}
I
Ingo Molnar 已提交
3213

P
Peter Zijlstra 已提交
3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
	graph_unlock();

	/*
	 * We must printk outside of the graph_lock:
	 */
	if (ret == 2) {
		printk("\nmarked lock as {%s}:\n", usage_str[new_bit]);
		print_lock(this);
		print_irqtrace_events(curr);
		dump_stack();
	}

	return ret;
}
I
Ingo Molnar 已提交
3228 3229 3230 3231

/*
 * Initialize a lock instance's lock-class mapping info:
 */
3232
static void __lockdep_init_map(struct lockdep_map *lock, const char *name,
3233
		      struct lock_class_key *key, int subclass)
I
Ingo Molnar 已提交
3234
{
3235 3236 3237 3238 3239 3240
	int i;

	kmemcheck_mark_initialized(lock, sizeof(*lock));

	for (i = 0; i < NR_LOCKDEP_CACHING_CLASSES; i++)
		lock->class_cache[i] = NULL;
3241

3242 3243 3244 3245
#ifdef CONFIG_LOCK_STAT
	lock->cpu = raw_smp_processor_id();
#endif

P
Peter Zijlstra 已提交
3246 3247 3248
	/*
	 * Can't be having no nameless bastards around this place!
	 */
3249 3250
	if (DEBUG_LOCKS_WARN_ON(!name)) {
		lock->name = "NULL";
I
Ingo Molnar 已提交
3251
		return;
3252 3253 3254
	}

	lock->name = name;
I
Ingo Molnar 已提交
3255

P
Peter Zijlstra 已提交
3256 3257 3258
	/*
	 * No key, no joy, we need to hash something.
	 */
I
Ingo Molnar 已提交
3259 3260 3261 3262 3263 3264 3265
	if (DEBUG_LOCKS_WARN_ON(!key))
		return;
	/*
	 * Sanity check, the lock-class key must be persistent:
	 */
	if (!static_obj(key)) {
		printk("BUG: key %p not in .data!\n", key);
P
Peter Zijlstra 已提交
3266 3267 3268
		/*
		 * What it says above ^^^^^, I suggest you read it.
		 */
I
Ingo Molnar 已提交
3269 3270 3271 3272
		DEBUG_LOCKS_WARN_ON(1);
		return;
	}
	lock->key = key;
3273 3274 3275 3276

	if (unlikely(!debug_locks))
		return;

3277 3278 3279 3280 3281 3282 3283 3284
	if (subclass) {
		unsigned long flags;

		if (DEBUG_LOCKS_WARN_ON(current->lockdep_recursion))
			return;

		raw_local_irq_save(flags);
		current->lockdep_recursion = 1;
3285
		register_lock_class(lock, subclass, 1);
3286 3287 3288
		current->lockdep_recursion = 0;
		raw_local_irq_restore(flags);
	}
I
Ingo Molnar 已提交
3289
}
3290 3291 3292 3293 3294 3295 3296

void lockdep_init_map(struct lockdep_map *lock, const char *name,
		      struct lock_class_key *key, int subclass)
{
	cross_init(lock, 0);
	__lockdep_init_map(lock, name, key, subclass);
}
I
Ingo Molnar 已提交
3297 3298
EXPORT_SYMBOL_GPL(lockdep_init_map);

3299 3300 3301 3302 3303 3304 3305 3306 3307 3308
#ifdef CONFIG_LOCKDEP_CROSSRELEASE
void lockdep_init_map_crosslock(struct lockdep_map *lock, const char *name,
		      struct lock_class_key *key, int subclass)
{
	cross_init(lock, 1);
	__lockdep_init_map(lock, name, key, subclass);
}
EXPORT_SYMBOL_GPL(lockdep_init_map_crosslock);
#endif

3309
struct lock_class_key __lockdep_no_validate__;
K
Kent Overstreet 已提交
3310
EXPORT_SYMBOL_GPL(__lockdep_no_validate__);
3311

3312 3313 3314 3315 3316 3317 3318 3319 3320 3321
static int
print_lock_nested_lock_not_held(struct task_struct *curr,
				struct held_lock *hlock,
				unsigned long ip)
{
	if (!debug_locks_off())
		return 0;
	if (debug_locks_silent)
		return 0;

3322
	pr_warn("\n");
3323 3324
	pr_warn("==================================\n");
	pr_warn("WARNING: Nested lock was not taken\n");
3325
	print_kernel_ident();
3326
	pr_warn("----------------------------------\n");
3327

3328
	pr_warn("%s/%d is trying to lock:\n", curr->comm, task_pid_nr(curr));
3329 3330
	print_lock(hlock);

3331 3332
	pr_warn("\nbut this task is not holding:\n");
	pr_warn("%s\n", hlock->nest_lock->name);
3333

3334
	pr_warn("\nstack backtrace:\n");
3335 3336
	dump_stack();

3337
	pr_warn("\nother info that might help us debug this:\n");
3338 3339
	lockdep_print_held_locks(curr);

3340
	pr_warn("\nstack backtrace:\n");
3341 3342 3343 3344 3345
	dump_stack();

	return 0;
}

3346
static int __lock_is_held(struct lockdep_map *lock, int read);
3347

I
Ingo Molnar 已提交
3348 3349 3350 3351 3352 3353
/*
 * This gets called for every mutex_lock*()/spin_lock*() operation.
 * We maintain the dependency maps and validate the locking attempt:
 */
static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
			  int trylock, int read, int check, int hardirqs_off,
3354
			  struct lockdep_map *nest_lock, unsigned long ip,
3355
			  int references, int pin_count)
I
Ingo Molnar 已提交
3356 3357
{
	struct task_struct *curr = current;
3358
	struct lock_class *class = NULL;
I
Ingo Molnar 已提交
3359
	struct held_lock *hlock;
3360
	unsigned int depth;
I
Ingo Molnar 已提交
3361
	int chain_head = 0;
3362
	int class_idx;
I
Ingo Molnar 已提交
3363
	u64 chain_key;
3364
	int ret;
I
Ingo Molnar 已提交
3365 3366 3367 3368

	if (unlikely(!debug_locks))
		return 0;

P
Peter Zijlstra 已提交
3369 3370 3371 3372 3373
	/*
	 * Lockdep should run with IRQs disabled, otherwise we could
	 * get an interrupt which would want to take locks, which would
	 * end up in lockdep and have you got a head-ache already?
	 */
I
Ingo Molnar 已提交
3374 3375 3376
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;

3377 3378
	if (!prove_locking || lock->key == &__lockdep_no_validate__)
		check = 0;
3379

3380 3381
	if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		class = lock->class_cache[subclass];
3382
	/*
3383
	 * Not cached?
3384
	 */
I
Ingo Molnar 已提交
3385
	if (unlikely(!class)) {
3386
		class = register_lock_class(lock, subclass, 0);
I
Ingo Molnar 已提交
3387 3388 3389
		if (!class)
			return 0;
	}
3390
	atomic_inc((atomic_t *)&class->ops);
I
Ingo Molnar 已提交
3391 3392 3393
	if (very_verbose(class)) {
		printk("\nacquire class [%p] %s", class->key, class->name);
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
3394 3395
			printk(KERN_CONT "#%d", class->name_version);
		printk(KERN_CONT "\n");
I
Ingo Molnar 已提交
3396 3397 3398 3399 3400 3401 3402 3403 3404
		dump_stack();
	}

	/*
	 * Add the lock to the list of currently held locks.
	 * (we dont increase the depth just yet, up until the
	 * dependency checks are done)
	 */
	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3405 3406 3407
	/*
	 * Ran out of static storage for our per-task lock stack again have we?
	 */
I
Ingo Molnar 已提交
3408 3409 3410
	if (DEBUG_LOCKS_WARN_ON(depth >= MAX_LOCK_DEPTH))
		return 0;

3411 3412
	class_idx = class - lock_classes + 1;

3413 3414
	/* TODO: nest_lock is not implemented for crosslock yet. */
	if (depth && !cross_lock(lock)) {
3415 3416
		hlock = curr->held_locks + depth - 1;
		if (hlock->class_idx == class_idx && nest_lock) {
3417 3418 3419 3420 3421 3422 3423
			if (hlock->references) {
				/*
				 * Check: unsigned int references:12, overflow.
				 */
				if (DEBUG_LOCKS_WARN_ON(hlock->references == (1 << 12)-1))
					return 0;

3424
				hlock->references++;
3425
			} else {
3426
				hlock->references = 2;
3427
			}
3428 3429 3430 3431 3432

			return 1;
		}
	}

I
Ingo Molnar 已提交
3433
	hlock = curr->held_locks + depth;
P
Peter Zijlstra 已提交
3434 3435 3436 3437
	/*
	 * Plain impossible, we just registered it and checked it weren't no
	 * NULL like.. I bet this mushroom I ate was good!
	 */
D
Dave Jones 已提交
3438 3439
	if (DEBUG_LOCKS_WARN_ON(!class))
		return 0;
3440
	hlock->class_idx = class_idx;
I
Ingo Molnar 已提交
3441 3442
	hlock->acquire_ip = ip;
	hlock->instance = lock;
P
Peter Zijlstra 已提交
3443
	hlock->nest_lock = nest_lock;
3444
	hlock->irq_context = task_irq_context(curr);
I
Ingo Molnar 已提交
3445 3446 3447
	hlock->trylock = trylock;
	hlock->read = read;
	hlock->check = check;
3448
	hlock->hardirqs_off = !!hardirqs_off;
3449
	hlock->references = references;
P
Peter Zijlstra 已提交
3450 3451
#ifdef CONFIG_LOCK_STAT
	hlock->waittime_stamp = 0;
3452
	hlock->holdtime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
3453
#endif
3454
	hlock->pin_count = pin_count;
I
Ingo Molnar 已提交
3455

3456
	if (check && !mark_irqflags(curr, hlock))
P
Peter Zijlstra 已提交
3457 3458
		return 0;

I
Ingo Molnar 已提交
3459
	/* mark it as used: */
3460
	if (!mark_lock(curr, hlock, LOCK_USED))
I
Ingo Molnar 已提交
3461
		return 0;
P
Peter Zijlstra 已提交
3462

I
Ingo Molnar 已提交
3463
	/*
3464
	 * Calculate the chain hash: it's the combined hash of all the
I
Ingo Molnar 已提交
3465 3466 3467 3468 3469 3470 3471 3472
	 * lock keys along the dependency chain. We save the hash value
	 * at every step so that we can get the current hash easily
	 * after unlock. The chain hash is then used to cache dependency
	 * results.
	 *
	 * The 'key ID' is what is the most compact key value to drive
	 * the hash, not class->key.
	 */
P
Peter Zijlstra 已提交
3473 3474 3475
	/*
	 * Whoops, we did it again.. ran straight out of our static allocation.
	 */
3476
	if (DEBUG_LOCKS_WARN_ON(class_idx > MAX_LOCKDEP_KEYS))
I
Ingo Molnar 已提交
3477 3478 3479 3480
		return 0;

	chain_key = curr->curr_chain_key;
	if (!depth) {
P
Peter Zijlstra 已提交
3481 3482 3483
		/*
		 * How can we have a chain hash when we ain't got no keys?!
		 */
I
Ingo Molnar 已提交
3484 3485 3486 3487 3488 3489
		if (DEBUG_LOCKS_WARN_ON(chain_key != 0))
			return 0;
		chain_head = 1;
	}

	hlock->prev_chain_key = chain_key;
P
Peter Zijlstra 已提交
3490 3491 3492
	if (separate_irq_context(curr, hlock)) {
		chain_key = 0;
		chain_head = 1;
I
Ingo Molnar 已提交
3493
	}
3494
	chain_key = iterate_chain_key(chain_key, class_idx);
I
Ingo Molnar 已提交
3495

3496
	if (nest_lock && !__lock_is_held(nest_lock, -1))
3497 3498
		return print_lock_nested_lock_not_held(curr, hlock, ip);

3499
	if (!validate_chain(curr, lock, hlock, chain_head, chain_key))
P
Peter Zijlstra 已提交
3500
		return 0;
3501

3502 3503 3504 3505 3506 3507 3508 3509
	ret = lock_acquire_crosslock(hlock);
	/*
	 * 2 means normal acquire operations are needed. Otherwise, it's
	 * ok just to return with '0:fail, 1:success'.
	 */
	if (ret != 2)
		return ret;

3510
	curr->curr_chain_key = chain_key;
I
Ingo Molnar 已提交
3511 3512
	curr->lockdep_depth++;
	check_chain_key(curr);
3513 3514 3515 3516
#ifdef CONFIG_DEBUG_LOCKDEP
	if (unlikely(!debug_locks))
		return 0;
#endif
I
Ingo Molnar 已提交
3517 3518
	if (unlikely(curr->lockdep_depth >= MAX_LOCK_DEPTH)) {
		debug_locks_off();
3519 3520
		print_lockdep_off("BUG: MAX_LOCK_DEPTH too low!");
		printk(KERN_DEBUG "depth: %i  max: %lu!\n",
3521 3522 3523 3524
		       curr->lockdep_depth, MAX_LOCK_DEPTH);

		lockdep_print_held_locks(current);
		debug_show_all_locks();
3525
		dump_stack();
3526

I
Ingo Molnar 已提交
3527 3528
		return 0;
	}
3529

I
Ingo Molnar 已提交
3530 3531 3532 3533 3534 3535 3536
	if (unlikely(curr->lockdep_depth > max_lockdep_depth))
		max_lockdep_depth = curr->lockdep_depth;

	return 1;
}

static int
3537
print_unlock_imbalance_bug(struct task_struct *curr, struct lockdep_map *lock,
I
Ingo Molnar 已提交
3538 3539 3540 3541 3542 3543 3544
			   unsigned long ip)
{
	if (!debug_locks_off())
		return 0;
	if (debug_locks_silent)
		return 0;

3545
	pr_warn("\n");
3546 3547
	pr_warn("=====================================\n");
	pr_warn("WARNING: bad unlock balance detected!\n");
3548
	print_kernel_ident();
3549
	pr_warn("-------------------------------------\n");
3550
	pr_warn("%s/%d is trying to release lock (",
3551
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
3552
	print_lockdep_cache(lock);
3553
	pr_cont(") at:\n");
I
Ingo Molnar 已提交
3554
	print_ip_sym(ip);
3555 3556
	pr_warn("but there are no more locks to release!\n");
	pr_warn("\nother info that might help us debug this:\n");
I
Ingo Molnar 已提交
3557 3558
	lockdep_print_held_locks(curr);

3559
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
3560 3561 3562 3563 3564
	dump_stack();

	return 0;
}

3565 3566 3567 3568 3569 3570
static int match_held_lock(struct held_lock *hlock, struct lockdep_map *lock)
{
	if (hlock->instance == lock)
		return 1;

	if (hlock->references) {
3571
		struct lock_class *class = lock->class_cache[0];
3572 3573 3574 3575

		if (!class)
			class = look_up_lock_class(lock, 0);

3576 3577 3578 3579 3580 3581
		/*
		 * If look_up_lock_class() failed to find a class, we're trying
		 * to test if we hold a lock that has never yet been acquired.
		 * Clearly if the lock hasn't been acquired _ever_, we're not
		 * holding it either, so report failure.
		 */
3582
		if (IS_ERR_OR_NULL(class))
3583 3584
			return 0;

P
Peter Zijlstra 已提交
3585 3586 3587 3588 3589
		/*
		 * References, but not a lock we're actually ref-counting?
		 * State got messed up, follow the sites that change ->references
		 * and try to make sense of it.
		 */
3590 3591 3592 3593 3594 3595 3596 3597 3598 3599
		if (DEBUG_LOCKS_WARN_ON(!hlock->nest_lock))
			return 0;

		if (hlock->class_idx == class - lock_classes + 1)
			return 1;
	}

	return 0;
}

3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635
/* @depth must not be zero */
static struct held_lock *find_held_lock(struct task_struct *curr,
					struct lockdep_map *lock,
					unsigned int depth, int *idx)
{
	struct held_lock *ret, *hlock, *prev_hlock;
	int i;

	i = depth - 1;
	hlock = curr->held_locks + i;
	ret = hlock;
	if (match_held_lock(hlock, lock))
		goto out;

	ret = NULL;
	for (i--, prev_hlock = hlock--;
	     i >= 0;
	     i--, prev_hlock = hlock--) {
		/*
		 * We must not cross into another context:
		 */
		if (prev_hlock->irq_context != hlock->irq_context) {
			ret = NULL;
			break;
		}
		if (match_held_lock(hlock, lock)) {
			ret = hlock;
			break;
		}
	}

out:
	*idx = i;
	return ret;
}

3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653
static int reacquire_held_locks(struct task_struct *curr, unsigned int depth,
			      int idx)
{
	struct held_lock *hlock;

	for (hlock = curr->held_locks + idx; idx < depth; idx++, hlock++) {
		if (!__lock_acquire(hlock->instance,
				    hlock_class(hlock)->subclass,
				    hlock->trylock,
				    hlock->read, hlock->check,
				    hlock->hardirqs_off,
				    hlock->nest_lock, hlock->acquire_ip,
				    hlock->references, hlock->pin_count))
			return 1;
	}
	return 0;
}

3654
static int
3655 3656 3657
__lock_set_class(struct lockdep_map *lock, const char *name,
		 struct lock_class_key *key, unsigned int subclass,
		 unsigned long ip)
3658 3659
{
	struct task_struct *curr = current;
3660
	struct held_lock *hlock;
3661 3662 3663 3664 3665
	struct lock_class *class;
	unsigned int depth;
	int i;

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3666 3667 3668 3669
	/*
	 * This function is about (re)setting the class of a held lock,
	 * yet we're not actually holding any locks. Naughty user!
	 */
3670 3671 3672
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return 0;

3673 3674 3675
	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock)
		return print_unlock_imbalance_bug(curr, lock, ip);
3676

3677
	lockdep_init_map(lock, name, key, 0);
3678
	class = register_lock_class(lock, subclass, 0);
D
Dave Jones 已提交
3679
	hlock->class_idx = class - lock_classes + 1;
3680 3681 3682 3683

	curr->lockdep_depth = i;
	curr->curr_chain_key = hlock->prev_chain_key;

3684 3685
	if (reacquire_held_locks(curr, depth, i))
		return 0;
3686

P
Peter Zijlstra 已提交
3687 3688 3689 3690
	/*
	 * I took it apart and put it back together again, except now I have
	 * these 'spare' parts.. where shall I put them.
	 */
3691 3692 3693 3694 3695
	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth))
		return 0;
	return 1;
}

3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733
static int __lock_downgrade(struct lockdep_map *lock, unsigned long ip)
{
	struct task_struct *curr = current;
	struct held_lock *hlock;
	unsigned int depth;
	int i;

	depth = curr->lockdep_depth;
	/*
	 * This function is about (re)setting the class of a held lock,
	 * yet we're not actually holding any locks. Naughty user!
	 */
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return 0;

	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock)
		return print_unlock_imbalance_bug(curr, lock, ip);

	curr->lockdep_depth = i;
	curr->curr_chain_key = hlock->prev_chain_key;

	WARN(hlock->read, "downgrading a read lock");
	hlock->read = 1;
	hlock->acquire_ip = ip;

	if (reacquire_held_locks(curr, depth, i))
		return 0;

	/*
	 * I took it apart and put it back together again, except now I have
	 * these 'spare' parts.. where shall I put them.
	 */
	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth))
		return 0;
	return 1;
}

I
Ingo Molnar 已提交
3734
/*
3735 3736 3737 3738 3739
 * Remove the lock to the list of currently held locks - this gets
 * called on mutex_unlock()/spin_unlock*() (or on a failed
 * mutex_lock_interruptible()).
 *
 * @nested is an hysterical artifact, needs a tree wide cleanup.
I
Ingo Molnar 已提交
3740 3741
 */
static int
3742
__lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
I
Ingo Molnar 已提交
3743
{
3744
	struct task_struct *curr = current;
3745
	struct held_lock *hlock;
I
Ingo Molnar 已提交
3746
	unsigned int depth;
3747
	int ret, i;
I
Ingo Molnar 已提交
3748

3749 3750 3751
	if (unlikely(!debug_locks))
		return 0;

3752 3753 3754 3755 3756 3757 3758 3759
	ret = lock_release_crosslock(lock);
	/*
	 * 2 means normal release operations are needed. Otherwise, it's
	 * ok just to return with '0:fail, 1:success'.
	 */
	if (ret != 2)
		return ret;

I
Ingo Molnar 已提交
3760
	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3761 3762 3763 3764
	/*
	 * So we're all set to release this lock.. wait what lock? We don't
	 * own any locks, you've been drinking again?
	 */
3765 3766
	if (DEBUG_LOCKS_WARN_ON(depth <= 0))
		 return print_unlock_imbalance_bug(curr, lock, ip);
I
Ingo Molnar 已提交
3767

3768 3769 3770 3771
	/*
	 * Check whether the lock exists in the current stack
	 * of held locks:
	 */
3772 3773 3774
	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock)
		return print_unlock_imbalance_bug(curr, lock, ip);
I
Ingo Molnar 已提交
3775

3776 3777 3778
	if (hlock->instance == lock)
		lock_release_holdtime(hlock);

P
Peter Zijlstra 已提交
3779 3780
	WARN(hlock->pin_count, "releasing a pinned lock\n");

3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791
	if (hlock->references) {
		hlock->references--;
		if (hlock->references) {
			/*
			 * We had, and after removing one, still have
			 * references, the current lock stack is still
			 * valid. We're done!
			 */
			return 1;
		}
	}
P
Peter Zijlstra 已提交
3792

I
Ingo Molnar 已提交
3793 3794 3795 3796 3797
	/*
	 * We have the right lock to unlock, 'hlock' points to it.
	 * Now we remove it from the stack, and add back the other
	 * entries (if any), recalculating the hash along the way:
	 */
3798

I
Ingo Molnar 已提交
3799 3800 3801
	curr->lockdep_depth = i;
	curr->curr_chain_key = hlock->prev_chain_key;

3802 3803
	if (reacquire_held_locks(curr, depth, i + 1))
		return 0;
I
Ingo Molnar 已提交
3804

P
Peter Zijlstra 已提交
3805 3806 3807 3808
	/*
	 * We had N bottles of beer on the wall, we drank one, but now
	 * there's not N-1 bottles of beer left on the wall...
	 */
I
Ingo Molnar 已提交
3809 3810
	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth - 1))
		return 0;
P
Peter Zijlstra 已提交
3811

I
Ingo Molnar 已提交
3812 3813 3814
	return 1;
}

3815
static int __lock_is_held(struct lockdep_map *lock, int read)
I
Ingo Molnar 已提交
3816
{
3817 3818
	struct task_struct *curr = current;
	int i;
I
Ingo Molnar 已提交
3819

3820
	for (i = 0; i < curr->lockdep_depth; i++) {
3821
		struct held_lock *hlock = curr->held_locks + i;
I
Ingo Molnar 已提交
3822

3823 3824 3825 3826 3827 3828
		if (match_held_lock(hlock, lock)) {
			if (read == -1 || hlock->read == read)
				return 1;

			return 0;
		}
3829
	}
P
Peter Zijlstra 已提交
3830

3831
	return 0;
I
Ingo Molnar 已提交
3832 3833
}

3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862
static struct pin_cookie __lock_pin_lock(struct lockdep_map *lock)
{
	struct pin_cookie cookie = NIL_COOKIE;
	struct task_struct *curr = current;
	int i;

	if (unlikely(!debug_locks))
		return cookie;

	for (i = 0; i < curr->lockdep_depth; i++) {
		struct held_lock *hlock = curr->held_locks + i;

		if (match_held_lock(hlock, lock)) {
			/*
			 * Grab 16bits of randomness; this is sufficient to not
			 * be guessable and still allows some pin nesting in
			 * our u32 pin_count.
			 */
			cookie.val = 1 + (prandom_u32() >> 16);
			hlock->pin_count += cookie.val;
			return cookie;
		}
	}

	WARN(1, "pinning an unheld lock\n");
	return cookie;
}

static void __lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
I
Ingo Molnar 已提交
3863 3864
{
	struct task_struct *curr = current;
P
Peter Zijlstra 已提交
3865
	int i;
I
Ingo Molnar 已提交
3866

P
Peter Zijlstra 已提交
3867
	if (unlikely(!debug_locks))
I
Ingo Molnar 已提交
3868 3869
		return;

P
Peter Zijlstra 已提交
3870 3871 3872 3873
	for (i = 0; i < curr->lockdep_depth; i++) {
		struct held_lock *hlock = curr->held_locks + i;

		if (match_held_lock(hlock, lock)) {
3874
			hlock->pin_count += cookie.val;
I
Ingo Molnar 已提交
3875
			return;
P
Peter Zijlstra 已提交
3876
		}
I
Ingo Molnar 已提交
3877 3878
	}

P
Peter Zijlstra 已提交
3879
	WARN(1, "pinning an unheld lock\n");
I
Ingo Molnar 已提交
3880 3881
}

3882
static void __lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
3883 3884 3885 3886
{
	struct task_struct *curr = current;
	int i;

P
Peter Zijlstra 已提交
3887 3888 3889
	if (unlikely(!debug_locks))
		return;

3890
	for (i = 0; i < curr->lockdep_depth; i++) {
3891 3892
		struct held_lock *hlock = curr->held_locks + i;

P
Peter Zijlstra 已提交
3893 3894 3895 3896
		if (match_held_lock(hlock, lock)) {
			if (WARN(!hlock->pin_count, "unpinning an unpinned lock\n"))
				return;

3897 3898 3899 3900 3901
			hlock->pin_count -= cookie.val;

			if (WARN((int)hlock->pin_count < 0, "pin count corrupted\n"))
				hlock->pin_count = 0;

P
Peter Zijlstra 已提交
3902 3903
			return;
		}
3904 3905
	}

P
Peter Zijlstra 已提交
3906
	WARN(1, "unpinning an unheld lock\n");
3907 3908
}

I
Ingo Molnar 已提交
3909 3910 3911
/*
 * Check whether we follow the irq-flags state precisely:
 */
3912
static void check_flags(unsigned long flags)
I
Ingo Molnar 已提交
3913
{
3914 3915
#if defined(CONFIG_PROVE_LOCKING) && defined(CONFIG_DEBUG_LOCKDEP) && \
    defined(CONFIG_TRACE_IRQFLAGS)
I
Ingo Molnar 已提交
3916 3917 3918
	if (!debug_locks)
		return;

3919 3920 3921 3922 3923 3924 3925 3926 3927
	if (irqs_disabled_flags(flags)) {
		if (DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled)) {
			printk("possible reason: unannotated irqs-off.\n");
		}
	} else {
		if (DEBUG_LOCKS_WARN_ON(!current->hardirqs_enabled)) {
			printk("possible reason: unannotated irqs-on.\n");
		}
	}
I
Ingo Molnar 已提交
3928 3929 3930 3931 3932 3933 3934

	/*
	 * We dont accurately track softirq state in e.g.
	 * hardirq contexts (such as on 4KSTACKS), so only
	 * check if not in hardirq contexts:
	 */
	if (!hardirq_count()) {
P
Peter Zijlstra 已提交
3935 3936
		if (softirq_count()) {
			/* like the above, but with softirqs */
I
Ingo Molnar 已提交
3937
			DEBUG_LOCKS_WARN_ON(current->softirqs_enabled);
P
Peter Zijlstra 已提交
3938 3939
		} else {
			/* lick the above, does it taste good? */
I
Ingo Molnar 已提交
3940
			DEBUG_LOCKS_WARN_ON(!current->softirqs_enabled);
P
Peter Zijlstra 已提交
3941
		}
I
Ingo Molnar 已提交
3942 3943 3944 3945 3946 3947 3948
	}

	if (!debug_locks)
		print_irqtrace_events(current);
#endif
}

3949 3950 3951
void lock_set_class(struct lockdep_map *lock, const char *name,
		    struct lock_class_key *key, unsigned int subclass,
		    unsigned long ip)
3952 3953 3954 3955 3956 3957 3958 3959 3960
{
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	current->lockdep_recursion = 1;
	check_flags(flags);
3961
	if (__lock_set_class(lock, name, key, subclass, ip))
3962 3963 3964 3965
		check_chain_key(current);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
3966
EXPORT_SYMBOL_GPL(lock_set_class);
3967

3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984
void lock_downgrade(struct lockdep_map *lock, unsigned long ip)
{
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	current->lockdep_recursion = 1;
	check_flags(flags);
	if (__lock_downgrade(lock, ip))
		check_chain_key(current);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_downgrade);

I
Ingo Molnar 已提交
3985 3986 3987 3988
/*
 * We are not always called with irqs disabled - do that here,
 * and also avoid lockdep recursion:
 */
3989
void lock_acquire(struct lockdep_map *lock, unsigned int subclass,
P
Peter Zijlstra 已提交
3990 3991
			  int trylock, int read, int check,
			  struct lockdep_map *nest_lock, unsigned long ip)
I
Ingo Molnar 已提交
3992 3993 3994 3995 3996 3997 3998 3999 4000 4001
{
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4002
	trace_lock_acquire(lock, subclass, trylock, read, check, nest_lock, ip);
I
Ingo Molnar 已提交
4003
	__lock_acquire(lock, subclass, trylock, read, check,
4004
		       irqs_disabled_flags(flags), nest_lock, ip, 0, 0);
I
Ingo Molnar 已提交
4005 4006 4007 4008 4009
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquire);

4010
void lock_release(struct lockdep_map *lock, int nested,
4011
			  unsigned long ip)
I
Ingo Molnar 已提交
4012 4013 4014 4015 4016 4017 4018 4019 4020
{
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
4021
	trace_lock_release(lock, ip);
4022 4023
	if (__lock_release(lock, nested, ip))
		check_chain_key(current);
I
Ingo Molnar 已提交
4024 4025 4026 4027 4028
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_release);

4029
int lock_is_held_type(struct lockdep_map *lock, int read)
4030 4031 4032 4033 4034
{
	unsigned long flags;
	int ret = 0;

	if (unlikely(current->lockdep_recursion))
4035
		return 1; /* avoid false negative lockdep_assert_held() */
4036 4037 4038 4039 4040

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4041
	ret = __lock_is_held(lock, read);
4042 4043 4044 4045 4046
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);

	return ret;
}
4047
EXPORT_SYMBOL_GPL(lock_is_held_type);
4048

4049
struct pin_cookie lock_pin_lock(struct lockdep_map *lock)
P
Peter Zijlstra 已提交
4050
{
4051
	struct pin_cookie cookie = NIL_COOKIE;
P
Peter Zijlstra 已提交
4052 4053 4054
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
4055
		return cookie;
P
Peter Zijlstra 已提交
4056 4057 4058 4059 4060

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4061
	cookie = __lock_pin_lock(lock);
P
Peter Zijlstra 已提交
4062 4063
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
4064 4065

	return cookie;
P
Peter Zijlstra 已提交
4066 4067 4068
}
EXPORT_SYMBOL_GPL(lock_pin_lock);

4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086
void lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
{
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
	__lock_repin_lock(lock, cookie);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_repin_lock);

void lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
P
Peter Zijlstra 已提交
4087 4088 4089 4090 4091 4092 4093 4094 4095 4096
{
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4097
	__lock_unpin_lock(lock, cookie);
P
Peter Zijlstra 已提交
4098 4099 4100 4101 4102
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_unpin_lock);

P
Peter Zijlstra 已提交
4103 4104 4105 4106 4107 4108 4109 4110 4111 4112
#ifdef CONFIG_LOCK_STAT
static int
print_lock_contention_bug(struct task_struct *curr, struct lockdep_map *lock,
			   unsigned long ip)
{
	if (!debug_locks_off())
		return 0;
	if (debug_locks_silent)
		return 0;

4113
	pr_warn("\n");
4114 4115
	pr_warn("=================================\n");
	pr_warn("WARNING: bad contention detected!\n");
4116
	print_kernel_ident();
4117
	pr_warn("---------------------------------\n");
4118
	pr_warn("%s/%d is trying to contend lock (",
4119
		curr->comm, task_pid_nr(curr));
P
Peter Zijlstra 已提交
4120
	print_lockdep_cache(lock);
4121
	pr_cont(") at:\n");
P
Peter Zijlstra 已提交
4122
	print_ip_sym(ip);
4123 4124
	pr_warn("but there are no locks held!\n");
	pr_warn("\nother info that might help us debug this:\n");
P
Peter Zijlstra 已提交
4125 4126
	lockdep_print_held_locks(curr);

4127
	pr_warn("\nstack backtrace:\n");
P
Peter Zijlstra 已提交
4128 4129 4130 4131 4132 4133 4134 4135 4136
	dump_stack();

	return 0;
}

static void
__lock_contended(struct lockdep_map *lock, unsigned long ip)
{
	struct task_struct *curr = current;
4137
	struct held_lock *hlock;
P
Peter Zijlstra 已提交
4138 4139
	struct lock_class_stats *stats;
	unsigned int depth;
P
Peter Zijlstra 已提交
4140
	int i, contention_point, contending_point;
P
Peter Zijlstra 已提交
4141 4142

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
4143 4144 4145 4146
	/*
	 * Whee, we contended on this lock, except it seems we're not
	 * actually trying to acquire anything much at all..
	 */
P
Peter Zijlstra 已提交
4147 4148 4149
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return;

4150 4151 4152 4153
	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock) {
		print_lock_contention_bug(curr, lock, ip);
		return;
P
Peter Zijlstra 已提交
4154 4155
	}

4156 4157 4158
	if (hlock->instance != lock)
		return;

4159
	hlock->waittime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
4160

P
Peter Zijlstra 已提交
4161 4162 4163
	contention_point = lock_point(hlock_class(hlock)->contention_point, ip);
	contending_point = lock_point(hlock_class(hlock)->contending_point,
				      lock->ip);
P
Peter Zijlstra 已提交
4164

D
Dave Jones 已提交
4165
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
4166 4167 4168 4169
	if (contention_point < LOCKSTAT_POINTS)
		stats->contention_point[contention_point]++;
	if (contending_point < LOCKSTAT_POINTS)
		stats->contending_point[contending_point]++;
P
Peter Zijlstra 已提交
4170 4171
	if (lock->cpu != smp_processor_id())
		stats->bounces[bounce_contended + !!hlock->read]++;
P
Peter Zijlstra 已提交
4172 4173 4174 4175
	put_lock_stats(stats);
}

static void
P
Peter Zijlstra 已提交
4176
__lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
4177 4178
{
	struct task_struct *curr = current;
4179
	struct held_lock *hlock;
P
Peter Zijlstra 已提交
4180 4181
	struct lock_class_stats *stats;
	unsigned int depth;
4182
	u64 now, waittime = 0;
P
Peter Zijlstra 已提交
4183
	int i, cpu;
P
Peter Zijlstra 已提交
4184 4185

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
4186 4187 4188 4189
	/*
	 * Yay, we acquired ownership of this lock we didn't try to
	 * acquire, how the heck did that happen?
	 */
P
Peter Zijlstra 已提交
4190 4191 4192
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return;

4193 4194 4195 4196
	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock) {
		print_lock_contention_bug(curr, lock, _RET_IP_);
		return;
P
Peter Zijlstra 已提交
4197 4198
	}

4199 4200 4201
	if (hlock->instance != lock)
		return;

P
Peter Zijlstra 已提交
4202 4203
	cpu = smp_processor_id();
	if (hlock->waittime_stamp) {
4204
		now = lockstat_clock();
P
Peter Zijlstra 已提交
4205 4206 4207
		waittime = now - hlock->waittime_stamp;
		hlock->holdtime_stamp = now;
	}
P
Peter Zijlstra 已提交
4208

4209
	trace_lock_acquired(lock, ip);
4210

D
Dave Jones 已提交
4211
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
4212 4213 4214 4215 4216 4217 4218 4219
	if (waittime) {
		if (hlock->read)
			lock_time_inc(&stats->read_waittime, waittime);
		else
			lock_time_inc(&stats->write_waittime, waittime);
	}
	if (lock->cpu != cpu)
		stats->bounces[bounce_acquired + !!hlock->read]++;
P
Peter Zijlstra 已提交
4220
	put_lock_stats(stats);
P
Peter Zijlstra 已提交
4221 4222

	lock->cpu = cpu;
P
Peter Zijlstra 已提交
4223
	lock->ip = ip;
P
Peter Zijlstra 已提交
4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238
}

void lock_contended(struct lockdep_map *lock, unsigned long ip)
{
	unsigned long flags;

	if (unlikely(!lock_stat))
		return;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
4239
	trace_lock_contended(lock, ip);
P
Peter Zijlstra 已提交
4240 4241 4242 4243 4244 4245
	__lock_contended(lock, ip);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_contended);

P
Peter Zijlstra 已提交
4246
void lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258
{
	unsigned long flags;

	if (unlikely(!lock_stat))
		return;

	if (unlikely(current->lockdep_recursion))
		return;

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
P
Peter Zijlstra 已提交
4259
	__lock_acquired(lock, ip);
P
Peter Zijlstra 已提交
4260 4261 4262 4263 4264 4265
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquired);
#endif

I
Ingo Molnar 已提交
4266 4267 4268 4269 4270 4271 4272 4273
/*
 * Used by the testsuite, sanitize the validator state
 * after a simulated failure:
 */

void lockdep_reset(void)
{
	unsigned long flags;
4274
	int i;
I
Ingo Molnar 已提交
4275 4276 4277 4278 4279 4280 4281 4282 4283 4284

	raw_local_irq_save(flags);
	current->curr_chain_key = 0;
	current->lockdep_depth = 0;
	current->lockdep_recursion = 0;
	memset(current->held_locks, 0, MAX_LOCK_DEPTH*sizeof(struct held_lock));
	nr_hardirq_chains = 0;
	nr_softirq_chains = 0;
	nr_process_chains = 0;
	debug_locks = 1;
4285
	for (i = 0; i < CHAINHASH_SIZE; i++)
4286
		INIT_HLIST_HEAD(chainhash_table + i);
I
Ingo Molnar 已提交
4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304
	raw_local_irq_restore(flags);
}

static void zap_class(struct lock_class *class)
{
	int i;

	/*
	 * Remove all dependencies this lock is
	 * involved in:
	 */
	for (i = 0; i < nr_list_entries; i++) {
		if (list_entries[i].class == class)
			list_del_rcu(&list_entries[i].entry);
	}
	/*
	 * Unhash the class and remove it from the all_lock_classes list:
	 */
4305
	hlist_del_rcu(&class->hash_entry);
I
Ingo Molnar 已提交
4306 4307
	list_del_rcu(&class->lock_entry);

4308 4309
	RCU_INIT_POINTER(class->key, NULL);
	RCU_INIT_POINTER(class->name, NULL);
I
Ingo Molnar 已提交
4310 4311
}

4312
static inline int within(const void *addr, void *start, unsigned long size)
I
Ingo Molnar 已提交
4313 4314 4315 4316
{
	return addr >= start && addr < start + size;
}

4317 4318 4319 4320 4321 4322 4323 4324
/*
 * Used in module.c to remove lock classes from memory that is going to be
 * freed; and possibly re-used by other modules.
 *
 * We will have had one sync_sched() before getting here, so we're guaranteed
 * nobody will look up these exact classes -- they're properly dead but still
 * allocated.
 */
I
Ingo Molnar 已提交
4325 4326
void lockdep_free_key_range(void *start, unsigned long size)
{
4327
	struct lock_class *class;
4328
	struct hlist_head *head;
I
Ingo Molnar 已提交
4329 4330
	unsigned long flags;
	int i;
4331
	int locked;
I
Ingo Molnar 已提交
4332 4333

	raw_local_irq_save(flags);
4334
	locked = graph_lock();
I
Ingo Molnar 已提交
4335 4336 4337 4338 4339 4340

	/*
	 * Unhash all classes that were created by this module:
	 */
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
4341
		hlist_for_each_entry_rcu(class, head, hash_entry) {
I
Ingo Molnar 已提交
4342 4343
			if (within(class->key, start, size))
				zap_class(class);
4344 4345 4346
			else if (within(class->name, start, size))
				zap_class(class);
		}
I
Ingo Molnar 已提交
4347 4348
	}

4349 4350
	if (locked)
		graph_unlock();
I
Ingo Molnar 已提交
4351
	raw_local_irq_restore(flags);
4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365

	/*
	 * Wait for any possible iterators from look_up_lock_class() to pass
	 * before continuing to free the memory they refer to.
	 *
	 * sync_sched() is sufficient because the read-side is IRQ disable.
	 */
	synchronize_sched();

	/*
	 * XXX at this point we could return the resources to the pool;
	 * instead we leak them. We would need to change to bitmap allocators
	 * instead of the linear allocators we have now.
	 */
I
Ingo Molnar 已提交
4366 4367 4368 4369
}

void lockdep_reset_lock(struct lockdep_map *lock)
{
4370
	struct lock_class *class;
4371
	struct hlist_head *head;
I
Ingo Molnar 已提交
4372 4373
	unsigned long flags;
	int i, j;
4374
	int locked;
I
Ingo Molnar 已提交
4375 4376 4377 4378

	raw_local_irq_save(flags);

	/*
4379 4380 4381 4382 4383 4384 4385
	 * Remove all classes this lock might have:
	 */
	for (j = 0; j < MAX_LOCKDEP_SUBCLASSES; j++) {
		/*
		 * If the class exists we look it up and zap it:
		 */
		class = look_up_lock_class(lock, j);
4386
		if (!IS_ERR_OR_NULL(class))
4387 4388 4389 4390 4391
			zap_class(class);
	}
	/*
	 * Debug check: in the end all mapped classes should
	 * be gone.
I
Ingo Molnar 已提交
4392
	 */
4393
	locked = graph_lock();
I
Ingo Molnar 已提交
4394 4395
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
4396
		hlist_for_each_entry_rcu(class, head, hash_entry) {
4397 4398 4399 4400 4401 4402
			int match = 0;

			for (j = 0; j < NR_LOCKDEP_CACHING_CLASSES; j++)
				match |= class == lock->class_cache[j];

			if (unlikely(match)) {
P
Peter Zijlstra 已提交
4403 4404 4405 4406
				if (debug_locks_off_graph_unlock()) {
					/*
					 * We all just reset everything, how did it match?
					 */
4407
					WARN_ON(1);
P
Peter Zijlstra 已提交
4408
				}
4409
				goto out_restore;
I
Ingo Molnar 已提交
4410 4411 4412
			}
		}
	}
4413 4414
	if (locked)
		graph_unlock();
4415 4416

out_restore:
I
Ingo Molnar 已提交
4417 4418 4419 4420 4421 4422 4423
	raw_local_irq_restore(flags);
}

void __init lockdep_info(void)
{
	printk("Lock dependency validator: Copyright (c) 2006 Red Hat, Inc., Ingo Molnar\n");

4424
	printk("... MAX_LOCKDEP_SUBCLASSES:  %lu\n", MAX_LOCKDEP_SUBCLASSES);
I
Ingo Molnar 已提交
4425 4426
	printk("... MAX_LOCK_DEPTH:          %lu\n", MAX_LOCK_DEPTH);
	printk("... MAX_LOCKDEP_KEYS:        %lu\n", MAX_LOCKDEP_KEYS);
4427
	printk("... CLASSHASH_SIZE:          %lu\n", CLASSHASH_SIZE);
I
Ingo Molnar 已提交
4428 4429 4430 4431 4432 4433 4434 4435 4436
	printk("... MAX_LOCKDEP_ENTRIES:     %lu\n", MAX_LOCKDEP_ENTRIES);
	printk("... MAX_LOCKDEP_CHAINS:      %lu\n", MAX_LOCKDEP_CHAINS);
	printk("... CHAINHASH_SIZE:          %lu\n", CHAINHASH_SIZE);

	printk(" memory used by lock dependency info: %lu kB\n",
		(sizeof(struct lock_class) * MAX_LOCKDEP_KEYS +
		sizeof(struct list_head) * CLASSHASH_SIZE +
		sizeof(struct lock_list) * MAX_LOCKDEP_ENTRIES +
		sizeof(struct lock_chain) * MAX_LOCKDEP_CHAINS +
M
Ming Lei 已提交
4437
		sizeof(struct list_head) * CHAINHASH_SIZE
4438
#ifdef CONFIG_PROVE_LOCKING
4439
		+ sizeof(struct circular_queue)
4440
#endif
M
Ming Lei 已提交
4441
		) / 1024
4442
		);
I
Ingo Molnar 已提交
4443 4444 4445 4446 4447 4448 4449

	printk(" per task-struct memory footprint: %lu bytes\n",
		sizeof(struct held_lock) * MAX_LOCK_DEPTH);
}

static void
print_freed_lock_bug(struct task_struct *curr, const void *mem_from,
4450
		     const void *mem_to, struct held_lock *hlock)
I
Ingo Molnar 已提交
4451 4452 4453 4454 4455 4456
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

4457
	pr_warn("\n");
4458 4459
	pr_warn("=========================\n");
	pr_warn("WARNING: held lock freed!\n");
4460
	print_kernel_ident();
4461
	pr_warn("-------------------------\n");
4462
	pr_warn("%s/%d is freeing memory %p-%p, with a lock still held there!\n",
4463
		curr->comm, task_pid_nr(curr), mem_from, mem_to-1);
4464
	print_lock(hlock);
I
Ingo Molnar 已提交
4465 4466
	lockdep_print_held_locks(curr);

4467
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
4468 4469 4470
	dump_stack();
}

O
Oleg Nesterov 已提交
4471 4472 4473 4474 4475 4476 4477
static inline int not_in_range(const void* mem_from, unsigned long mem_len,
				const void* lock_from, unsigned long lock_len)
{
	return lock_from + lock_len <= mem_from ||
		mem_from + mem_len <= lock_from;
}

I
Ingo Molnar 已提交
4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496
/*
 * Called when kernel memory is freed (or unmapped), or if a lock
 * is destroyed or reinitialized - this code checks whether there is
 * any held lock in the memory range of <from> to <to>:
 */
void debug_check_no_locks_freed(const void *mem_from, unsigned long mem_len)
{
	struct task_struct *curr = current;
	struct held_lock *hlock;
	unsigned long flags;
	int i;

	if (unlikely(!debug_locks))
		return;

	local_irq_save(flags);
	for (i = 0; i < curr->lockdep_depth; i++) {
		hlock = curr->held_locks + i;

O
Oleg Nesterov 已提交
4497 4498
		if (not_in_range(mem_from, mem_len, hlock->instance,
					sizeof(*hlock->instance)))
I
Ingo Molnar 已提交
4499 4500
			continue;

O
Oleg Nesterov 已提交
4501
		print_freed_lock_bug(curr, mem_from, mem_from + mem_len, hlock);
I
Ingo Molnar 已提交
4502 4503 4504 4505
		break;
	}
	local_irq_restore(flags);
}
4506
EXPORT_SYMBOL_GPL(debug_check_no_locks_freed);
I
Ingo Molnar 已提交
4507

4508
static void print_held_locks_bug(void)
I
Ingo Molnar 已提交
4509 4510 4511 4512 4513 4514
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

4515
	pr_warn("\n");
4516 4517
	pr_warn("====================================\n");
	pr_warn("WARNING: %s/%d still has locks held!\n",
4518
	       current->comm, task_pid_nr(current));
4519
	print_kernel_ident();
4520
	pr_warn("------------------------------------\n");
4521
	lockdep_print_held_locks(current);
4522
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
4523 4524 4525
	dump_stack();
}

4526
void debug_check_no_locks_held(void)
I
Ingo Molnar 已提交
4527
{
4528 4529
	if (unlikely(current->lockdep_depth > 0))
		print_held_locks_bug();
I
Ingo Molnar 已提交
4530
}
4531
EXPORT_SYMBOL_GPL(debug_check_no_locks_held);
I
Ingo Molnar 已提交
4532

4533
#ifdef __KERNEL__
I
Ingo Molnar 已提交
4534 4535 4536 4537 4538 4539
void debug_show_all_locks(void)
{
	struct task_struct *g, *p;
	int count = 10;
	int unlock = 1;

4540
	if (unlikely(!debug_locks)) {
4541
		pr_warn("INFO: lockdep is turned off.\n");
4542 4543
		return;
	}
4544
	pr_warn("\nShowing all locks held in the system:\n");
I
Ingo Molnar 已提交
4545 4546 4547 4548 4549 4550 4551 4552 4553 4554

	/*
	 * Here we try to get the tasklist_lock as hard as possible,
	 * if not successful after 2 seconds we ignore it (but keep
	 * trying). This is to enable a debug printout even if a
	 * tasklist_lock-holding task deadlocks or crashes.
	 */
retry:
	if (!read_trylock(&tasklist_lock)) {
		if (count == 10)
4555
			pr_warn("hm, tasklist_lock locked, retrying... ");
I
Ingo Molnar 已提交
4556 4557
		if (count) {
			count--;
4558
			pr_cont(" #%d", 10-count);
I
Ingo Molnar 已提交
4559 4560 4561
			mdelay(200);
			goto retry;
		}
4562
		pr_cont(" ignoring it.\n");
I
Ingo Molnar 已提交
4563
		unlock = 0;
4564 4565
	} else {
		if (count != 10)
4566
			pr_cont(" locked it.\n");
I
Ingo Molnar 已提交
4567 4568 4569
	}

	do_each_thread(g, p) {
I
Ingo Molnar 已提交
4570 4571 4572 4573 4574 4575 4576
		/*
		 * It's not reliable to print a task's held locks
		 * if it's not sleeping (or if it's not the current
		 * task):
		 */
		if (p->state == TASK_RUNNING && p != current)
			continue;
I
Ingo Molnar 已提交
4577 4578 4579 4580 4581 4582 4583
		if (p->lockdep_depth)
			lockdep_print_held_locks(p);
		if (!unlock)
			if (read_trylock(&tasklist_lock))
				unlock = 1;
	} while_each_thread(g, p);

4584
	pr_warn("\n");
4585
	pr_warn("=============================================\n\n");
I
Ingo Molnar 已提交
4586 4587 4588 4589 4590

	if (unlock)
		read_unlock(&tasklist_lock);
}
EXPORT_SYMBOL_GPL(debug_show_all_locks);
4591
#endif
I
Ingo Molnar 已提交
4592

4593 4594 4595 4596
/*
 * Careful: only use this function if you are sure that
 * the task cannot run in parallel!
 */
4597
void debug_show_held_locks(struct task_struct *task)
I
Ingo Molnar 已提交
4598
{
4599 4600 4601 4602
	if (unlikely(!debug_locks)) {
		printk("INFO: lockdep is turned off.\n");
		return;
	}
I
Ingo Molnar 已提交
4603 4604 4605
	lockdep_print_held_locks(task);
}
EXPORT_SYMBOL_GPL(debug_show_held_locks);
P
Peter Zijlstra 已提交
4606

4607
asmlinkage __visible void lockdep_sys_exit(void)
P
Peter Zijlstra 已提交
4608 4609 4610 4611 4612 4613
{
	struct task_struct *curr = current;

	if (unlikely(curr->lockdep_depth)) {
		if (!debug_locks_off())
			return;
4614
		pr_warn("\n");
4615 4616
		pr_warn("================================================\n");
		pr_warn("WARNING: lock held when returning to user space!\n");
4617
		print_kernel_ident();
4618
		pr_warn("------------------------------------------------\n");
4619
		pr_warn("%s/%d is leaving the kernel with locks still held!\n",
P
Peter Zijlstra 已提交
4620 4621 4622
				curr->comm, curr->pid);
		lockdep_print_held_locks(curr);
	}
4623 4624 4625 4626 4627

	/*
	 * The lock history for each syscall should be independent. So wipe the
	 * slate clean on return to userspace.
	 */
4628
	lockdep_invariant_state(false);
P
Peter Zijlstra 已提交
4629
}
4630

4631
void lockdep_rcu_suspicious(const char *file, const int line, const char *s)
4632 4633 4634
{
	struct task_struct *curr = current;

4635
	/* Note: the following can be executed concurrently, so be careful. */
4636
	pr_warn("\n");
4637 4638
	pr_warn("=============================\n");
	pr_warn("WARNING: suspicious RCU usage\n");
4639
	print_kernel_ident();
4640
	pr_warn("-----------------------------\n");
4641 4642 4643
	pr_warn("%s:%d %s!\n", file, line, s);
	pr_warn("\nother info that might help us debug this:\n\n");
	pr_warn("\n%srcu_scheduler_active = %d, debug_locks = %d\n",
4644 4645
	       !rcu_lockdep_current_cpu_online()
			? "RCU used illegally from offline CPU!\n"
4646
			: !rcu_is_watching()
4647 4648 4649
				? "RCU used illegally from idle CPU!\n"
				: "",
	       rcu_scheduler_active, debug_locks);
4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668

	/*
	 * If a CPU is in the RCU-free window in idle (ie: in the section
	 * between rcu_idle_enter() and rcu_idle_exit(), then RCU
	 * considers that CPU to be in an "extended quiescent state",
	 * which means that RCU will be completely ignoring that CPU.
	 * Therefore, rcu_read_lock() and friends have absolutely no
	 * effect on a CPU running in that state. In other words, even if
	 * such an RCU-idle CPU has called rcu_read_lock(), RCU might well
	 * delete data structures out from under it.  RCU really has no
	 * choice here: we need to keep an RCU-free window in idle where
	 * the CPU may possibly enter into low power mode. This way we can
	 * notice an extended quiescent state to other CPUs that started a grace
	 * period. Otherwise we would delay any grace period as long as we run
	 * in the idle task.
	 *
	 * So complain bitterly if someone does call rcu_read_lock(),
	 * rcu_read_lock_bh() and so on from extended quiescent states.
	 */
4669
	if (!rcu_is_watching())
4670
		pr_warn("RCU used illegally from extended quiescent state!\n");
4671

4672
	lockdep_print_held_locks(curr);
4673
	pr_warn("\nstack backtrace:\n");
4674 4675
	dump_stack();
}
4676
EXPORT_SYMBOL_GPL(lockdep_rcu_suspicious);
4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709

#ifdef CONFIG_LOCKDEP_CROSSRELEASE

/*
 * Crossrelease works by recording a lock history for each thread and
 * connecting those historic locks that were taken after the
 * wait_for_completion() in the complete() context.
 *
 * Task-A				Task-B
 *
 *					mutex_lock(&A);
 *					mutex_unlock(&A);
 *
 * wait_for_completion(&C);
 *   lock_acquire_crosslock();
 *     atomic_inc_return(&cross_gen_id);
 *                                |
 *				  |	mutex_lock(&B);
 *				  |	mutex_unlock(&B);
 *                                |
 *				  |	complete(&C);
 *				  `--	  lock_commit_crosslock();
 *
 * Which will then add a dependency between B and C.
 */

#define xhlock(i)         (current->xhlocks[(i) % MAX_XHLOCKS_NR])

/*
 * Whenever a crosslock is held, cross_gen_id will be increased.
 */
static atomic_t cross_gen_id; /* Can be wrapped */

4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720
/*
 * Make an entry of the ring buffer invalid.
 */
static inline void invalidate_xhlock(struct hist_lock *xhlock)
{
	/*
	 * Normally, xhlock->hlock.instance must be !NULL.
	 */
	xhlock->hlock.instance = NULL;
}

4721
/*
4722
 * Lock history stacks; we have 2 nested lock history stacks:
4723
 *
4724 4725
 *   HARD(IRQ)
 *   SOFT(IRQ)
4726
 *
4727 4728 4729 4730
 * The thing is that once we complete a HARD/SOFT IRQ the future task locks
 * should not depend on any of the locks observed while running the IRQ.  So
 * what we do is rewind the history buffer and erase all our knowledge of that
 * temporal event.
4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762
 */

void crossrelease_hist_start(enum xhlock_context_t c)
{
	struct task_struct *cur = current;

	if (!cur->xhlocks)
		return;

	cur->xhlock_idx_hist[c] = cur->xhlock_idx;
	cur->hist_id_save[c]    = cur->hist_id;
}

void crossrelease_hist_end(enum xhlock_context_t c)
{
	struct task_struct *cur = current;

	if (cur->xhlocks) {
		unsigned int idx = cur->xhlock_idx_hist[c];
		struct hist_lock *h = &xhlock(idx);

		cur->xhlock_idx = idx;

		/* Check if the ring was overwritten. */
		if (h->hist_id != cur->hist_id_save[c])
			invalidate_xhlock(h);
	}
}

/*
 * lockdep_invariant_state() is used to annotate independence inside a task, to
 * make one task look like multiple independent 'tasks'.
4763 4764 4765 4766 4767
 *
 * Take for instance workqueues; each work is independent of the last. The
 * completion of a future work does not depend on the completion of a past work
 * (in general). Therefore we must not carry that (lock) dependency across
 * works.
4768 4769
 *
 * This is true for many things; pretty much all kthreads fall into this
4770
 * pattern, where they have an invariant state and future completions do not
4771 4772 4773 4774 4775 4776 4777
 * depend on past completions. Its just that since they all have the 'same'
 * form -- the kthread does the same over and over -- it doesn't typically
 * matter.
 *
 * The same is true for system-calls, once a system call is completed (we've
 * returned to userspace) the next system call does not depend on the lock
 * history of the previous system call.
4778 4779 4780 4781 4782 4783
 *
 * They key property for independence, this invariant state, is that it must be
 * a point where we hold no locks and have no history. Because if we were to
 * hold locks, the restore at _end() would not necessarily recover it's history
 * entry. Similarly, independence per-definition means it does not depend on
 * prior state.
4784
 */
4785
void lockdep_invariant_state(bool force)
4786
{
4787 4788
	/*
	 * We call this at an invariant point, no current state, no history.
4789
	 * Verify the former, enforce the latter.
4790
	 */
4791 4792
	WARN_ON_ONCE(!force && current->lockdep_depth);
	invalidate_xhlock(&xhlock(current->xhlock_idx));
4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837
}

static int cross_lock(struct lockdep_map *lock)
{
	return lock ? lock->cross : 0;
}

/*
 * This is needed to decide the relationship between wrapable variables.
 */
static inline int before(unsigned int a, unsigned int b)
{
	return (int)(a - b) < 0;
}

static inline struct lock_class *xhlock_class(struct hist_lock *xhlock)
{
	return hlock_class(&xhlock->hlock);
}

static inline struct lock_class *xlock_class(struct cross_lock *xlock)
{
	return hlock_class(&xlock->hlock);
}

/*
 * Should we check a dependency with previous one?
 */
static inline int depend_before(struct held_lock *hlock)
{
	return hlock->read != 2 && hlock->check && !hlock->trylock;
}

/*
 * Should we check a dependency with next one?
 */
static inline int depend_after(struct held_lock *hlock)
{
	return hlock->read != 2 && hlock->check;
}

/*
 * Check if the xhlock is valid, which would be false if,
 *
 *    1. Has not used after initializaion yet.
4838
 *    2. Got invalidated.
4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869
 *
 * Remind hist_lock is implemented as a ring buffer.
 */
static inline int xhlock_valid(struct hist_lock *xhlock)
{
	/*
	 * xhlock->hlock.instance must be !NULL.
	 */
	return !!xhlock->hlock.instance;
}

/*
 * Record a hist_lock entry.
 *
 * Irq disable is only required.
 */
static void add_xhlock(struct held_lock *hlock)
{
	unsigned int idx = ++current->xhlock_idx;
	struct hist_lock *xhlock = &xhlock(idx);

#ifdef CONFIG_DEBUG_LOCKDEP
	/*
	 * This can be done locklessly because they are all task-local
	 * state, we must however ensure IRQs are disabled.
	 */
	WARN_ON_ONCE(!irqs_disabled());
#endif

	/* Initialize hist_lock's members */
	xhlock->hlock = *hlock;
4870
	xhlock->hist_id = ++current->hist_id;
4871 4872 4873 4874

	xhlock->trace.nr_entries = 0;
	xhlock->trace.max_entries = MAX_XHLOCK_TRACE_ENTRIES;
	xhlock->trace.entries = xhlock->trace_entries;
4875 4876 4877 4878 4879 4880 4881 4882

	if (crossrelease_fullstack) {
		xhlock->trace.skip = 3;
		save_stack_trace(&xhlock->trace);
	} else {
		xhlock->trace.nr_entries = 1;
		xhlock->trace.entries[0] = hlock->acquire_ip;
	}
4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920
}

static inline int same_context_xhlock(struct hist_lock *xhlock)
{
	return xhlock->hlock.irq_context == task_irq_context(current);
}

/*
 * This should be lockless as far as possible because this would be
 * called very frequently.
 */
static void check_add_xhlock(struct held_lock *hlock)
{
	/*
	 * Record a hist_lock, only in case that acquisitions ahead
	 * could depend on the held_lock. For example, if the held_lock
	 * is trylock then acquisitions ahead never depends on that.
	 * In that case, we don't need to record it. Just return.
	 */
	if (!current->xhlocks || !depend_before(hlock))
		return;

	add_xhlock(hlock);
}

/*
 * For crosslock.
 */
static int add_xlock(struct held_lock *hlock)
{
	struct cross_lock *xlock;
	unsigned int gen_id;

	if (!graph_lock())
		return 0;

	xlock = &((struct lockdep_map_cross *)hlock->instance)->xlock;

4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937
	/*
	 * When acquisitions for a crosslock are overlapped, we use
	 * nr_acquire to perform commit for them, based on cross_gen_id
	 * of the first acquisition, which allows to add additional
	 * dependencies.
	 *
	 * Moreover, when no acquisition of a crosslock is in progress,
	 * we should not perform commit because the lock might not exist
	 * any more, which might cause incorrect memory access. So we
	 * have to track the number of acquisitions of a crosslock.
	 *
	 * depend_after() is necessary to initialize only the first
	 * valid xlock so that the xlock can be used on its commit.
	 */
	if (xlock->nr_acquire++ && depend_after(&xlock->hlock))
		goto unlock;

4938 4939 4940
	gen_id = (unsigned int)atomic_inc_return(&cross_gen_id);
	xlock->hlock = *hlock;
	xlock->hlock.gen_id = gen_id;
4941
unlock:
4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031
	graph_unlock();
	return 1;
}

/*
 * Called for both normal and crosslock acquires. Normal locks will be
 * pushed on the hist_lock queue. Cross locks will record state and
 * stop regular lock_acquire() to avoid being placed on the held_lock
 * stack.
 *
 * Return: 0 - failure;
 *         1 - crosslock, done;
 *         2 - normal lock, continue to held_lock[] ops.
 */
static int lock_acquire_crosslock(struct held_lock *hlock)
{
	/*
	 *	CONTEXT 1		CONTEXT 2
	 *	---------		---------
	 *	lock A (cross)
	 *	X = atomic_inc_return(&cross_gen_id)
	 *	~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
	 *				Y = atomic_read_acquire(&cross_gen_id)
	 *				lock B
	 *
	 * atomic_read_acquire() is for ordering between A and B,
	 * IOW, A happens before B, when CONTEXT 2 see Y >= X.
	 *
	 * Pairs with atomic_inc_return() in add_xlock().
	 */
	hlock->gen_id = (unsigned int)atomic_read_acquire(&cross_gen_id);

	if (cross_lock(hlock->instance))
		return add_xlock(hlock);

	check_add_xhlock(hlock);
	return 2;
}

static int copy_trace(struct stack_trace *trace)
{
	unsigned long *buf = stack_trace + nr_stack_trace_entries;
	unsigned int max_nr = MAX_STACK_TRACE_ENTRIES - nr_stack_trace_entries;
	unsigned int nr = min(max_nr, trace->nr_entries);

	trace->nr_entries = nr;
	memcpy(buf, trace->entries, nr * sizeof(trace->entries[0]));
	trace->entries = buf;
	nr_stack_trace_entries += nr;

	if (nr_stack_trace_entries >= MAX_STACK_TRACE_ENTRIES-1) {
		if (!debug_locks_off_graph_unlock())
			return 0;

		print_lockdep_off("BUG: MAX_STACK_TRACE_ENTRIES too low!");
		dump_stack();

		return 0;
	}

	return 1;
}

static int commit_xhlock(struct cross_lock *xlock, struct hist_lock *xhlock)
{
	unsigned int xid, pid;
	u64 chain_key;

	xid = xlock_class(xlock) - lock_classes;
	chain_key = iterate_chain_key((u64)0, xid);
	pid = xhlock_class(xhlock) - lock_classes;
	chain_key = iterate_chain_key(chain_key, pid);

	if (lookup_chain_cache(chain_key))
		return 1;

	if (!add_chain_cache_classes(xid, pid, xhlock->hlock.irq_context,
				chain_key))
		return 0;

	if (!check_prev_add(current, &xlock->hlock, &xhlock->hlock, 1,
			    &xhlock->trace, copy_trace))
		return 0;

	return 1;
}

static void commit_xhlocks(struct cross_lock *xlock)
{
	unsigned int cur = current->xhlock_idx;
5032
	unsigned int prev_hist_id = xhlock(cur).hist_id;
5033 5034 5035 5036 5037
	unsigned int i;

	if (!graph_lock())
		return;

5038 5039 5040
	if (xlock->nr_acquire) {
		for (i = 0; i < MAX_XHLOCKS_NR; i++) {
			struct hist_lock *xhlock = &xhlock(cur - i);
5041

5042 5043
			if (!xhlock_valid(xhlock))
				break;
5044

5045 5046
			if (before(xhlock->hlock.gen_id, xlock->hlock.gen_id))
				break;
5047

5048 5049
			if (!same_context_xhlock(xhlock))
				break;
5050

5051
			/*
5052 5053 5054 5055
			 * Filter out the cases where the ring buffer was
			 * overwritten and the current entry has a bigger
			 * hist_id than the previous one, which is impossible
			 * otherwise:
5056
			 */
5057
			if (unlikely(before(prev_hist_id, xhlock->hist_id)))
5058
				break;
5059

5060
			prev_hist_id = xhlock->hist_id;
5061

5062 5063 5064 5065 5066 5067 5068
			/*
			 * commit_xhlock() returns 0 with graph_lock already
			 * released if fail.
			 */
			if (!commit_xhlock(xlock, xhlock))
				return;
		}
5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111
	}

	graph_unlock();
}

void lock_commit_crosslock(struct lockdep_map *lock)
{
	struct cross_lock *xlock;
	unsigned long flags;

	if (unlikely(!debug_locks || current->lockdep_recursion))
		return;

	if (!current->xhlocks)
		return;

	/*
	 * Do commit hist_locks with the cross_lock, only in case that
	 * the cross_lock could depend on acquisitions after that.
	 *
	 * For example, if the cross_lock does not have the 'check' flag
	 * then we don't need to check dependencies and commit for that.
	 * Just skip it. In that case, of course, the cross_lock does
	 * not depend on acquisitions ahead, either.
	 *
	 * WARNING: Don't do that in add_xlock() in advance. When an
	 * acquisition context is different from the commit context,
	 * invalid(skipped) cross_lock might be accessed.
	 */
	if (!depend_after(&((struct lockdep_map_cross *)lock)->xlock.hlock))
		return;

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
	xlock = &((struct lockdep_map_cross *)lock)->xlock;
	commit_xhlocks(xlock);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_commit_crosslock);

/*
5112 5113
 * Return: 0 - failure;
 *         1 - crosslock, done;
5114 5115 5116 5117
 *         2 - normal lock, continue to held_lock[] ops.
 */
static int lock_release_crosslock(struct lockdep_map *lock)
{
5118 5119 5120 5121 5122 5123 5124 5125
	if (cross_lock(lock)) {
		if (!graph_lock())
			return 0;
		((struct lockdep_map_cross *)lock)->xlock.nr_acquire--;
		graph_unlock();
		return 1;
	}
	return 2;
5126 5127 5128 5129
}

static void cross_init(struct lockdep_map *lock, int cross)
{
5130 5131 5132
	if (cross)
		((struct lockdep_map_cross *)lock)->xlock.nr_acquire = 0;

5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146
	lock->cross = cross;

	/*
	 * Crossrelease assumes that the ring buffer size of xhlocks
	 * is aligned with power of 2. So force it on build.
	 */
	BUILD_BUG_ON(MAX_XHLOCKS_NR & (MAX_XHLOCKS_NR - 1));
}

void lockdep_init_task(struct task_struct *task)
{
	int i;

	task->xhlock_idx = UINT_MAX;
5147
	task->hist_id = 0;
5148

5149
	for (i = 0; i < XHLOCK_CTX_NR; i++) {
5150
		task->xhlock_idx_hist[i] = UINT_MAX;
5151 5152
		task->hist_id_save[i] = 0;
	}
5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167

	task->xhlocks = kzalloc(sizeof(struct hist_lock) * MAX_XHLOCKS_NR,
				GFP_KERNEL);
}

void lockdep_free_task(struct task_struct *task)
{
	if (task->xhlocks) {
		void *tmp = task->xhlocks;
		/* Diable crossrelease for current */
		task->xhlocks = NULL;
		kfree(tmp);
	}
}
#endif