lockdep.c 126.4 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
#ifdef CONFIG_BOOTPARAM_LOCKDEP_CROSSRELEASE_FULLSTACK
static int crossrelease_fullstack = 1;
#else
82
static int crossrelease_fullstack;
83
#endif
84 85 86 87 88 89 90 91
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
92
/*
93 94
 * lockdep_lock: protects the lockdep graph, the hashes and the
 *               class/list/hash allocators.
I
Ingo Molnar 已提交
95 96 97
 *
 * This is one of the rare exceptions where it's justified
 * to use a raw spinlock - we really dont want the spinlock
98
 * code to recurse back into the lockdep code...
I
Ingo Molnar 已提交
99
 */
100
static arch_spinlock_t lockdep_lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
101 102 103

static int graph_lock(void)
{
104
	arch_spin_lock(&lockdep_lock);
105 106 107 108 109 110 111
	/*
	 * 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) {
112
		arch_spin_unlock(&lockdep_lock);
113 114
		return 0;
	}
115 116
	/* prevent any recursions within lockdep from causing deadlocks */
	current->lockdep_recursion++;
117 118 119 120 121
	return 1;
}

static inline int graph_unlock(void)
{
P
Peter Zijlstra 已提交
122 123 124 125 126
	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!
		 */
127
		return DEBUG_LOCKS_WARN_ON(1);
P
Peter Zijlstra 已提交
128
	}
129

130
	current->lockdep_recursion--;
131
	arch_spin_unlock(&lockdep_lock);
132 133 134 135 136 137 138 139 140 141 142
	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();

143
	arch_spin_unlock(&lockdep_lock);
144 145 146

	return ret;
}
I
Ingo Molnar 已提交
147 148

unsigned long nr_list_entries;
P
Peter Zijlstra 已提交
149
static struct lock_list list_entries[MAX_LOCKDEP_ENTRIES];
I
Ingo Molnar 已提交
150 151 152 153 154 155 156 157 158 159

/*
 * 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
160 161 162
static inline struct lock_class *hlock_class(struct held_lock *hlock)
{
	if (!hlock->class_idx) {
P
Peter Zijlstra 已提交
163 164 165
		/*
		 * Someone passed in garbage, we give up.
		 */
D
Dave Jones 已提交
166 167 168 169 170 171
		DEBUG_LOCKS_WARN_ON(1);
		return NULL;
	}
	return lock_classes + hlock->class_idx - 1;
}

P
Peter Zijlstra 已提交
172
#ifdef CONFIG_LOCK_STAT
173
static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
P
Peter Zijlstra 已提交
174

175 176
static inline u64 lockstat_clock(void)
{
177
	return local_clock();
178 179
}

P
Peter Zijlstra 已提交
180
static int lock_point(unsigned long points[], unsigned long ip)
P
Peter Zijlstra 已提交
181 182 183
{
	int i;

P
Peter Zijlstra 已提交
184 185 186
	for (i = 0; i < LOCKSTAT_POINTS; i++) {
		if (points[i] == 0) {
			points[i] = ip;
P
Peter Zijlstra 已提交
187 188
			break;
		}
P
Peter Zijlstra 已提交
189
		if (points[i] == ip)
P
Peter Zijlstra 已提交
190 191 192 193 194 195
			break;
	}

	return i;
}

196
static void lock_time_inc(struct lock_time *lt, u64 time)
P
Peter Zijlstra 已提交
197 198 199 200
{
	if (time > lt->max)
		lt->max = time;

201
	if (time < lt->min || !lt->nr)
P
Peter Zijlstra 已提交
202 203 204 205 206 207
		lt->min = time;

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

208 209
static inline void lock_time_add(struct lock_time *src, struct lock_time *dst)
{
210 211 212 213 214 215 216 217 218
	if (!src->nr)
		return;

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

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

219 220 221 222 223 224 225 226 227 228 229 230
	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 =
231
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
232 233 234 235

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

P
Peter Zijlstra 已提交
236 237 238
		for (i = 0; i < ARRAY_SIZE(stats.contending_point); i++)
			stats.contending_point[i] += pcs->contending_point[i];

239 240 241 242 243
		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
244 245 246

		for (i = 0; i < ARRAY_SIZE(stats.bounces); i++)
			stats.bounces[i] += pcs->bounces[i];
247 248 249 250 251 252 253 254 255 256 257
	}

	return stats;
}

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

	for_each_possible_cpu(cpu) {
		struct lock_class_stats *cpu_stats =
258
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
259 260 261 262

		memset(cpu_stats, 0, sizeof(struct lock_class_stats));
	}
	memset(class->contention_point, 0, sizeof(class->contention_point));
P
Peter Zijlstra 已提交
263
	memset(class->contending_point, 0, sizeof(class->contending_point));
264 265
}

P
Peter Zijlstra 已提交
266 267
static struct lock_class_stats *get_lock_stats(struct lock_class *class)
{
268
	return &get_cpu_var(cpu_lock_stats)[class - lock_classes];
P
Peter Zijlstra 已提交
269 270 271 272
}

static void put_lock_stats(struct lock_class_stats *stats)
{
273
	put_cpu_var(cpu_lock_stats);
P
Peter Zijlstra 已提交
274 275 276 277 278
}

static void lock_release_holdtime(struct held_lock *hlock)
{
	struct lock_class_stats *stats;
279
	u64 holdtime;
P
Peter Zijlstra 已提交
280 281 282 283

	if (!lock_stat)
		return;

284
	holdtime = lockstat_clock() - hlock->holdtime_stamp;
P
Peter Zijlstra 已提交
285

D
Dave Jones 已提交
286
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
287 288 289 290 291 292 293 294 295 296 297 298
	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
299 300 301 302 303 304 305 306 307 308 309 310
/*
 * 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
311
#define __classhashfn(key)	hash_long((unsigned long)key, CLASSHASH_BITS)
I
Ingo Molnar 已提交
312 313
#define classhashentry(key)	(classhash_table + __classhashfn((key)))

314
static struct hlist_head classhash_table[CLASSHASH_SIZE];
I
Ingo Molnar 已提交
315 316 317 318 319 320 321

/*
 * 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
322
#define __chainhashfn(chain)	hash_long(chain, CHAINHASH_BITS)
I
Ingo Molnar 已提交
323 324
#define chainhashentry(chain)	(chainhash_table + __chainhashfn((chain)))

325
static struct hlist_head chainhash_table[CHAINHASH_SIZE];
I
Ingo Molnar 已提交
326 327 328 329 330 331 332

/*
 * 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.
 */
333 334 335 336 337 338 339 340
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
341

342
void lockdep_off(void)
I
Ingo Molnar 已提交
343 344 345 346 347
{
	current->lockdep_recursion++;
}
EXPORT_SYMBOL(lockdep_off);

348
void lockdep_on(void)
I
Ingo Molnar 已提交
349 350 351 352 353 354 355 356 357 358
{
	current->lockdep_recursion--;
}
EXPORT_SYMBOL(lockdep_on);

/*
 * Debugging switches:
 */

#define VERBOSE			0
359
#define VERY_VERBOSE		0
I
Ingo Molnar 已提交
360 361 362 363 364 365 366 367 368

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

369
#if VERBOSE || HARDIRQ_VERBOSE || SOFTIRQ_VERBOSE
I
Ingo Molnar 已提交
370 371 372 373 374
/*
 * Quick filtering for interesting events:
 */
static int class_filter(struct lock_class *class)
{
A
Andi Kleen 已提交
375 376
#if 0
	/* Example */
I
Ingo Molnar 已提交
377
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
378
			!strcmp(class->name, "lockname"))
I
Ingo Molnar 已提交
379 380
		return 1;
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
381
			!strcmp(class->name, "&struct->lockfield"))
I
Ingo Molnar 已提交
382
		return 1;
A
Andi Kleen 已提交
383
#endif
384 385
	/* Filter everything else. 1 would be to allow everything else */
	return 0;
I
Ingo Molnar 已提交
386 387 388 389 390 391 392 393 394 395 396 397 398
}
#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
399
 * addresses. Protected by the graph_lock.
I
Ingo Molnar 已提交
400 401 402 403
 */
unsigned long nr_stack_trace_entries;
static unsigned long stack_trace[MAX_STACK_TRACE_ENTRIES];

404 405 406 407
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");
408
#ifdef CONFIG_LOCK_STAT
409
	printk(KERN_DEBUG "Please attach the output of /proc/lock_stat to the bug report\n");
410
#endif
411 412
}

I
Ingo Molnar 已提交
413 414 415 416 417 418
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;

419 420
	trace->skip = 3;

C
Christoph Hellwig 已提交
421
	save_stack_trace(trace);
I
Ingo Molnar 已提交
422

P
Peter Zijlstra 已提交
423 424 425 426 427 428 429
	/*
	 * 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>
	 */
430 431
	if (trace->nr_entries != 0 &&
	    trace->entries[trace->nr_entries-1] == ULONG_MAX)
P
Peter Zijlstra 已提交
432 433
		trace->nr_entries--;

I
Ingo Molnar 已提交
434 435 436 437
	trace->max_entries = trace->nr_entries;

	nr_stack_trace_entries += trace->nr_entries;

P
Peter Zijlstra 已提交
438
	if (nr_stack_trace_entries >= MAX_STACK_TRACE_ENTRIES-1) {
439 440 441
		if (!debug_locks_off_graph_unlock())
			return 0;

442
		print_lockdep_off("BUG: MAX_STACK_TRACE_ENTRIES too low!");
443 444
		dump_stack();

I
Ingo Molnar 已提交
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
		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:
 */
460
DEFINE_PER_CPU(struct lockdep_stats, lockdep_stats);
I
Ingo Molnar 已提交
461 462 463 464 465 466
#endif

/*
 * Locking printouts:
 */

P
Peter Zijlstra 已提交
467
#define __USAGE(__STATE)						\
P
Peter Zijlstra 已提交
468 469 470 471
	[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
472

I
Ingo Molnar 已提交
473 474
static const char *usage_str[] =
{
P
Peter Zijlstra 已提交
475 476 477 478
#define LOCKDEP_STATE(__STATE) __USAGE(__STATE)
#include "lockdep_states.h"
#undef LOCKDEP_STATE
	[LOCK_USED] = "INITIAL USE",
I
Ingo Molnar 已提交
479 480 481 482
};

const char * __get_key_name(struct lockdep_subclass_key *key, char *str)
{
A
Alexey Dobriyan 已提交
483
	return kallsyms_lookup((unsigned long)key, NULL, NULL, NULL, str);
I
Ingo Molnar 已提交
484 485
}

486
static inline unsigned long lock_flag(enum lock_usage_bit bit)
I
Ingo Molnar 已提交
487
{
488 489
	return 1UL << bit;
}
I
Ingo Molnar 已提交
490

491 492 493 494 495 496 497 498 499 500
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
501 502
	}

503 504
	return c;
}
505

P
Peter Zijlstra 已提交
506
void get_usage_chars(struct lock_class *class, char usage[LOCK_USAGE_CHARS])
507
{
P
Peter Zijlstra 已提交
508
	int i = 0;
509

P
Peter Zijlstra 已提交
510 511 512 513 514 515 516
#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
517 518
}

519
static void __print_lock_name(struct lock_class *class)
520 521 522 523
{
	char str[KSYM_NAME_LEN];
	const char *name;

I
Ingo Molnar 已提交
524 525 526
	name = class->name;
	if (!name) {
		name = __get_key_name(class->key, str);
D
Dmitry Vyukov 已提交
527
		printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
528
	} else {
D
Dmitry Vyukov 已提交
529
		printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
530
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
531
			printk(KERN_CONT "#%d", class->name_version);
I
Ingo Molnar 已提交
532
		if (class->subclass)
D
Dmitry Vyukov 已提交
533
			printk(KERN_CONT "/%d", class->subclass);
I
Ingo Molnar 已提交
534
	}
535 536 537 538 539 540 541 542
}

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
543
	printk(KERN_CONT " (");
544
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
545
	printk(KERN_CONT "){%s}", usage);
I
Ingo Molnar 已提交
546 547 548 549 550
}

static void print_lockdep_cache(struct lockdep_map *lock)
{
	const char *name;
551
	char str[KSYM_NAME_LEN];
I
Ingo Molnar 已提交
552 553 554 555 556

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

D
Dmitry Vyukov 已提交
557
	printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
558 559 560 561
}

static void print_lock(struct held_lock *hlock)
{
562 563 564 565 566 567 568 569 570 571
	/*
	 * 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
572
		printk(KERN_CONT "<RELEASED>\n");
573 574 575 576
		return;
	}

	print_lock_name(lock_classes + class_idx - 1);
D
Dmitry Vyukov 已提交
577 578
	printk(KERN_CONT ", at: [<%p>] %pS\n",
		(void *)hlock->acquire_ip, (void *)hlock->acquire_ip);
I
Ingo Molnar 已提交
579 580 581 582 583 584 585
}

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

	if (!depth) {
586
		printk("no locks held by %s/%d.\n", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
587 588 589
		return;
	}
	printk("%d lock%s held by %s/%d:\n",
590
		depth, depth > 1 ? "s" : "", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
591 592 593 594 595 596 597

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

598
static void print_kernel_ident(void)
P
Peter Zijlstra 已提交
599
{
600
	printk("%s %.*s %s\n", init_utsname()->release,
P
Peter Zijlstra 已提交
601
		(int)strcspn(init_utsname()->version, " "),
602 603
		init_utsname()->version,
		print_tainted());
P
Peter Zijlstra 已提交
604 605 606 607 608 609 610 611 612 613
}

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

I
Ingo Molnar 已提交
614
/*
P
Peter Zijlstra 已提交
615
 * Is this the address of a static object:
I
Ingo Molnar 已提交
616
 */
617
#ifdef __KERNEL__
P
Peter Zijlstra 已提交
618
static int static_obj(void *obj)
I
Ingo Molnar 已提交
619
{
P
Peter Zijlstra 已提交
620 621 622 623
	unsigned long start = (unsigned long) &_stext,
		      end   = (unsigned long) &_end,
		      addr  = (unsigned long) obj;

I
Ingo Molnar 已提交
624
	/*
P
Peter Zijlstra 已提交
625
	 * static variable?
I
Ingo Molnar 已提交
626
	 */
P
Peter Zijlstra 已提交
627 628
	if ((addr >= start) && (addr < end))
		return 1;
I
Ingo Molnar 已提交
629

630 631 632
	if (arch_is_kernel_data(addr))
		return 1;

I
Ingo Molnar 已提交
633
	/*
634
	 * in-kernel percpu var?
I
Ingo Molnar 已提交
635
	 */
636 637
	if (is_kernel_percpu_address(addr))
		return 1;
I
Ingo Molnar 已提交
638

P
Peter Zijlstra 已提交
639
	/*
640
	 * module static or percpu var?
P
Peter Zijlstra 已提交
641
	 */
642
	return is_module_address(addr) || is_module_percpu_address(addr);
643
}
644
#endif
645

I
Ingo Molnar 已提交
646
/*
P
Peter Zijlstra 已提交
647 648
 * To make lock name printouts unique, we calculate a unique
 * class->name_version generation counter:
I
Ingo Molnar 已提交
649
 */
P
Peter Zijlstra 已提交
650
static int count_matching_names(struct lock_class *new_class)
I
Ingo Molnar 已提交
651
{
P
Peter Zijlstra 已提交
652 653
	struct lock_class *class;
	int count = 0;
I
Ingo Molnar 已提交
654

P
Peter Zijlstra 已提交
655
	if (!new_class->name)
I
Ingo Molnar 已提交
656 657
		return 0;

658
	list_for_each_entry_rcu(class, &all_lock_classes, lock_entry) {
P
Peter Zijlstra 已提交
659 660 661 662 663
		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
664

P
Peter Zijlstra 已提交
665
	return count + 1;
I
Ingo Molnar 已提交
666 667
}

P
Peter Zijlstra 已提交
668 669 670 671 672 673 674
/*
 * 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
675
{
P
Peter Zijlstra 已提交
676
	struct lockdep_subclass_key *key;
677
	struct hlist_head *hash_head;
P
Peter Zijlstra 已提交
678
	struct lock_class *class;
679
	bool is_static = false;
I
Ingo Molnar 已提交
680

681 682 683 684 685 686 687 688 689 690
	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
691 692
	/*
	 * Static locks do not have their class-keys yet - for them the key
693 694 695
	 * 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
696
	 */
697 698 699 700 701 702 703 704 705 706 707 708 709
	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
710

P
Peter Zijlstra 已提交
711 712 713 714 715 716
	/*
	 * 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
717 718
	BUILD_BUG_ON(sizeof(struct lock_class_key) >
			sizeof(struct lockdep_map));
I
Ingo Molnar 已提交
719

P
Peter Zijlstra 已提交
720
	key = lock->key->subkeys + subclass;
721

P
Peter Zijlstra 已提交
722
	hash_head = classhashentry(key);
723

P
Peter Zijlstra 已提交
724
	/*
725
	 * We do an RCU walk of the hash, see lockdep_free_key_range().
P
Peter Zijlstra 已提交
726
	 */
727 728 729
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return NULL;

730
	hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
P
Peter Zijlstra 已提交
731
		if (class->key == key) {
P
Peter Zijlstra 已提交
732 733 734 735
			/*
			 * Huh! same key, different name? Did someone trample
			 * on some memory? We're most confused.
			 */
P
Peter Zijlstra 已提交
736
			WARN_ON_ONCE(class->name != lock->name);
P
Peter Zijlstra 已提交
737
			return class;
P
Peter Zijlstra 已提交
738 739
		}
	}
I
Ingo Molnar 已提交
740

741
	return is_static || static_obj(lock->key) ? NULL : ERR_PTR(-EINVAL);
I
Ingo Molnar 已提交
742 743
}

744 745 746 747 748 749 750 751 752 753 754 755
#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
756
/*
P
Peter Zijlstra 已提交
757 758 759
 * 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
760
 */
761
static struct lock_class *
P
Peter Zijlstra 已提交
762
register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
I
Ingo Molnar 已提交
763
{
P
Peter Zijlstra 已提交
764
	struct lockdep_subclass_key *key;
765
	struct hlist_head *hash_head;
P
Peter Zijlstra 已提交
766
	struct lock_class *class;
767 768

	DEBUG_LOCKS_WARN_ON(!irqs_disabled());
P
Peter Zijlstra 已提交
769 770

	class = look_up_lock_class(lock, subclass);
771
	if (likely(!IS_ERR_OR_NULL(class)))
772
		goto out_set_class_cache;
P
Peter Zijlstra 已提交
773 774 775

	/*
	 * Debug-check: all keys must be persistent!
776 777
	 */
	if (IS_ERR(class)) {
P
Peter Zijlstra 已提交
778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795
		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:
	 */
796
	hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
P
Peter Zijlstra 已提交
797 798
		if (class->key == key)
			goto out_unlock_set;
799 800
	}

P
Peter Zijlstra 已提交
801 802 803 804 805 806 807 808 809
	/*
	 * 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;
		}

810
		print_lockdep_off("BUG: MAX_LOCKDEP_KEYS too low!");
811
		dump_stack();
P
Peter Zijlstra 已提交
812 813 814
		return NULL;
	}
	class = lock_classes + nr_lock_classes++;
815
	debug_atomic_inc(nr_unused_locks);
P
Peter Zijlstra 已提交
816 817 818 819 820 821 822 823 824 825 826
	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:
	 */
827
	hlist_add_head_rcu(&class->hash_entry, hash_head);
828 829 830 831
	/*
	 * Add it to the global list of classes:
	 */
	list_add_tail_rcu(&class->lock_entry, &all_lock_classes);
P
Peter Zijlstra 已提交
832 833 834 835 836 837

	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
838 839
			printk(KERN_CONT "#%d", class->name_version);
		printk(KERN_CONT "\n");
P
Peter Zijlstra 已提交
840 841 842 843 844 845 846 847 848
		dump_stack();

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

849
out_set_class_cache:
P
Peter Zijlstra 已提交
850
	if (!subclass || force)
851 852 853
		lock->class_cache[0] = class;
	else if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		lock->class_cache[subclass] = class;
P
Peter Zijlstra 已提交
854

P
Peter Zijlstra 已提交
855 856 857 858
	/*
	 * 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
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
	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;

876
		print_lockdep_off("BUG: MAX_LOCKDEP_ENTRIES too low!");
877
		dump_stack();
P
Peter Zijlstra 已提交
878 879 880 881 882 883 884 885
		return NULL;
	}
	return list_entries + nr_list_entries++;
}

/*
 * Add a new dependency to the head of the list:
 */
886 887 888
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
889 890 891 892 893 894 895 896 897 898
{
	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;

899 900
	entry->class = this;
	entry->distance = distance;
Y
Yong Zhang 已提交
901
	entry->trace = *trace;
P
Peter Zijlstra 已提交
902
	/*
903 904 905
	 * 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
906 907 908 909 910 911
	 */
	list_add_tail_rcu(&entry->entry, head);

	return 1;
}

P
Peter Zijlstra 已提交
912 913 914
/*
 * For good efficiency of modular, we use power of 2
 */
P
Peter Zijlstra 已提交
915 916 917
#define MAX_CIRCULAR_QUEUE_SIZE		4096UL
#define CQ_MASK				(MAX_CIRCULAR_QUEUE_SIZE-1)

P
Peter Zijlstra 已提交
918 919
/*
 * The circular_queue and helpers is used to implement the
P
Peter Zijlstra 已提交
920 921 922
 * 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
923
 */
P
Peter Zijlstra 已提交
924 925 926 927 928 929 930
struct circular_queue {
	unsigned long element[MAX_CIRCULAR_QUEUE_SIZE];
	unsigned int  front, rear;
};

static struct circular_queue lock_cq;

931
unsigned int max_bfs_queue_depth;
P
Peter Zijlstra 已提交
932

933 934
static unsigned int lockdep_dependency_gen_id;

P
Peter Zijlstra 已提交
935 936 937
static inline void __cq_init(struct circular_queue *cq)
{
	cq->front = cq->rear = 0;
938
	lockdep_dependency_gen_id++;
P
Peter Zijlstra 已提交
939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
}

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
980

P
Peter Zijlstra 已提交
981
	nr = lock - list_entries;
P
Peter Zijlstra 已提交
982
	WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
P
Peter Zijlstra 已提交
983
	lock->parent = parent;
984
	lock->class->dep_gen_id = lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
985 986 987 988 989
}

static inline unsigned long lock_accessed(struct lock_list *lock)
{
	unsigned long nr;
P
Peter Zijlstra 已提交
990

P
Peter Zijlstra 已提交
991
	nr = lock - list_entries;
P
Peter Zijlstra 已提交
992
	WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
993
	return lock->class->dep_gen_id == lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
}

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;
}

1013
static int __bfs(struct lock_list *source_entry,
P
Peter Zijlstra 已提交
1014 1015 1016 1017
		 void *data,
		 int (*match)(struct lock_list *entry, void *data),
		 struct lock_list **target_entry,
		 int forward)
1018 1019
{
	struct lock_list *entry;
1020
	struct list_head *head;
1021 1022 1023
	struct circular_queue *cq = &lock_cq;
	int ret = 1;

1024
	if (match(source_entry, data)) {
1025 1026 1027 1028 1029
		*target_entry = source_entry;
		ret = 0;
		goto exit;
	}

1030 1031 1032 1033 1034 1035 1036 1037 1038
	if (forward)
		head = &source_entry->class->locks_after;
	else
		head = &source_entry->class->locks_before;

	if (list_empty(head))
		goto exit;

	__cq_init(cq);
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
	__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;

1056 1057 1058
		DEBUG_LOCKS_WARN_ON(!irqs_disabled());

		list_for_each_entry_rcu(entry, head, entry) {
1059
			if (!lock_accessed(entry)) {
1060
				unsigned int cq_depth;
1061
				mark_lock_accessed(entry, lock);
1062
				if (match(entry, data)) {
1063 1064 1065 1066 1067 1068 1069 1070 1071
					*target_entry = entry;
					ret = 0;
					goto exit;
				}

				if (__cq_enqueue(cq, (unsigned long)entry)) {
					ret = -1;
					goto exit;
				}
1072 1073 1074
				cq_depth = __cq_get_elem_count(cq);
				if (max_bfs_queue_depth < cq_depth)
					max_bfs_queue_depth = cq_depth;
1075 1076 1077 1078 1079 1080 1081
			}
		}
	}
exit:
	return ret;
}

1082
static inline int __bfs_forwards(struct lock_list *src_entry,
1083 1084 1085
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1086
{
1087
	return __bfs(src_entry, data, match, target_entry, 1);
1088 1089 1090

}

1091
static inline int __bfs_backwards(struct lock_list *src_entry,
1092 1093 1094
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1095
{
1096
	return __bfs(src_entry, data, match, target_entry, 0);
1097 1098 1099

}

P
Peter Zijlstra 已提交
1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
/*
 * 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
1111
print_circular_bug_entry(struct lock_list *target, int depth)
P
Peter Zijlstra 已提交
1112 1113 1114 1115 1116
{
	if (debug_locks_silent)
		return 0;
	printk("\n-> #%u", depth);
	print_lock_name(target->class);
D
Dmitry Vyukov 已提交
1117
	printk(KERN_CONT ":\n");
P
Peter Zijlstra 已提交
1118 1119 1120 1121 1122
	print_stack_trace(&target->trace, 6);

	return 0;
}

1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
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
1148
		printk(KERN_CONT " --> ");
1149
		__print_lock_name(parent);
D
Dmitry Vyukov 已提交
1150
		printk(KERN_CONT " --> ");
1151
		__print_lock_name(target);
D
Dmitry Vyukov 已提交
1152
		printk(KERN_CONT "\n\n");
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 1186 1187 1188 1189
	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");
	}
1190 1191
}

P
Peter Zijlstra 已提交
1192 1193 1194 1195 1196
/*
 * When a circular dependency is detected, print the
 * header first:
 */
static noinline int
1197 1198 1199
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
1200 1201 1202
{
	struct task_struct *curr = current;

1203
	if (debug_locks_silent)
P
Peter Zijlstra 已提交
1204 1205
		return 0;

1206
	pr_warn("\n");
1207 1208
	pr_warn("======================================================\n");
	pr_warn("WARNING: possible circular locking dependency detected\n");
1209
	print_kernel_ident();
1210
	pr_warn("------------------------------------------------------\n");
1211
	pr_warn("%s/%d is trying to acquire lock:\n",
1212
		curr->comm, task_pid_nr(curr));
1213
	print_lock(check_src);
1214 1215 1216 1217 1218 1219

	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");

1220
	print_lock(check_tgt);
1221 1222
	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
1223 1224 1225 1226 1227 1228

	print_circular_bug_entry(entry, depth);

	return 0;
}

1229 1230 1231 1232 1233
static inline int class_equal(struct lock_list *entry, void *data)
{
	return entry->class == data;
}

1234 1235 1236
static noinline int print_circular_bug(struct lock_list *this,
				struct lock_list *target,
				struct held_lock *check_src,
1237 1238
				struct held_lock *check_tgt,
				struct stack_trace *trace)
P
Peter Zijlstra 已提交
1239 1240
{
	struct task_struct *curr = current;
1241
	struct lock_list *parent;
1242
	struct lock_list *first_parent;
1243
	int depth;
P
Peter Zijlstra 已提交
1244

1245
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
P
Peter Zijlstra 已提交
1246 1247
		return 0;

1248 1249 1250
	if (cross_lock(check_tgt->instance))
		this->trace = *trace;
	else if (!save_trace(&this->trace))
P
Peter Zijlstra 已提交
1251 1252
		return 0;

1253 1254
	depth = get_lock_depth(target);

1255
	print_circular_bug_header(target, depth, check_src, check_tgt);
1256 1257

	parent = get_lock_parent(target);
1258
	first_parent = parent;
1259 1260 1261 1262 1263

	while (parent) {
		print_circular_bug_entry(parent, --depth);
		parent = get_lock_parent(parent);
	}
P
Peter Zijlstra 已提交
1264 1265

	printk("\nother info that might help us debug this:\n\n");
1266 1267 1268
	print_circular_lock_scenario(check_src, check_tgt,
				     first_parent);

P
Peter Zijlstra 已提交
1269 1270 1271 1272 1273 1274 1275 1276
	lockdep_print_held_locks(curr);

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

	return 0;
}

1277 1278 1279 1280 1281
static noinline int print_bfs_bug(int ret)
{
	if (!debug_locks_off_graph_unlock())
		return 0;

P
Peter Zijlstra 已提交
1282 1283 1284
	/*
	 * Breadth-first-search failed, graph got corrupted?
	 */
1285 1286 1287 1288 1289
	WARN(1, "lockdep bfs error:%d\n", ret);

	return 0;
}

1290
static int noop_count(struct lock_list *entry, void *data)
1291
{
1292 1293 1294
	(*(unsigned long *)data)++;
	return 0;
}
1295

1296
static unsigned long __lockdep_count_forward_deps(struct lock_list *this)
1297 1298 1299
{
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1300

1301
	__bfs_forwards(this, (void *)&count, noop_count, &target_entry);
1302

1303
	return count;
1304 1305 1306 1307
}
unsigned long lockdep_count_forward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1308 1309 1310 1311
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1312 1313

	local_irq_save(flags);
1314
	arch_spin_lock(&lockdep_lock);
1315
	ret = __lockdep_count_forward_deps(&this);
1316
	arch_spin_unlock(&lockdep_lock);
1317 1318 1319 1320 1321
	local_irq_restore(flags);

	return ret;
}

1322
static unsigned long __lockdep_count_backward_deps(struct lock_list *this)
1323
{
1324 1325
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1326

1327
	__bfs_backwards(this, (void *)&count, noop_count, &target_entry);
1328

1329
	return count;
1330 1331 1332 1333 1334
}

unsigned long lockdep_count_backward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1335 1336 1337 1338
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1339 1340

	local_irq_save(flags);
1341
	arch_spin_lock(&lockdep_lock);
1342
	ret = __lockdep_count_backward_deps(&this);
1343
	arch_spin_unlock(&lockdep_lock);
1344 1345 1346 1347 1348
	local_irq_restore(flags);

	return ret;
}

P
Peter Zijlstra 已提交
1349 1350 1351 1352 1353
/*
 * 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
1354 1355
check_noncircular(struct lock_list *root, struct lock_class *target,
		struct lock_list **target_entry)
P
Peter Zijlstra 已提交
1356
{
1357
	int result;
P
Peter Zijlstra 已提交
1358

1359
	debug_atomic_inc(nr_cyclic_checks);
1360

1361
	result = __bfs_forwards(root, target, class_equal, target_entry);
I
Ingo Molnar 已提交
1362

1363 1364
	return result;
}
1365

1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
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;
}

1379
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
1380 1381 1382 1383 1384 1385
/*
 * 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.
 */

1386 1387 1388 1389 1390 1391 1392
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
1393 1394
/*
 * Find a node in the forwards-direction dependency sub-graph starting
1395
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1396
 *
1397 1398
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1399
 *
1400 1401
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1402
 */
1403 1404 1405
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
1406
{
1407
	int result;
I
Ingo Molnar 已提交
1408

1409
	debug_atomic_inc(nr_find_usage_forwards_checks);
I
Ingo Molnar 已提交
1410

1411 1412 1413
	result = __bfs_forwards(root, (void *)bit, usage_match, target_entry);

	return result;
I
Ingo Molnar 已提交
1414 1415 1416 1417
}

/*
 * Find a node in the backwards-direction dependency sub-graph starting
1418
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1419
 *
1420 1421
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1422
 *
1423 1424
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1425
 */
1426 1427 1428
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
1429
{
1430
	int result;
I
Ingo Molnar 已提交
1431

1432
	debug_atomic_inc(nr_find_usage_backwards_checks);
I
Ingo Molnar 已提交
1433

1434
	result = __bfs_backwards(root, (void *)bit, usage_match, target_entry);
D
Dave Jones 已提交
1435

1436
	return result;
I
Ingo Molnar 已提交
1437 1438
}

P
Peter Zijlstra 已提交
1439 1440 1441 1442 1443 1444
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
1445 1446
	printk(KERN_CONT " ops: %lu", class->ops);
	printk(KERN_CONT " {\n");
P
Peter Zijlstra 已提交
1447 1448 1449 1450 1451 1452

	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
1453
			len += printk(KERN_CONT " at:\n");
P
Peter Zijlstra 已提交
1454 1455 1456 1457 1458
			print_stack_trace(class->usage_traces + bit, len);
		}
	}
	printk("%*s }\n", depth, "");

D
Dmitry Vyukov 已提交
1459 1460
	printk("%*s ... key      at: [<%p>] %pS\n",
		depth, "", class->key, class->key);
P
Peter Zijlstra 已提交
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482
}

/*
 * 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)) {
1483
			printk("lockdep:%s bad path found in chain graph\n", __func__);
P
Peter Zijlstra 已提交
1484 1485 1486 1487 1488 1489 1490 1491 1492
			break;
		}

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

	return;
}
1493

1494 1495 1496
static void
print_irq_lock_scenario(struct lock_list *safe_entry,
			struct lock_list *unsafe_entry,
1497 1498
			struct lock_class *prev_class,
			struct lock_class *next_class)
1499 1500 1501
{
	struct lock_class *safe_class = safe_entry->class;
	struct lock_class *unsafe_class = unsafe_entry->class;
1502
	struct lock_class *middle_class = prev_class;
1503 1504

	if (middle_class == safe_class)
1505
		middle_class = next_class;
1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522

	/*
	 * 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
1523
		printk(KERN_CONT " --> ");
1524
		__print_lock_name(middle_class);
D
Dmitry Vyukov 已提交
1525
		printk(KERN_CONT " --> ");
1526
		__print_lock_name(unsafe_class);
D
Dmitry Vyukov 已提交
1527
		printk(KERN_CONT "\n\n");
1528 1529 1530 1531 1532 1533 1534
	}

	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
1535
	printk(KERN_CONT ");\n");
1536 1537 1538
	printk("                               local_irq_disable();\n");
	printk("                               lock(");
	__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1539
	printk(KERN_CONT ");\n");
1540 1541
	printk("                               lock(");
	__print_lock_name(middle_class);
D
Dmitry Vyukov 已提交
1542
	printk(KERN_CONT ");\n");
1543 1544 1545
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1546
	printk(KERN_CONT ");\n");
1547 1548 1549
	printk("\n *** DEADLOCK ***\n\n");
}

I
Ingo Molnar 已提交
1550 1551
static int
print_bad_irq_dependency(struct task_struct *curr,
1552 1553 1554 1555
			 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
1556 1557 1558 1559 1560 1561
			 struct held_lock *prev,
			 struct held_lock *next,
			 enum lock_usage_bit bit1,
			 enum lock_usage_bit bit2,
			 const char *irqclass)
{
1562
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1563 1564
		return 0;

1565
	pr_warn("\n");
1566 1567
	pr_warn("=====================================================\n");
	pr_warn("WARNING: %s-safe -> %s-unsafe lock order detected\n",
I
Ingo Molnar 已提交
1568
		irqclass, irqclass);
1569
	print_kernel_ident();
1570
	pr_warn("-----------------------------------------------------\n");
1571
	pr_warn("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] is trying to acquire:\n",
1572
		curr->comm, task_pid_nr(curr),
I
Ingo Molnar 已提交
1573 1574 1575 1576 1577 1578
		curr->hardirq_context, hardirq_count() >> HARDIRQ_SHIFT,
		curr->softirq_context, softirq_count() >> SOFTIRQ_SHIFT,
		curr->hardirqs_enabled,
		curr->softirqs_enabled);
	print_lock(next);

1579
	pr_warn("\nand this task is already holding:\n");
I
Ingo Molnar 已提交
1580
	print_lock(prev);
1581
	pr_warn("which would create a new lock dependency:\n");
D
Dave Jones 已提交
1582
	print_lock_name(hlock_class(prev));
1583
	pr_cont(" ->");
D
Dave Jones 已提交
1584
	print_lock_name(hlock_class(next));
1585
	pr_cont("\n");
I
Ingo Molnar 已提交
1586

1587
	pr_warn("\nbut this new dependency connects a %s-irq-safe lock:\n",
I
Ingo Molnar 已提交
1588
		irqclass);
1589
	print_lock_name(backwards_entry->class);
1590
	pr_warn("\n... which became %s-irq-safe at:\n", irqclass);
I
Ingo Molnar 已提交
1591

1592
	print_stack_trace(backwards_entry->class->usage_traces + bit1, 1);
I
Ingo Molnar 已提交
1593

1594
	pr_warn("\nto a %s-irq-unsafe lock:\n", irqclass);
1595
	print_lock_name(forwards_entry->class);
1596 1597
	pr_warn("\n... which became %s-irq-unsafe at:\n", irqclass);
	pr_warn("...");
I
Ingo Molnar 已提交
1598

1599
	print_stack_trace(forwards_entry->class->usage_traces + bit2, 1);
I
Ingo Molnar 已提交
1600

1601
	pr_warn("\nother info that might help us debug this:\n\n");
1602 1603
	print_irq_lock_scenario(backwards_entry, forwards_entry,
				hlock_class(prev), hlock_class(next));
1604

I
Ingo Molnar 已提交
1605 1606
	lockdep_print_held_locks(curr);

1607
	pr_warn("\nthe dependencies between %s-irq-safe lock and the holding lock:\n", irqclass);
1608 1609 1610
	if (!save_trace(&prev_root->trace))
		return 0;
	print_shortest_lock_dependencies(backwards_entry, prev_root);
I
Ingo Molnar 已提交
1611

1612 1613
	pr_warn("\nthe dependencies between the lock to be acquired");
	pr_warn(" and %s-irq-unsafe lock:\n", irqclass);
1614 1615 1616
	if (!save_trace(&next_root->trace))
		return 0;
	print_shortest_lock_dependencies(forwards_entry, next_root);
I
Ingo Molnar 已提交
1617

1618
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629
	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;
1630
	struct lock_list this, that;
1631
	struct lock_list *uninitialized_var(target_entry);
1632
	struct lock_list *uninitialized_var(target_entry1);
1633 1634 1635 1636 1637

	this.parent = NULL;

	this.class = hlock_class(prev);
	ret = find_usage_backwards(&this, bit_backwards, &target_entry);
P
Peter Zijlstra 已提交
1638 1639 1640 1641
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
1642

1643 1644 1645
	that.parent = NULL;
	that.class = hlock_class(next);
	ret = find_usage_forwards(&that, bit_forwards, &target_entry1);
P
Peter Zijlstra 已提交
1646 1647 1648 1649
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
1650

1651 1652 1653
	return print_bad_irq_dependency(curr, &this, &that,
			target_entry, target_entry1,
			prev, next,
I
Ingo Molnar 已提交
1654 1655 1656
			bit_backwards, bit_forwards, irqclass);
}

1657 1658
static const char *state_names[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1659
	__stringify(__STATE),
1660 1661 1662 1663 1664 1665
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static const char *state_rnames[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1666
	__stringify(__STATE)"-READ",
1667 1668 1669 1670 1671
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static inline const char *state_name(enum lock_usage_bit bit)
P
Peter Zijlstra 已提交
1672
{
1673 1674
	return (bit & 1) ? state_rnames[bit >> 2] : state_names[bit >> 2];
}
P
Peter Zijlstra 已提交
1675

1676 1677
static int exclusive_bit(int new_bit)
{
P
Peter Zijlstra 已提交
1678
	/*
1679 1680 1681 1682 1683 1684 1685 1686
	 * 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
1687
	 */
1688 1689 1690

	int state = new_bit & ~3;
	int dir = new_bit & 2;
P
Peter Zijlstra 已提交
1691 1692

	/*
1693
	 * keep state, bit flip the direction and strip read.
P
Peter Zijlstra 已提交
1694
	 */
1695 1696 1697 1698 1699 1700
	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
1701
	/*
1702 1703
	 * 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
1704 1705 1706
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1707 1708
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
P
Peter Zijlstra 已提交
1709 1710
		return 0;

1711 1712
	bit++; /* _READ */

1713
	/*
1714 1715
	 * Prove that the new dependency does not connect a hardirq-safe-read
	 * lock with a hardirq-unsafe lock - to achieve this we search
1716 1717 1718
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1719 1720
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
1721 1722
		return 0;

1723 1724 1725 1726 1727 1728 1729 1730 1731
	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))	\
1732
		return 0;
1733 1734
#include "lockdep_states.h"
#undef LOCKDEP_STATE
1735

P
Peter Zijlstra 已提交
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 1761 1762 1763 1764
	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
1765 1766
#endif

1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
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
1779
	printk(KERN_CONT ");\n");
1780 1781
	printk("  lock(");
	__print_lock_name(next);
D
Dmitry Vyukov 已提交
1782
	printk(KERN_CONT ");\n");
1783 1784 1785 1786
	printk("\n *** DEADLOCK ***\n\n");
	printk(" May be due to missing lock nesting notation\n\n");
}

I
Ingo Molnar 已提交
1787 1788 1789 1790
static int
print_deadlock_bug(struct task_struct *curr, struct held_lock *prev,
		   struct held_lock *next)
{
1791
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1792 1793
		return 0;

1794
	pr_warn("\n");
1795 1796
	pr_warn("============================================\n");
	pr_warn("WARNING: possible recursive locking detected\n");
1797
	print_kernel_ident();
1798
	pr_warn("--------------------------------------------\n");
1799
	pr_warn("%s/%d is trying to acquire lock:\n",
1800
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
1801
	print_lock(next);
1802
	pr_warn("\nbut task is already holding lock:\n");
I
Ingo Molnar 已提交
1803 1804
	print_lock(prev);

1805
	pr_warn("\nother info that might help us debug this:\n");
1806
	print_deadlock_scenario(next, prev);
I
Ingo Molnar 已提交
1807 1808
	lockdep_print_held_locks(curr);

1809
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827
	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
1828
	struct held_lock *nest = NULL;
I
Ingo Molnar 已提交
1829 1830 1831 1832
	int i;

	for (i = 0; i < curr->lockdep_depth; i++) {
		prev = curr->held_locks + i;
P
Peter Zijlstra 已提交
1833 1834 1835 1836

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

D
Dave Jones 已提交
1837
		if (hlock_class(prev) != hlock_class(next))
I
Ingo Molnar 已提交
1838
			continue;
P
Peter Zijlstra 已提交
1839

I
Ingo Molnar 已提交
1840 1841
		/*
		 * Allow read-after-read recursion of the same
1842
		 * lock class (i.e. read_lock(lock)+read_lock(lock)):
I
Ingo Molnar 已提交
1843
		 */
1844
		if ((read == 2) && prev->read)
I
Ingo Molnar 已提交
1845
			return 2;
P
Peter Zijlstra 已提交
1846 1847 1848 1849 1850 1851 1852 1853

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

1854 1855 1856
		if (cross_lock(prev->instance))
			continue;

I
Ingo Molnar 已提交
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 1882 1883 1884 1885
		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,
1886 1887
	       struct held_lock *next, int distance, struct stack_trace *trace,
	       int (*save)(struct stack_trace *trace))
I
Ingo Molnar 已提交
1888
{
1889
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
1890
	struct lock_list *entry;
1891
	struct lock_list this;
1892
	int ret;
I
Ingo Molnar 已提交
1893 1894 1895 1896 1897 1898 1899 1900 1901 1902

	/*
	 * 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:
	 */
1903 1904 1905
	this.class = hlock_class(next);
	this.parent = NULL;
	ret = check_noncircular(&this, hlock_class(prev), &target_entry);
1906 1907 1908 1909 1910 1911 1912 1913 1914
	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);
		}
1915
		return print_circular_bug(&this, target_entry, next, prev, trace);
1916
	}
1917 1918
	else if (unlikely(ret < 0))
		return print_bfs_bug(ret);
1919

P
Peter Zijlstra 已提交
1920
	if (!check_prev_add_irq(curr, prev, next))
I
Ingo Molnar 已提交
1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940
		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
1941 1942
	list_for_each_entry(entry, &hlock_class(prev)->locks_after, entry) {
		if (entry->class == hlock_class(next)) {
1943 1944
			if (distance == 1)
				entry->distance = 1;
1945
			return 1;
1946
		}
I
Ingo Molnar 已提交
1947 1948
	}

1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962
	/*
	 * 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);


1963
	if (!trace->entries && !save(trace))
1964
		return 0;
Y
Yong Zhang 已提交
1965

I
Ingo Molnar 已提交
1966 1967 1968 1969
	/*
	 * Ok, all validations passed, add the new lock
	 * to the previous lock's dependency list:
	 */
1970
	ret = add_lock_to_list(hlock_class(next),
D
Dave Jones 已提交
1971
			       &hlock_class(prev)->locks_after,
1972
			       next->acquire_ip, distance, trace);
1973

I
Ingo Molnar 已提交
1974 1975
	if (!ret)
		return 0;
1976

1977
	ret = add_lock_to_list(hlock_class(prev),
D
Dave Jones 已提交
1978
			       &hlock_class(next)->locks_before,
1979
			       next->acquire_ip, distance, trace);
1980 1981
	if (!ret)
		return 0;
I
Ingo Molnar 已提交
1982

1983
	return 2;
P
Peter Zijlstra 已提交
1984
}
I
Ingo Molnar 已提交
1985

P
Peter Zijlstra 已提交
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
/*
 * 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;
1997 1998 1999 2000 2001 2002
	struct stack_trace trace = {
		.nr_entries = 0,
		.max_entries = 0,
		.entries = NULL,
		.skip = 0,
	};
2003

I
Ingo Molnar 已提交
2004
	/*
P
Peter Zijlstra 已提交
2005 2006 2007
	 * Debugging checks.
	 *
	 * Depth must not be zero for a non-head lock:
I
Ingo Molnar 已提交
2008
	 */
P
Peter Zijlstra 已提交
2009 2010
	if (!depth)
		goto out_bug;
I
Ingo Molnar 已提交
2011
	/*
P
Peter Zijlstra 已提交
2012 2013
	 * At least two relevant locks must exist for this
	 * to be a head:
I
Ingo Molnar 已提交
2014
	 */
P
Peter Zijlstra 已提交
2015 2016 2017
	if (curr->held_locks[depth].irq_context !=
			curr->held_locks[depth-1].irq_context)
		goto out_bug;
2018

P
Peter Zijlstra 已提交
2019 2020
	for (;;) {
		int distance = curr->lockdep_depth - depth + 1;
2021
		hlock = curr->held_locks + depth - 1;
P
Peter Zijlstra 已提交
2022
		/*
2023 2024
		 * Only non-crosslock entries get new dependencies added.
		 * Crosslock entries will be added by commit later:
P
Peter Zijlstra 已提交
2025
		 */
2026
		if (!cross_lock(hlock->instance)) {
2027
			/*
2028 2029
			 * Only non-recursive-read entries get new dependencies
			 * added:
2030
			 */
2031 2032
			if (hlock->read != 2 && hlock->check) {
				int ret = check_prev_add(curr, hlock, next,
2033
							 distance, &trace, save_trace);
2034 2035
				if (!ret)
					return 0;
2036

2037 2038 2039 2040 2041 2042 2043 2044 2045
				/*
				 * 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;
			}
2046
		}
P
Peter Zijlstra 已提交
2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058
		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
2059
	}
P
Peter Zijlstra 已提交
2060 2061 2062 2063
	return 1;
out_bug:
	if (!debug_locks_off_graph_unlock())
		return 0;
I
Ingo Molnar 已提交
2064

P
Peter Zijlstra 已提交
2065 2066 2067 2068 2069
	/*
	 * 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
2070
	WARN_ON(1);
I
Ingo Molnar 已提交
2071

P
Peter Zijlstra 已提交
2072
	return 0;
I
Ingo Molnar 已提交
2073 2074
}

P
Peter Zijlstra 已提交
2075
unsigned long nr_lock_chains;
2076
struct lock_chain lock_chains[MAX_LOCKDEP_CHAINS];
2077
int nr_chain_hlocks;
2078 2079 2080 2081 2082 2083
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
2084

2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103
/*
 * 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;
}

2104
#ifdef CONFIG_DEBUG_LOCKDEP
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 2154 2155 2156 2157
/*
 * 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)
{
2158
	pr_warn("\n");
2159 2160
	pr_warn("============================\n");
	pr_warn("WARNING: chain_key collision\n");
2161
	print_kernel_ident();
2162
	pr_warn("----------------------------\n");
2163 2164
	pr_warn("%s/%d: ", current->comm, task_pid_nr(current));
	pr_warn("Hash chain already cached but the contents don't match!\n");
2165

2166
	pr_warn("Held locks:");
2167 2168
	print_chain_keys_held_locks(curr, hlock_next);

2169
	pr_warn("Locks in cached chain:");
2170 2171
	print_chain_keys_chain(chain);

2172
	pr_warn("\nstack backtrace:\n");
2173 2174
	dump_stack();
}
2175
#endif
2176

2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
/*
 * 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);

2192 2193
	if (DEBUG_LOCKS_WARN_ON(chain->depth != curr->lockdep_depth - (i - 1))) {
		print_collision(curr, hlock, chain);
2194
		return 0;
2195
	}
2196 2197 2198 2199

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

2200 2201
		if (DEBUG_LOCKS_WARN_ON(chain_hlocks[chain->base + j] != id)) {
			print_collision(curr, hlock, chain);
2202
			return 0;
2203
		}
2204 2205 2206 2207 2208
	}
#endif
	return 1;
}

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 2275 2276 2277 2278
/*
 * 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
2279
/*
2280 2281 2282 2283 2284
 * 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
2285
 */
2286 2287 2288
static inline int add_chain_cache(struct task_struct *curr,
				  struct held_lock *hlock,
				  u64 chain_key)
I
Ingo Molnar 已提交
2289
{
D
Dave Jones 已提交
2290
	struct lock_class *class = hlock_class(hlock);
2291
	struct hlist_head *hash_head = chainhashentry(chain_key);
I
Ingo Molnar 已提交
2292
	struct lock_chain *chain;
2293
	int i, j;
I
Ingo Molnar 已提交
2294

2295 2296 2297 2298 2299
	/*
	 * Allocate a new chain entry from the static array, and add
	 * it to the hash:
	 */

P
Peter Zijlstra 已提交
2300 2301 2302 2303 2304
	/*
	 * 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.
	 */
2305 2306
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;
2307

I
Ingo Molnar 已提交
2308
	if (unlikely(nr_lock_chains >= MAX_LOCKDEP_CHAINS)) {
2309 2310 2311
		if (!debug_locks_off_graph_unlock())
			return 0;

2312
		print_lockdep_off("BUG: MAX_LOCKDEP_CHAINS too low!");
2313
		dump_stack();
I
Ingo Molnar 已提交
2314 2315 2316 2317
		return 0;
	}
	chain = lock_chains + nr_lock_chains++;
	chain->chain_key = chain_key;
2318
	chain->irq_context = hlock->irq_context;
2319
	i = get_first_held_lock(curr, hlock);
2320
	chain->depth = curr->lockdep_depth + 1 - i;
2321 2322 2323 2324 2325

	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));

2326 2327
	if (likely(nr_chain_hlocks + chain->depth <= MAX_LOCKDEP_CHAIN_HLOCKS)) {
		chain->base = nr_chain_hlocks;
2328
		for (j = 0; j < chain->depth - 1; j++, i++) {
D
Dave Jones 已提交
2329
			int lock_id = curr->held_locks[i].class_idx - 1;
2330 2331 2332 2333
			chain_hlocks[chain->base + j] = lock_id;
		}
		chain_hlocks[chain->base + j] = class - lock_classes;
	}
2334 2335 2336 2337 2338 2339 2340 2341 2342

	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)) {
2343
		if (!debug_locks_off_graph_unlock())
2344 2345 2346 2347 2348 2349 2350 2351
			return 0;

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

2352
	hlist_add_head_rcu(&chain->entry, hash_head);
2353
	debug_atomic_inc(chain_lookup_misses);
P
Peter Zijlstra 已提交
2354 2355 2356 2357 2358
	inc_chains();

	return 1;
}

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 2427 2428 2429 2430
/*
 * 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
2431
static int validate_chain(struct task_struct *curr, struct lockdep_map *lock,
2432
		struct held_lock *hlock, int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2433 2434 2435 2436 2437 2438 2439 2440
{
	/*
	 * 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.
2441
	 * (If lookup_chain_cache_add() return with 1 it acquires
P
Peter Zijlstra 已提交
2442 2443
	 * graph_lock for us)
	 */
2444
	if (!hlock->trylock && hlock->check &&
2445
	    lookup_chain_cache_add(curr, hlock, chain_key)) {
P
Peter Zijlstra 已提交
2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472
		/*
		 * 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:
		 */
2473
		if (!chain_head && ret != 2) {
P
Peter Zijlstra 已提交
2474 2475
			if (!check_prevs_add(curr, hlock))
				return 0;
2476 2477
		}

P
Peter Zijlstra 已提交
2478
		graph_unlock();
2479 2480
	} else {
		/* after lookup_chain_cache_add(): */
P
Peter Zijlstra 已提交
2481 2482
		if (unlikely(!debug_locks))
			return 0;
2483
	}
I
Ingo Molnar 已提交
2484 2485 2486

	return 1;
}
P
Peter Zijlstra 已提交
2487 2488 2489
#else
static inline int validate_chain(struct task_struct *curr,
	       	struct lockdep_map *lock, struct held_lock *hlock,
2490
		int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2491 2492 2493
{
	return 1;
}
2494
#endif
I
Ingo Molnar 已提交
2495 2496 2497 2498 2499

/*
 * We are building curr_chain_key incrementally, so double-check
 * it from scratch, to make sure that it's done correctly:
 */
2500
static void check_chain_key(struct task_struct *curr)
I
Ingo Molnar 已提交
2501 2502 2503
{
#ifdef CONFIG_DEBUG_LOCKDEP
	struct held_lock *hlock, *prev_hlock = NULL;
2504
	unsigned int i;
I
Ingo Molnar 已提交
2505 2506 2507 2508 2509 2510
	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
2511 2512 2513 2514
			/*
			 * We got mighty confused, our chain keys don't match
			 * with what we expect, someone trample on our task state?
			 */
2515
			WARN(1, "hm#1, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2516 2517 2518 2519 2520
				curr->lockdep_depth, i,
				(unsigned long long)chain_key,
				(unsigned long long)hlock->prev_chain_key);
			return;
		}
P
Peter Zijlstra 已提交
2521 2522 2523
		/*
		 * Whoops ran out of static storage again?
		 */
2524
		if (DEBUG_LOCKS_WARN_ON(hlock->class_idx > MAX_LOCKDEP_KEYS))
2525 2526
			return;

I
Ingo Molnar 已提交
2527 2528 2529
		if (prev_hlock && (prev_hlock->irq_context !=
							hlock->irq_context))
			chain_key = 0;
2530
		chain_key = iterate_chain_key(chain_key, hlock->class_idx);
I
Ingo Molnar 已提交
2531 2532 2533 2534
		prev_hlock = hlock;
	}
	if (chain_key != curr->curr_chain_key) {
		debug_locks_off();
P
Peter Zijlstra 已提交
2535 2536 2537 2538
		/*
		 * More smoking hash instead of calculating it, damn see these
		 * numbers float.. I bet that a pink elephant stepped on my memory.
		 */
2539
		WARN(1, "hm#2, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2540 2541 2542 2543 2544 2545 2546
			curr->lockdep_depth, i,
			(unsigned long long)chain_key,
			(unsigned long long)curr->curr_chain_key);
	}
#endif
}

2547 2548 2549 2550 2551 2552 2553 2554 2555 2556
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
2557
	printk(KERN_CONT ");\n");
2558 2559 2560
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
2561
	printk(KERN_CONT ");\n");
2562 2563 2564
	printk("\n *** DEADLOCK ***\n\n");
}

P
Peter Zijlstra 已提交
2565 2566 2567 2568 2569 2570 2571
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;

2572
	pr_warn("\n");
2573 2574
	pr_warn("================================\n");
	pr_warn("WARNING: inconsistent lock state\n");
2575
	print_kernel_ident();
2576
	pr_warn("--------------------------------\n");
P
Peter Zijlstra 已提交
2577

2578
	pr_warn("inconsistent {%s} -> {%s} usage.\n",
P
Peter Zijlstra 已提交
2579 2580
		usage_str[prev_bit], usage_str[new_bit]);

2581
	pr_warn("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] takes:\n",
2582
		curr->comm, task_pid_nr(curr),
P
Peter Zijlstra 已提交
2583 2584 2585 2586 2587 2588
		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);

2589
	pr_warn("{%s} state was registered at:\n", usage_str[prev_bit]);
D
Dave Jones 已提交
2590
	print_stack_trace(hlock_class(this)->usage_traces + prev_bit, 1);
P
Peter Zijlstra 已提交
2591 2592

	print_irqtrace_events(curr);
2593
	pr_warn("\nother info that might help us debug this:\n");
2594 2595
	print_usage_bug_scenario(this);

P
Peter Zijlstra 已提交
2596 2597
	lockdep_print_held_locks(curr);

2598
	pr_warn("\nstack backtrace:\n");
P
Peter Zijlstra 已提交
2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610
	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
2611
	if (unlikely(hlock_class(this)->usage_mask & (1 << bad_bit)))
P
Peter Zijlstra 已提交
2612 2613 2614 2615 2616 2617 2618
		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);

2619
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
2620 2621 2622 2623 2624

/*
 * print irq inversion bug:
 */
static int
2625 2626
print_irq_inversion_bug(struct task_struct *curr,
			struct lock_list *root, struct lock_list *other,
I
Ingo Molnar 已提交
2627 2628 2629
			struct held_lock *this, int forwards,
			const char *irqclass)
{
2630 2631 2632 2633
	struct lock_list *entry = other;
	struct lock_list *middle = NULL;
	int depth;

2634
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
2635 2636
		return 0;

2637
	pr_warn("\n");
2638 2639
	pr_warn("========================================================\n");
	pr_warn("WARNING: possible irq lock inversion dependency detected\n");
2640
	print_kernel_ident();
2641
	pr_warn("--------------------------------------------------------\n");
2642
	pr_warn("%s/%d just changed the state of lock:\n",
2643
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
2644 2645
	print_lock(this);
	if (forwards)
2646
		pr_warn("but this lock took another, %s-unsafe lock in the past:\n", irqclass);
I
Ingo Molnar 已提交
2647
	else
2648
		pr_warn("but this lock was taken by another, %s-safe lock in the past:\n", irqclass);
2649
	print_lock_name(other->class);
2650
	pr_warn("\n\nand interrupts could create inverse lock ordering between them.\n\n");
I
Ingo Molnar 已提交
2651

2652
	pr_warn("\nother info that might help us debug this:\n");
2653 2654 2655 2656 2657

	/* Find a middle lock (if one exists) */
	depth = get_lock_depth(other);
	do {
		if (depth == 0 && (entry != root)) {
2658
			pr_warn("lockdep:%s bad path found in chain graph\n", __func__);
2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671
			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
2672 2673
	lockdep_print_held_locks(curr);

2674
	pr_warn("\nthe shortest dependencies between 2nd lock and 1st lock:\n");
2675 2676 2677
	if (!save_trace(&root->trace))
		return 0;
	print_shortest_lock_dependencies(other, root);
I
Ingo Molnar 已提交
2678

2679
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693
	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;
2694 2695
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2696

2697 2698 2699
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_forwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2700 2701 2702 2703
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2704

2705
	return print_irq_inversion_bug(curr, &root, target_entry,
2706
					this, 1, irqclass);
I
Ingo Molnar 已提交
2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717
}

/*
 * 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;
2718 2719
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2720

2721 2722 2723
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_backwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2724 2725 2726 2727
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2728

2729
	return print_irq_inversion_bug(curr, &root, target_entry,
2730
					this, 0, irqclass);
I
Ingo Molnar 已提交
2731 2732
}

2733
void print_irqtrace_events(struct task_struct *curr)
I
Ingo Molnar 已提交
2734 2735
{
	printk("irq event stamp: %u\n", curr->irq_events);
D
Dmitry Vyukov 已提交
2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
	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
2748 2749
}

2750
static int HARDIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2751
{
P
Peter Zijlstra 已提交
2752 2753 2754
#if HARDIRQ_VERBOSE
	return class_filter(class);
#endif
I
Ingo Molnar 已提交
2755 2756 2757
	return 0;
}

2758
static int SOFTIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2759
{
P
Peter Zijlstra 已提交
2760 2761 2762 2763
#if SOFTIRQ_VERBOSE
	return class_filter(class);
#endif
	return 0;
I
Ingo Molnar 已提交
2764 2765 2766 2767
}

#define STRICT_READ_CHECKS	1

2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780
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);
}

2781 2782 2783
typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
			     enum lock_usage_bit bit, const char *name);

2784
static int
2785 2786
mark_lock_irq(struct task_struct *curr, struct held_lock *this,
		enum lock_usage_bit new_bit)
2787
{
2788
	int excl_bit = exclusive_bit(new_bit);
2789
	int read = new_bit & 1;
2790 2791
	int dir = new_bit & 2;

2792 2793 2794 2795 2796 2797 2798
	/*
	 * 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.
	 */
2799 2800
	check_usage_f usage = dir ?
		check_usage_backwards : check_usage_forwards;
2801

2802 2803 2804 2805
	/*
	 * Validate that this particular lock does not have conflicting
	 * usage states.
	 */
2806 2807
	if (!valid_state(curr, this, new_bit, excl_bit))
		return 0;
2808

2809 2810 2811 2812 2813
	/*
	 * Validate that the lock dependencies don't have conflicting usage
	 * states.
	 */
	if ((!read || !dir || STRICT_READ_CHECKS) &&
2814
			!usage(curr, this, excl_bit, state_name(new_bit & ~1)))
2815
		return 0;
2816

2817 2818 2819 2820 2821 2822 2823 2824
	/*
	 * Check for read in write conflicts
	 */
	if (!read) {
		if (!valid_state(curr, this, new_bit, excl_bit + 1))
			return 0;

		if (STRICT_READ_CHECKS &&
2825 2826
			!usage(curr, this, excl_bit + 1,
				state_name(new_bit + 1)))
2827 2828
			return 0;
	}
2829

2830
	if (state_verbose(new_bit, hlock_class(this)))
2831 2832 2833 2834 2835
		return 2;

	return 1;
}

2836
enum mark_type {
2837 2838 2839
#define LOCKDEP_STATE(__STATE)	__STATE,
#include "lockdep_states.h"
#undef LOCKDEP_STATE
2840 2841
};

I
Ingo Molnar 已提交
2842 2843 2844
/*
 * Mark all held locks with a usage bit:
 */
2845
static int
2846
mark_held_locks(struct task_struct *curr, enum mark_type mark)
I
Ingo Molnar 已提交
2847 2848 2849 2850 2851 2852 2853 2854
{
	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;

2855 2856 2857 2858 2859
		usage_bit = 2 + (mark << 2); /* ENABLED */
		if (hlock->read)
			usage_bit += 1; /* READ */

		BUG_ON(usage_bit >= LOCK_USAGE_STATES);
2860

2861
		if (!hlock->check)
2862 2863
			continue;

2864
		if (!mark_lock(curr, hlock, usage_bit))
I
Ingo Molnar 已提交
2865 2866 2867 2868 2869 2870 2871 2872 2873
			return 0;
	}

	return 1;
}

/*
 * Hardirqs will be enabled:
 */
2874
static void __trace_hardirqs_on_caller(unsigned long ip)
I
Ingo Molnar 已提交
2875 2876 2877 2878 2879 2880 2881 2882 2883 2884
{
	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:
	 */
2885
	if (!mark_held_locks(curr, HARDIRQ))
I
Ingo Molnar 已提交
2886 2887 2888 2889 2890 2891 2892
		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)
2893
		if (!mark_held_locks(curr, SOFTIRQ))
I
Ingo Molnar 已提交
2894 2895
			return;

P
Peter Zijlstra 已提交
2896 2897
	curr->hardirq_enable_ip = ip;
	curr->hardirq_enable_event = ++curr->irq_events;
2898
	debug_atomic_inc(hardirqs_on_events);
P
Peter Zijlstra 已提交
2899
}
2900

2901
__visible void trace_hardirqs_on_caller(unsigned long ip)
2902 2903 2904 2905 2906 2907
{
	time_hardirqs_on(CALLER_ADDR0, ip);

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

2908 2909 2910 2911 2912 2913 2914 2915 2916 2917
	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
2918 2919 2920 2921 2922
	/*
	 * 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.
	 */
2923 2924 2925
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

P
Peter Zijlstra 已提交
2926 2927 2928
	/*
	 * See the fine text that goes along with this variable definition.
	 */
2929 2930 2931
	if (DEBUG_LOCKS_WARN_ON(unlikely(early_boot_irqs_disabled)))
		return;

P
Peter Zijlstra 已提交
2932 2933 2934 2935
	/*
	 * Can't allow enabling interrupts while in an interrupt handler,
	 * that's general bad form and such. Recursion, limited stack etc..
	 */
2936 2937 2938
	if (DEBUG_LOCKS_WARN_ON(current->hardirq_context))
		return;

2939 2940 2941 2942
	current->lockdep_recursion = 1;
	__trace_hardirqs_on_caller(ip);
	current->lockdep_recursion = 0;
}
2943
EXPORT_SYMBOL(trace_hardirqs_on_caller);
P
Peter Zijlstra 已提交
2944

2945
void trace_hardirqs_on(void)
2946 2947 2948
{
	trace_hardirqs_on_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2949 2950 2951 2952 2953
EXPORT_SYMBOL(trace_hardirqs_on);

/*
 * Hardirqs were disabled:
 */
2954
__visible void trace_hardirqs_off_caller(unsigned long ip)
P
Peter Zijlstra 已提交
2955 2956 2957
{
	struct task_struct *curr = current;

2958
	time_hardirqs_off(CALLER_ADDR0, ip);
2959

P
Peter Zijlstra 已提交
2960 2961 2962
	if (unlikely(!debug_locks || current->lockdep_recursion))
		return;

P
Peter Zijlstra 已提交
2963 2964 2965 2966
	/*
	 * 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
2967 2968 2969 2970 2971 2972 2973 2974
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->hardirqs_enabled) {
		/*
		 * We have done an ON -> OFF transition:
		 */
		curr->hardirqs_enabled = 0;
2975
		curr->hardirq_disable_ip = ip;
P
Peter Zijlstra 已提交
2976
		curr->hardirq_disable_event = ++curr->irq_events;
2977
		debug_atomic_inc(hardirqs_off_events);
P
Peter Zijlstra 已提交
2978
	} else
2979
		debug_atomic_inc(redundant_hardirqs_off);
P
Peter Zijlstra 已提交
2980
}
2981
EXPORT_SYMBOL(trace_hardirqs_off_caller);
P
Peter Zijlstra 已提交
2982

2983
void trace_hardirqs_off(void)
2984 2985 2986
{
	trace_hardirqs_off_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2987 2988 2989 2990 2991 2992 2993 2994 2995
EXPORT_SYMBOL(trace_hardirqs_off);

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

2996
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
2997 2998
		return;

P
Peter Zijlstra 已提交
2999 3000 3001 3002
	/*
	 * We fancy IRQs being disabled here, see softirq.c, avoids
	 * funny state and nesting things.
	 */
P
Peter Zijlstra 已提交
3003 3004 3005 3006
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->softirqs_enabled) {
3007
		debug_atomic_inc(redundant_softirqs_on);
P
Peter Zijlstra 已提交
3008 3009 3010
		return;
	}

3011
	current->lockdep_recursion = 1;
P
Peter Zijlstra 已提交
3012 3013 3014 3015 3016 3017
	/*
	 * We'll do an OFF -> ON transition:
	 */
	curr->softirqs_enabled = 1;
	curr->softirq_enable_ip = ip;
	curr->softirq_enable_event = ++curr->irq_events;
3018
	debug_atomic_inc(softirqs_on_events);
P
Peter Zijlstra 已提交
3019 3020 3021 3022 3023 3024
	/*
	 * 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)
3025
		mark_held_locks(curr, SOFTIRQ);
3026
	current->lockdep_recursion = 0;
P
Peter Zijlstra 已提交
3027 3028 3029 3030 3031 3032 3033 3034 3035
}

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

3036
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
3037 3038
		return;

P
Peter Zijlstra 已提交
3039 3040 3041
	/*
	 * We fancy IRQs being disabled here, see softirq.c
	 */
P
Peter Zijlstra 已提交
3042 3043 3044 3045 3046 3047 3048 3049 3050 3051
	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;
3052
		debug_atomic_inc(softirqs_off_events);
P
Peter Zijlstra 已提交
3053 3054 3055
		/*
		 * Whoops, we wanted softirqs off, so why aren't they?
		 */
P
Peter Zijlstra 已提交
3056 3057
		DEBUG_LOCKS_WARN_ON(!softirq_count());
	} else
3058
		debug_atomic_inc(redundant_softirqs_off);
P
Peter Zijlstra 已提交
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 3085 3086 3087 3088
}

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
3089
					LOCK_ENABLED_HARDIRQ_READ))
P
Peter Zijlstra 已提交
3090 3091 3092
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
3093
						LOCK_ENABLED_SOFTIRQ_READ))
P
Peter Zijlstra 已提交
3094 3095 3096
					return 0;
		} else {
			if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
3097
					LOCK_ENABLED_HARDIRQ))
P
Peter Zijlstra 已提交
3098 3099 3100
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
3101
						LOCK_ENABLED_SOFTIRQ))
P
Peter Zijlstra 已提交
3102 3103 3104 3105 3106 3107 3108
					return 0;
		}
	}

	return 1;
}

3109 3110 3111 3112 3113
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
3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134
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
3135 3136
}

P
Peter Zijlstra 已提交
3137
#else /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
I
Ingo Molnar 已提交
3138

P
Peter Zijlstra 已提交
3139 3140 3141
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
3142
{
P
Peter Zijlstra 已提交
3143
	WARN_ON(1); /* Impossible innit? when we don't have TRACE_IRQFLAG */
P
Peter Zijlstra 已提交
3144 3145
	return 1;
}
I
Ingo Molnar 已提交
3146

P
Peter Zijlstra 已提交
3147 3148 3149 3150 3151
static inline int mark_irqflags(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 1;
}
I
Ingo Molnar 已提交
3152

3153 3154 3155 3156 3157
static inline unsigned int task_irq_context(struct task_struct *task)
{
	return 0;
}

P
Peter Zijlstra 已提交
3158 3159 3160 3161
static inline int separate_irq_context(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 0;
I
Ingo Molnar 已提交
3162 3163
}

P
Peter Zijlstra 已提交
3164
#endif /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
I
Ingo Molnar 已提交
3165 3166

/*
P
Peter Zijlstra 已提交
3167
 * Mark a lock with a usage bit, and validate the state transition:
I
Ingo Molnar 已提交
3168
 */
3169
static int mark_lock(struct task_struct *curr, struct held_lock *this,
3170
			     enum lock_usage_bit new_bit)
I
Ingo Molnar 已提交
3171
{
P
Peter Zijlstra 已提交
3172
	unsigned int new_mask = 1 << new_bit, ret = 1;
I
Ingo Molnar 已提交
3173 3174

	/*
P
Peter Zijlstra 已提交
3175 3176
	 * If already set then do not dirty the cacheline,
	 * nor do any checks:
I
Ingo Molnar 已提交
3177
	 */
D
Dave Jones 已提交
3178
	if (likely(hlock_class(this)->usage_mask & new_mask))
P
Peter Zijlstra 已提交
3179 3180 3181 3182
		return 1;

	if (!graph_lock())
		return 0;
I
Ingo Molnar 已提交
3183
	/*
L
Lucas De Marchi 已提交
3184
	 * Make sure we didn't race:
I
Ingo Molnar 已提交
3185
	 */
D
Dave Jones 已提交
3186
	if (unlikely(hlock_class(this)->usage_mask & new_mask)) {
P
Peter Zijlstra 已提交
3187 3188 3189
		graph_unlock();
		return 1;
	}
I
Ingo Molnar 已提交
3190

D
Dave Jones 已提交
3191
	hlock_class(this)->usage_mask |= new_mask;
I
Ingo Molnar 已提交
3192

D
Dave Jones 已提交
3193
	if (!save_trace(hlock_class(this)->usage_traces + new_bit))
P
Peter Zijlstra 已提交
3194
		return 0;
I
Ingo Molnar 已提交
3195

P
Peter Zijlstra 已提交
3196
	switch (new_bit) {
P
Peter Zijlstra 已提交
3197 3198 3199 3200 3201 3202 3203
#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
3204 3205 3206 3207 3208
		ret = mark_lock_irq(curr, this, new_bit);
		if (!ret)
			return 0;
		break;
	case LOCK_USED:
3209
		debug_atomic_dec(nr_unused_locks);
P
Peter Zijlstra 已提交
3210 3211 3212 3213 3214 3215 3216
		break;
	default:
		if (!debug_locks_off_graph_unlock())
			return 0;
		WARN_ON(1);
		return 0;
	}
I
Ingo Molnar 已提交
3217

P
Peter Zijlstra 已提交
3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231
	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
3232 3233 3234 3235

/*
 * Initialize a lock instance's lock-class mapping info:
 */
3236
static void __lockdep_init_map(struct lockdep_map *lock, const char *name,
3237
		      struct lock_class_key *key, int subclass)
I
Ingo Molnar 已提交
3238
{
3239 3240 3241 3242 3243 3244
	int i;

	kmemcheck_mark_initialized(lock, sizeof(*lock));

	for (i = 0; i < NR_LOCKDEP_CACHING_CLASSES; i++)
		lock->class_cache[i] = NULL;
3245

3246 3247 3248 3249
#ifdef CONFIG_LOCK_STAT
	lock->cpu = raw_smp_processor_id();
#endif

P
Peter Zijlstra 已提交
3250 3251 3252
	/*
	 * Can't be having no nameless bastards around this place!
	 */
3253 3254
	if (DEBUG_LOCKS_WARN_ON(!name)) {
		lock->name = "NULL";
I
Ingo Molnar 已提交
3255
		return;
3256 3257 3258
	}

	lock->name = name;
I
Ingo Molnar 已提交
3259

P
Peter Zijlstra 已提交
3260 3261 3262
	/*
	 * No key, no joy, we need to hash something.
	 */
I
Ingo Molnar 已提交
3263 3264 3265 3266 3267 3268 3269
	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
3270 3271 3272
		/*
		 * What it says above ^^^^^, I suggest you read it.
		 */
I
Ingo Molnar 已提交
3273 3274 3275 3276
		DEBUG_LOCKS_WARN_ON(1);
		return;
	}
	lock->key = key;
3277 3278 3279 3280

	if (unlikely(!debug_locks))
		return;

3281 3282 3283 3284 3285 3286 3287 3288
	if (subclass) {
		unsigned long flags;

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

		raw_local_irq_save(flags);
		current->lockdep_recursion = 1;
3289
		register_lock_class(lock, subclass, 1);
3290 3291 3292
		current->lockdep_recursion = 0;
		raw_local_irq_restore(flags);
	}
I
Ingo Molnar 已提交
3293
}
3294 3295 3296 3297 3298 3299 3300

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
3301 3302
EXPORT_SYMBOL_GPL(lockdep_init_map);

3303 3304 3305 3306 3307 3308 3309 3310 3311 3312
#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

3313
struct lock_class_key __lockdep_no_validate__;
K
Kent Overstreet 已提交
3314
EXPORT_SYMBOL_GPL(__lockdep_no_validate__);
3315

3316 3317 3318 3319 3320 3321 3322 3323 3324 3325
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;

3326
	pr_warn("\n");
3327 3328
	pr_warn("==================================\n");
	pr_warn("WARNING: Nested lock was not taken\n");
3329
	print_kernel_ident();
3330
	pr_warn("----------------------------------\n");
3331

3332
	pr_warn("%s/%d is trying to lock:\n", curr->comm, task_pid_nr(curr));
3333 3334
	print_lock(hlock);

3335 3336
	pr_warn("\nbut this task is not holding:\n");
	pr_warn("%s\n", hlock->nest_lock->name);
3337

3338
	pr_warn("\nstack backtrace:\n");
3339 3340
	dump_stack();

3341
	pr_warn("\nother info that might help us debug this:\n");
3342 3343
	lockdep_print_held_locks(curr);

3344
	pr_warn("\nstack backtrace:\n");
3345 3346 3347 3348 3349
	dump_stack();

	return 0;
}

3350
static int __lock_is_held(struct lockdep_map *lock, int read);
3351

I
Ingo Molnar 已提交
3352 3353 3354 3355 3356 3357
/*
 * 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,
3358
			  struct lockdep_map *nest_lock, unsigned long ip,
3359
			  int references, int pin_count)
I
Ingo Molnar 已提交
3360 3361
{
	struct task_struct *curr = current;
3362
	struct lock_class *class = NULL;
I
Ingo Molnar 已提交
3363
	struct held_lock *hlock;
3364
	unsigned int depth;
I
Ingo Molnar 已提交
3365
	int chain_head = 0;
3366
	int class_idx;
I
Ingo Molnar 已提交
3367
	u64 chain_key;
3368
	int ret;
I
Ingo Molnar 已提交
3369 3370 3371 3372

	if (unlikely(!debug_locks))
		return 0;

P
Peter Zijlstra 已提交
3373 3374 3375 3376 3377
	/*
	 * 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
3378 3379 3380
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;

3381 3382
	if (!prove_locking || lock->key == &__lockdep_no_validate__)
		check = 0;
3383

3384 3385
	if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		class = lock->class_cache[subclass];
3386
	/*
3387
	 * Not cached?
3388
	 */
I
Ingo Molnar 已提交
3389
	if (unlikely(!class)) {
3390
		class = register_lock_class(lock, subclass, 0);
I
Ingo Molnar 已提交
3391 3392 3393
		if (!class)
			return 0;
	}
3394
	atomic_inc((atomic_t *)&class->ops);
I
Ingo Molnar 已提交
3395 3396 3397
	if (very_verbose(class)) {
		printk("\nacquire class [%p] %s", class->key, class->name);
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
3398 3399
			printk(KERN_CONT "#%d", class->name_version);
		printk(KERN_CONT "\n");
I
Ingo Molnar 已提交
3400 3401 3402 3403 3404 3405 3406 3407 3408
		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
3409 3410 3411
	/*
	 * Ran out of static storage for our per-task lock stack again have we?
	 */
I
Ingo Molnar 已提交
3412 3413 3414
	if (DEBUG_LOCKS_WARN_ON(depth >= MAX_LOCK_DEPTH))
		return 0;

3415 3416
	class_idx = class - lock_classes + 1;

3417 3418
	/* TODO: nest_lock is not implemented for crosslock yet. */
	if (depth && !cross_lock(lock)) {
3419 3420
		hlock = curr->held_locks + depth - 1;
		if (hlock->class_idx == class_idx && nest_lock) {
3421 3422 3423 3424 3425 3426 3427
			if (hlock->references) {
				/*
				 * Check: unsigned int references:12, overflow.
				 */
				if (DEBUG_LOCKS_WARN_ON(hlock->references == (1 << 12)-1))
					return 0;

3428
				hlock->references++;
3429
			} else {
3430
				hlock->references = 2;
3431
			}
3432 3433 3434 3435 3436

			return 1;
		}
	}

I
Ingo Molnar 已提交
3437
	hlock = curr->held_locks + depth;
P
Peter Zijlstra 已提交
3438 3439 3440 3441
	/*
	 * 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
3442 3443
	if (DEBUG_LOCKS_WARN_ON(!class))
		return 0;
3444
	hlock->class_idx = class_idx;
I
Ingo Molnar 已提交
3445 3446
	hlock->acquire_ip = ip;
	hlock->instance = lock;
P
Peter Zijlstra 已提交
3447
	hlock->nest_lock = nest_lock;
3448
	hlock->irq_context = task_irq_context(curr);
I
Ingo Molnar 已提交
3449 3450 3451
	hlock->trylock = trylock;
	hlock->read = read;
	hlock->check = check;
3452
	hlock->hardirqs_off = !!hardirqs_off;
3453
	hlock->references = references;
P
Peter Zijlstra 已提交
3454 3455
#ifdef CONFIG_LOCK_STAT
	hlock->waittime_stamp = 0;
3456
	hlock->holdtime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
3457
#endif
3458
	hlock->pin_count = pin_count;
I
Ingo Molnar 已提交
3459

3460
	if (check && !mark_irqflags(curr, hlock))
P
Peter Zijlstra 已提交
3461 3462
		return 0;

I
Ingo Molnar 已提交
3463
	/* mark it as used: */
3464
	if (!mark_lock(curr, hlock, LOCK_USED))
I
Ingo Molnar 已提交
3465
		return 0;
P
Peter Zijlstra 已提交
3466

I
Ingo Molnar 已提交
3467
	/*
3468
	 * Calculate the chain hash: it's the combined hash of all the
I
Ingo Molnar 已提交
3469 3470 3471 3472 3473 3474 3475 3476
	 * 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
3477 3478 3479
	/*
	 * Whoops, we did it again.. ran straight out of our static allocation.
	 */
3480
	if (DEBUG_LOCKS_WARN_ON(class_idx > MAX_LOCKDEP_KEYS))
I
Ingo Molnar 已提交
3481 3482 3483 3484
		return 0;

	chain_key = curr->curr_chain_key;
	if (!depth) {
P
Peter Zijlstra 已提交
3485 3486 3487
		/*
		 * How can we have a chain hash when we ain't got no keys?!
		 */
I
Ingo Molnar 已提交
3488 3489 3490 3491 3492 3493
		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
3494 3495 3496
	if (separate_irq_context(curr, hlock)) {
		chain_key = 0;
		chain_head = 1;
I
Ingo Molnar 已提交
3497
	}
3498
	chain_key = iterate_chain_key(chain_key, class_idx);
I
Ingo Molnar 已提交
3499

3500
	if (nest_lock && !__lock_is_held(nest_lock, -1))
3501 3502
		return print_lock_nested_lock_not_held(curr, hlock, ip);

3503
	if (!validate_chain(curr, lock, hlock, chain_head, chain_key))
P
Peter Zijlstra 已提交
3504
		return 0;
3505

3506 3507 3508 3509 3510 3511 3512 3513
	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;

3514
	curr->curr_chain_key = chain_key;
I
Ingo Molnar 已提交
3515 3516
	curr->lockdep_depth++;
	check_chain_key(curr);
3517 3518 3519 3520
#ifdef CONFIG_DEBUG_LOCKDEP
	if (unlikely(!debug_locks))
		return 0;
#endif
I
Ingo Molnar 已提交
3521 3522
	if (unlikely(curr->lockdep_depth >= MAX_LOCK_DEPTH)) {
		debug_locks_off();
3523 3524
		print_lockdep_off("BUG: MAX_LOCK_DEPTH too low!");
		printk(KERN_DEBUG "depth: %i  max: %lu!\n",
3525 3526 3527 3528
		       curr->lockdep_depth, MAX_LOCK_DEPTH);

		lockdep_print_held_locks(current);
		debug_show_all_locks();
3529
		dump_stack();
3530

I
Ingo Molnar 已提交
3531 3532
		return 0;
	}
3533

I
Ingo Molnar 已提交
3534 3535 3536 3537 3538 3539 3540
	if (unlikely(curr->lockdep_depth > max_lockdep_depth))
		max_lockdep_depth = curr->lockdep_depth;

	return 1;
}

static int
3541
print_unlock_imbalance_bug(struct task_struct *curr, struct lockdep_map *lock,
I
Ingo Molnar 已提交
3542 3543 3544 3545 3546 3547 3548
			   unsigned long ip)
{
	if (!debug_locks_off())
		return 0;
	if (debug_locks_silent)
		return 0;

3549
	pr_warn("\n");
3550 3551
	pr_warn("=====================================\n");
	pr_warn("WARNING: bad unlock balance detected!\n");
3552
	print_kernel_ident();
3553
	pr_warn("-------------------------------------\n");
3554
	pr_warn("%s/%d is trying to release lock (",
3555
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
3556
	print_lockdep_cache(lock);
3557
	pr_cont(") at:\n");
I
Ingo Molnar 已提交
3558
	print_ip_sym(ip);
3559 3560
	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
3561 3562
	lockdep_print_held_locks(curr);

3563
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
3564 3565 3566 3567 3568
	dump_stack();

	return 0;
}

3569 3570 3571 3572 3573 3574
static int match_held_lock(struct held_lock *hlock, struct lockdep_map *lock)
{
	if (hlock->instance == lock)
		return 1;

	if (hlock->references) {
3575
		struct lock_class *class = lock->class_cache[0];
3576 3577 3578 3579

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

3580 3581 3582 3583 3584 3585
		/*
		 * 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.
		 */
3586
		if (IS_ERR_OR_NULL(class))
3587 3588
			return 0;

P
Peter Zijlstra 已提交
3589 3590 3591 3592 3593
		/*
		 * 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.
		 */
3594 3595 3596 3597 3598 3599 3600 3601 3602 3603
		if (DEBUG_LOCKS_WARN_ON(!hlock->nest_lock))
			return 0;

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

	return 0;
}

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 3636 3637 3638 3639
/* @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;
}

3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657
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;
}

3658
static int
3659 3660 3661
__lock_set_class(struct lockdep_map *lock, const char *name,
		 struct lock_class_key *key, unsigned int subclass,
		 unsigned long ip)
3662 3663
{
	struct task_struct *curr = current;
3664
	struct held_lock *hlock;
3665 3666 3667 3668 3669
	struct lock_class *class;
	unsigned int depth;
	int i;

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3670 3671 3672 3673
	/*
	 * This function is about (re)setting the class of a held lock,
	 * yet we're not actually holding any locks. Naughty user!
	 */
3674 3675 3676
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return 0;

3677 3678 3679
	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock)
		return print_unlock_imbalance_bug(curr, lock, ip);
3680

3681
	lockdep_init_map(lock, name, key, 0);
3682
	class = register_lock_class(lock, subclass, 0);
D
Dave Jones 已提交
3683
	hlock->class_idx = class - lock_classes + 1;
3684 3685 3686 3687

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

3688 3689
	if (reacquire_held_locks(curr, depth, i))
		return 0;
3690

P
Peter Zijlstra 已提交
3691 3692 3693 3694
	/*
	 * I took it apart and put it back together again, except now I have
	 * these 'spare' parts.. where shall I put them.
	 */
3695 3696 3697 3698 3699
	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth))
		return 0;
	return 1;
}

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 3734 3735 3736 3737
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
3738
/*
3739 3740 3741 3742 3743
 * 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
3744 3745
 */
static int
3746
__lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
I
Ingo Molnar 已提交
3747
{
3748
	struct task_struct *curr = current;
3749
	struct held_lock *hlock;
I
Ingo Molnar 已提交
3750
	unsigned int depth;
3751
	int ret, i;
I
Ingo Molnar 已提交
3752

3753 3754 3755
	if (unlikely(!debug_locks))
		return 0;

3756 3757 3758 3759 3760 3761 3762 3763
	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
3764
	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3765 3766 3767 3768
	/*
	 * So we're all set to release this lock.. wait what lock? We don't
	 * own any locks, you've been drinking again?
	 */
3769 3770
	if (DEBUG_LOCKS_WARN_ON(depth <= 0))
		 return print_unlock_imbalance_bug(curr, lock, ip);
I
Ingo Molnar 已提交
3771

3772 3773 3774 3775
	/*
	 * Check whether the lock exists in the current stack
	 * of held locks:
	 */
3776 3777 3778
	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock)
		return print_unlock_imbalance_bug(curr, lock, ip);
I
Ingo Molnar 已提交
3779

3780 3781 3782
	if (hlock->instance == lock)
		lock_release_holdtime(hlock);

P
Peter Zijlstra 已提交
3783 3784
	WARN(hlock->pin_count, "releasing a pinned lock\n");

3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795
	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
3796

I
Ingo Molnar 已提交
3797 3798 3799 3800 3801
	/*
	 * 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:
	 */
3802

I
Ingo Molnar 已提交
3803 3804 3805
	curr->lockdep_depth = i;
	curr->curr_chain_key = hlock->prev_chain_key;

3806 3807
	if (reacquire_held_locks(curr, depth, i + 1))
		return 0;
I
Ingo Molnar 已提交
3808

P
Peter Zijlstra 已提交
3809 3810 3811 3812
	/*
	 * 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
3813 3814
	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth - 1))
		return 0;
P
Peter Zijlstra 已提交
3815

I
Ingo Molnar 已提交
3816 3817 3818
	return 1;
}

3819
static int __lock_is_held(struct lockdep_map *lock, int read)
I
Ingo Molnar 已提交
3820
{
3821 3822
	struct task_struct *curr = current;
	int i;
I
Ingo Molnar 已提交
3823

3824
	for (i = 0; i < curr->lockdep_depth; i++) {
3825
		struct held_lock *hlock = curr->held_locks + i;
I
Ingo Molnar 已提交
3826

3827 3828 3829 3830 3831 3832
		if (match_held_lock(hlock, lock)) {
			if (read == -1 || hlock->read == read)
				return 1;

			return 0;
		}
3833
	}
P
Peter Zijlstra 已提交
3834

3835
	return 0;
I
Ingo Molnar 已提交
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 3863 3864 3865 3866
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
3867 3868
{
	struct task_struct *curr = current;
P
Peter Zijlstra 已提交
3869
	int i;
I
Ingo Molnar 已提交
3870

P
Peter Zijlstra 已提交
3871
	if (unlikely(!debug_locks))
I
Ingo Molnar 已提交
3872 3873
		return;

P
Peter Zijlstra 已提交
3874 3875 3876 3877
	for (i = 0; i < curr->lockdep_depth; i++) {
		struct held_lock *hlock = curr->held_locks + i;

		if (match_held_lock(hlock, lock)) {
3878
			hlock->pin_count += cookie.val;
I
Ingo Molnar 已提交
3879
			return;
P
Peter Zijlstra 已提交
3880
		}
I
Ingo Molnar 已提交
3881 3882
	}

P
Peter Zijlstra 已提交
3883
	WARN(1, "pinning an unheld lock\n");
I
Ingo Molnar 已提交
3884 3885
}

3886
static void __lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
3887 3888 3889 3890
{
	struct task_struct *curr = current;
	int i;

P
Peter Zijlstra 已提交
3891 3892 3893
	if (unlikely(!debug_locks))
		return;

3894
	for (i = 0; i < curr->lockdep_depth; i++) {
3895 3896
		struct held_lock *hlock = curr->held_locks + i;

P
Peter Zijlstra 已提交
3897 3898 3899 3900
		if (match_held_lock(hlock, lock)) {
			if (WARN(!hlock->pin_count, "unpinning an unpinned lock\n"))
				return;

3901 3902 3903 3904 3905
			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
3906 3907
			return;
		}
3908 3909
	}

P
Peter Zijlstra 已提交
3910
	WARN(1, "unpinning an unheld lock\n");
3911 3912
}

I
Ingo Molnar 已提交
3913 3914 3915
/*
 * Check whether we follow the irq-flags state precisely:
 */
3916
static void check_flags(unsigned long flags)
I
Ingo Molnar 已提交
3917
{
3918 3919
#if defined(CONFIG_PROVE_LOCKING) && defined(CONFIG_DEBUG_LOCKDEP) && \
    defined(CONFIG_TRACE_IRQFLAGS)
I
Ingo Molnar 已提交
3920 3921 3922
	if (!debug_locks)
		return;

3923 3924 3925 3926 3927 3928 3929 3930 3931
	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
3932 3933 3934 3935 3936 3937 3938

	/*
	 * 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
3939 3940
		if (softirq_count()) {
			/* like the above, but with softirqs */
I
Ingo Molnar 已提交
3941
			DEBUG_LOCKS_WARN_ON(current->softirqs_enabled);
P
Peter Zijlstra 已提交
3942 3943
		} else {
			/* lick the above, does it taste good? */
I
Ingo Molnar 已提交
3944
			DEBUG_LOCKS_WARN_ON(!current->softirqs_enabled);
P
Peter Zijlstra 已提交
3945
		}
I
Ingo Molnar 已提交
3946 3947 3948 3949 3950 3951 3952
	}

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

3953 3954 3955
void lock_set_class(struct lockdep_map *lock, const char *name,
		    struct lock_class_key *key, unsigned int subclass,
		    unsigned long ip)
3956 3957 3958 3959 3960 3961 3962 3963 3964
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	current->lockdep_recursion = 1;
	check_flags(flags);
3965
	if (__lock_set_class(lock, name, key, subclass, ip))
3966 3967 3968 3969
		check_chain_key(current);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
3970
EXPORT_SYMBOL_GPL(lock_set_class);
3971

3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988
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
3989 3990 3991 3992
/*
 * We are not always called with irqs disabled - do that here,
 * and also avoid lockdep recursion:
 */
3993
void lock_acquire(struct lockdep_map *lock, unsigned int subclass,
P
Peter Zijlstra 已提交
3994 3995
			  int trylock, int read, int check,
			  struct lockdep_map *nest_lock, unsigned long ip)
I
Ingo Molnar 已提交
3996 3997 3998 3999 4000 4001 4002 4003 4004 4005
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4006
	trace_lock_acquire(lock, subclass, trylock, read, check, nest_lock, ip);
I
Ingo Molnar 已提交
4007
	__lock_acquire(lock, subclass, trylock, read, check,
4008
		       irqs_disabled_flags(flags), nest_lock, ip, 0, 0);
I
Ingo Molnar 已提交
4009 4010 4011 4012 4013
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquire);

4014
void lock_release(struct lockdep_map *lock, int nested,
4015
			  unsigned long ip)
I
Ingo Molnar 已提交
4016 4017 4018 4019 4020 4021 4022 4023 4024
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
4025
	trace_lock_release(lock, ip);
4026 4027
	if (__lock_release(lock, nested, ip))
		check_chain_key(current);
I
Ingo Molnar 已提交
4028 4029 4030 4031 4032
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_release);

4033
int lock_is_held_type(struct lockdep_map *lock, int read)
4034 4035 4036 4037 4038
{
	unsigned long flags;
	int ret = 0;

	if (unlikely(current->lockdep_recursion))
4039
		return 1; /* avoid false negative lockdep_assert_held() */
4040 4041 4042 4043 4044

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4045
	ret = __lock_is_held(lock, read);
4046 4047 4048 4049 4050
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);

	return ret;
}
4051
EXPORT_SYMBOL_GPL(lock_is_held_type);
4052

4053
struct pin_cookie lock_pin_lock(struct lockdep_map *lock)
P
Peter Zijlstra 已提交
4054
{
4055
	struct pin_cookie cookie = NIL_COOKIE;
P
Peter Zijlstra 已提交
4056 4057 4058
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
4059
		return cookie;
P
Peter Zijlstra 已提交
4060 4061 4062 4063 4064

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4065
	cookie = __lock_pin_lock(lock);
P
Peter Zijlstra 已提交
4066 4067
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
4068 4069

	return cookie;
P
Peter Zijlstra 已提交
4070 4071 4072
}
EXPORT_SYMBOL_GPL(lock_pin_lock);

4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090
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
4091 4092 4093 4094 4095 4096 4097 4098 4099 4100
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
4101
	__lock_unpin_lock(lock, cookie);
P
Peter Zijlstra 已提交
4102 4103 4104 4105 4106
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_unpin_lock);

P
Peter Zijlstra 已提交
4107 4108 4109 4110 4111 4112 4113 4114 4115 4116
#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;

4117
	pr_warn("\n");
4118 4119
	pr_warn("=================================\n");
	pr_warn("WARNING: bad contention detected!\n");
4120
	print_kernel_ident();
4121
	pr_warn("---------------------------------\n");
4122
	pr_warn("%s/%d is trying to contend lock (",
4123
		curr->comm, task_pid_nr(curr));
P
Peter Zijlstra 已提交
4124
	print_lockdep_cache(lock);
4125
	pr_cont(") at:\n");
P
Peter Zijlstra 已提交
4126
	print_ip_sym(ip);
4127 4128
	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
4129 4130
	lockdep_print_held_locks(curr);

4131
	pr_warn("\nstack backtrace:\n");
P
Peter Zijlstra 已提交
4132 4133 4134 4135 4136 4137 4138 4139 4140
	dump_stack();

	return 0;
}

static void
__lock_contended(struct lockdep_map *lock, unsigned long ip)
{
	struct task_struct *curr = current;
4141
	struct held_lock *hlock;
P
Peter Zijlstra 已提交
4142 4143
	struct lock_class_stats *stats;
	unsigned int depth;
P
Peter Zijlstra 已提交
4144
	int i, contention_point, contending_point;
P
Peter Zijlstra 已提交
4145 4146

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
4147 4148 4149 4150
	/*
	 * 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
4151 4152 4153
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return;

4154 4155 4156 4157
	hlock = find_held_lock(curr, lock, depth, &i);
	if (!hlock) {
		print_lock_contention_bug(curr, lock, ip);
		return;
P
Peter Zijlstra 已提交
4158 4159
	}

4160 4161 4162
	if (hlock->instance != lock)
		return;

4163
	hlock->waittime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
4164

P
Peter Zijlstra 已提交
4165 4166 4167
	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
4168

D
Dave Jones 已提交
4169
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
4170 4171 4172 4173
	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
4174 4175
	if (lock->cpu != smp_processor_id())
		stats->bounces[bounce_contended + !!hlock->read]++;
P
Peter Zijlstra 已提交
4176 4177 4178 4179
	put_lock_stats(stats);
}

static void
P
Peter Zijlstra 已提交
4180
__lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
4181 4182
{
	struct task_struct *curr = current;
4183
	struct held_lock *hlock;
P
Peter Zijlstra 已提交
4184 4185
	struct lock_class_stats *stats;
	unsigned int depth;
4186
	u64 now, waittime = 0;
P
Peter Zijlstra 已提交
4187
	int i, cpu;
P
Peter Zijlstra 已提交
4188 4189

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
4190 4191 4192 4193
	/*
	 * 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
4194 4195 4196
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return;

4197 4198 4199 4200
	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
4201 4202
	}

4203 4204 4205
	if (hlock->instance != lock)
		return;

P
Peter Zijlstra 已提交
4206 4207
	cpu = smp_processor_id();
	if (hlock->waittime_stamp) {
4208
		now = lockstat_clock();
P
Peter Zijlstra 已提交
4209 4210 4211
		waittime = now - hlock->waittime_stamp;
		hlock->holdtime_stamp = now;
	}
P
Peter Zijlstra 已提交
4212

4213
	trace_lock_acquired(lock, ip);
4214

D
Dave Jones 已提交
4215
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
4216 4217 4218 4219 4220 4221 4222 4223
	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
4224
	put_lock_stats(stats);
P
Peter Zijlstra 已提交
4225 4226

	lock->cpu = cpu;
P
Peter Zijlstra 已提交
4227
	lock->ip = ip;
P
Peter Zijlstra 已提交
4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242
}

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;
4243
	trace_lock_contended(lock, ip);
P
Peter Zijlstra 已提交
4244 4245 4246 4247 4248 4249
	__lock_contended(lock, ip);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_contended);

P
Peter Zijlstra 已提交
4250
void lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262
{
	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
4263
	__lock_acquired(lock, ip);
P
Peter Zijlstra 已提交
4264 4265 4266 4267 4268 4269
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquired);
#endif

I
Ingo Molnar 已提交
4270 4271 4272 4273 4274 4275 4276 4277
/*
 * Used by the testsuite, sanitize the validator state
 * after a simulated failure:
 */

void lockdep_reset(void)
{
	unsigned long flags;
4278
	int i;
I
Ingo Molnar 已提交
4279 4280 4281 4282 4283 4284 4285 4286 4287 4288

	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;
4289
	for (i = 0; i < CHAINHASH_SIZE; i++)
4290
		INIT_HLIST_HEAD(chainhash_table + i);
I
Ingo Molnar 已提交
4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308
	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:
	 */
4309
	hlist_del_rcu(&class->hash_entry);
I
Ingo Molnar 已提交
4310 4311
	list_del_rcu(&class->lock_entry);

4312 4313
	RCU_INIT_POINTER(class->key, NULL);
	RCU_INIT_POINTER(class->name, NULL);
I
Ingo Molnar 已提交
4314 4315
}

4316
static inline int within(const void *addr, void *start, unsigned long size)
I
Ingo Molnar 已提交
4317 4318 4319 4320
{
	return addr >= start && addr < start + size;
}

4321 4322 4323 4324 4325 4326 4327 4328
/*
 * 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
4329 4330
void lockdep_free_key_range(void *start, unsigned long size)
{
4331
	struct lock_class *class;
4332
	struct hlist_head *head;
I
Ingo Molnar 已提交
4333 4334
	unsigned long flags;
	int i;
4335
	int locked;
I
Ingo Molnar 已提交
4336 4337

	raw_local_irq_save(flags);
4338
	locked = graph_lock();
I
Ingo Molnar 已提交
4339 4340 4341 4342 4343 4344

	/*
	 * Unhash all classes that were created by this module:
	 */
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
4345
		hlist_for_each_entry_rcu(class, head, hash_entry) {
I
Ingo Molnar 已提交
4346 4347
			if (within(class->key, start, size))
				zap_class(class);
4348 4349 4350
			else if (within(class->name, start, size))
				zap_class(class);
		}
I
Ingo Molnar 已提交
4351 4352
	}

4353 4354
	if (locked)
		graph_unlock();
I
Ingo Molnar 已提交
4355
	raw_local_irq_restore(flags);
4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369

	/*
	 * 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
4370 4371 4372 4373
}

void lockdep_reset_lock(struct lockdep_map *lock)
{
4374
	struct lock_class *class;
4375
	struct hlist_head *head;
I
Ingo Molnar 已提交
4376 4377
	unsigned long flags;
	int i, j;
4378
	int locked;
I
Ingo Molnar 已提交
4379 4380 4381 4382

	raw_local_irq_save(flags);

	/*
4383 4384 4385 4386 4387 4388 4389
	 * 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);
4390
		if (!IS_ERR_OR_NULL(class))
4391 4392 4393 4394 4395
			zap_class(class);
	}
	/*
	 * Debug check: in the end all mapped classes should
	 * be gone.
I
Ingo Molnar 已提交
4396
	 */
4397
	locked = graph_lock();
I
Ingo Molnar 已提交
4398 4399
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
4400
		hlist_for_each_entry_rcu(class, head, hash_entry) {
4401 4402 4403 4404 4405 4406
			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
4407 4408 4409 4410
				if (debug_locks_off_graph_unlock()) {
					/*
					 * We all just reset everything, how did it match?
					 */
4411
					WARN_ON(1);
P
Peter Zijlstra 已提交
4412
				}
4413
				goto out_restore;
I
Ingo Molnar 已提交
4414 4415 4416
			}
		}
	}
4417 4418
	if (locked)
		graph_unlock();
4419 4420

out_restore:
I
Ingo Molnar 已提交
4421 4422 4423 4424 4425 4426 4427
	raw_local_irq_restore(flags);
}

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

4428
	printk("... MAX_LOCKDEP_SUBCLASSES:  %lu\n", MAX_LOCKDEP_SUBCLASSES);
I
Ingo Molnar 已提交
4429 4430
	printk("... MAX_LOCK_DEPTH:          %lu\n", MAX_LOCK_DEPTH);
	printk("... MAX_LOCKDEP_KEYS:        %lu\n", MAX_LOCKDEP_KEYS);
4431
	printk("... CLASSHASH_SIZE:          %lu\n", CLASSHASH_SIZE);
I
Ingo Molnar 已提交
4432 4433 4434 4435 4436 4437 4438 4439 4440
	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
4441
		sizeof(struct list_head) * CHAINHASH_SIZE
4442
#ifdef CONFIG_PROVE_LOCKING
4443
		+ sizeof(struct circular_queue)
4444
#endif
M
Ming Lei 已提交
4445
		) / 1024
4446
		);
I
Ingo Molnar 已提交
4447 4448 4449 4450 4451 4452 4453

	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,
4454
		     const void *mem_to, struct held_lock *hlock)
I
Ingo Molnar 已提交
4455 4456 4457 4458 4459 4460
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

4461
	pr_warn("\n");
4462 4463
	pr_warn("=========================\n");
	pr_warn("WARNING: held lock freed!\n");
4464
	print_kernel_ident();
4465
	pr_warn("-------------------------\n");
4466
	pr_warn("%s/%d is freeing memory %p-%p, with a lock still held there!\n",
4467
		curr->comm, task_pid_nr(curr), mem_from, mem_to-1);
4468
	print_lock(hlock);
I
Ingo Molnar 已提交
4469 4470
	lockdep_print_held_locks(curr);

4471
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
4472 4473 4474
	dump_stack();
}

O
Oleg Nesterov 已提交
4475 4476 4477 4478 4479 4480 4481
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
4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500
/*
 * 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
4501 4502
		if (not_in_range(mem_from, mem_len, hlock->instance,
					sizeof(*hlock->instance)))
I
Ingo Molnar 已提交
4503 4504
			continue;

O
Oleg Nesterov 已提交
4505
		print_freed_lock_bug(curr, mem_from, mem_from + mem_len, hlock);
I
Ingo Molnar 已提交
4506 4507 4508 4509
		break;
	}
	local_irq_restore(flags);
}
4510
EXPORT_SYMBOL_GPL(debug_check_no_locks_freed);
I
Ingo Molnar 已提交
4511

4512
static void print_held_locks_bug(void)
I
Ingo Molnar 已提交
4513 4514 4515 4516 4517 4518
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

4519
	pr_warn("\n");
4520 4521
	pr_warn("====================================\n");
	pr_warn("WARNING: %s/%d still has locks held!\n",
4522
	       current->comm, task_pid_nr(current));
4523
	print_kernel_ident();
4524
	pr_warn("------------------------------------\n");
4525
	lockdep_print_held_locks(current);
4526
	pr_warn("\nstack backtrace:\n");
I
Ingo Molnar 已提交
4527 4528 4529
	dump_stack();
}

4530
void debug_check_no_locks_held(void)
I
Ingo Molnar 已提交
4531
{
4532 4533
	if (unlikely(current->lockdep_depth > 0))
		print_held_locks_bug();
I
Ingo Molnar 已提交
4534
}
4535
EXPORT_SYMBOL_GPL(debug_check_no_locks_held);
I
Ingo Molnar 已提交
4536

4537
#ifdef __KERNEL__
I
Ingo Molnar 已提交
4538 4539 4540 4541 4542 4543
void debug_show_all_locks(void)
{
	struct task_struct *g, *p;
	int count = 10;
	int unlock = 1;

4544
	if (unlikely(!debug_locks)) {
4545
		pr_warn("INFO: lockdep is turned off.\n");
4546 4547
		return;
	}
4548
	pr_warn("\nShowing all locks held in the system:\n");
I
Ingo Molnar 已提交
4549 4550 4551 4552 4553 4554 4555 4556 4557 4558

	/*
	 * 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)
4559
			pr_warn("hm, tasklist_lock locked, retrying... ");
I
Ingo Molnar 已提交
4560 4561
		if (count) {
			count--;
4562
			pr_cont(" #%d", 10-count);
I
Ingo Molnar 已提交
4563 4564 4565
			mdelay(200);
			goto retry;
		}
4566
		pr_cont(" ignoring it.\n");
I
Ingo Molnar 已提交
4567
		unlock = 0;
4568 4569
	} else {
		if (count != 10)
4570
			pr_cont(" locked it.\n");
I
Ingo Molnar 已提交
4571 4572 4573
	}

	do_each_thread(g, p) {
I
Ingo Molnar 已提交
4574 4575 4576 4577 4578 4579 4580
		/*
		 * 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
4581 4582 4583 4584 4585 4586 4587
		if (p->lockdep_depth)
			lockdep_print_held_locks(p);
		if (!unlock)
			if (read_trylock(&tasklist_lock))
				unlock = 1;
	} while_each_thread(g, p);

4588
	pr_warn("\n");
4589
	pr_warn("=============================================\n\n");
I
Ingo Molnar 已提交
4590 4591 4592 4593 4594

	if (unlock)
		read_unlock(&tasklist_lock);
}
EXPORT_SYMBOL_GPL(debug_show_all_locks);
4595
#endif
I
Ingo Molnar 已提交
4596

4597 4598 4599 4600
/*
 * Careful: only use this function if you are sure that
 * the task cannot run in parallel!
 */
4601
void debug_show_held_locks(struct task_struct *task)
I
Ingo Molnar 已提交
4602
{
4603 4604 4605 4606
	if (unlikely(!debug_locks)) {
		printk("INFO: lockdep is turned off.\n");
		return;
	}
I
Ingo Molnar 已提交
4607 4608 4609
	lockdep_print_held_locks(task);
}
EXPORT_SYMBOL_GPL(debug_show_held_locks);
P
Peter Zijlstra 已提交
4610

4611
asmlinkage __visible void lockdep_sys_exit(void)
P
Peter Zijlstra 已提交
4612 4613 4614 4615 4616 4617
{
	struct task_struct *curr = current;

	if (unlikely(curr->lockdep_depth)) {
		if (!debug_locks_off())
			return;
4618
		pr_warn("\n");
4619 4620
		pr_warn("================================================\n");
		pr_warn("WARNING: lock held when returning to user space!\n");
4621
		print_kernel_ident();
4622
		pr_warn("------------------------------------------------\n");
4623
		pr_warn("%s/%d is leaving the kernel with locks still held!\n",
P
Peter Zijlstra 已提交
4624 4625 4626
				curr->comm, curr->pid);
		lockdep_print_held_locks(curr);
	}
4627 4628 4629 4630 4631

	/*
	 * The lock history for each syscall should be independent. So wipe the
	 * slate clean on return to userspace.
	 */
4632
	lockdep_invariant_state(false);
P
Peter Zijlstra 已提交
4633
}
4634

4635
void lockdep_rcu_suspicious(const char *file, const int line, const char *s)
4636 4637 4638
{
	struct task_struct *curr = current;

4639
	/* Note: the following can be executed concurrently, so be careful. */
4640
	pr_warn("\n");
4641 4642
	pr_warn("=============================\n");
	pr_warn("WARNING: suspicious RCU usage\n");
4643
	print_kernel_ident();
4644
	pr_warn("-----------------------------\n");
4645 4646 4647
	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",
4648 4649
	       !rcu_lockdep_current_cpu_online()
			? "RCU used illegally from offline CPU!\n"
4650
			: !rcu_is_watching()
4651 4652 4653
				? "RCU used illegally from idle CPU!\n"
				: "",
	       rcu_scheduler_active, debug_locks);
4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672

	/*
	 * 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.
	 */
4673
	if (!rcu_is_watching())
4674
		pr_warn("RCU used illegally from extended quiescent state!\n");
4675

4676
	lockdep_print_held_locks(curr);
4677
	pr_warn("\nstack backtrace:\n");
4678 4679
	dump_stack();
}
4680
EXPORT_SYMBOL_GPL(lockdep_rcu_suspicious);
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 4710 4711 4712 4713

#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 */

4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724
/*
 * 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;
}

4725
/*
4726
 * Lock history stacks; we have 2 nested lock history stacks:
4727
 *
4728 4729
 *   HARD(IRQ)
 *   SOFT(IRQ)
4730
 *
4731 4732 4733 4734
 * 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.
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 4763 4764 4765 4766
 */

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'.
4767 4768 4769 4770 4771
 *
 * 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.
4772 4773
 *
 * This is true for many things; pretty much all kthreads fall into this
4774
 * pattern, where they have an invariant state and future completions do not
4775 4776 4777 4778 4779 4780 4781
 * 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.
4782 4783 4784 4785 4786 4787
 *
 * 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.
4788
 */
4789
void lockdep_invariant_state(bool force)
4790
{
4791 4792
	/*
	 * We call this at an invariant point, no current state, no history.
4793
	 * Verify the former, enforce the latter.
4794
	 */
4795 4796
	WARN_ON_ONCE(!force && current->lockdep_depth);
	invalidate_xhlock(&xhlock(current->xhlock_idx));
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 4838 4839 4840 4841
}

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.
4842
 *    2. Got invalidated.
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 4870 4871 4872 4873
 *
 * 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;
4874
	xhlock->hist_id = ++current->hist_id;
4875 4876 4877 4878

	xhlock->trace.nr_entries = 0;
	xhlock->trace.max_entries = MAX_XHLOCK_TRACE_ENTRIES;
	xhlock->trace.entries = xhlock->trace_entries;
4879 4880 4881 4882 4883 4884 4885 4886

	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;
	}
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 4921 4922 4923 4924
}

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;

4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941
	/*
	 * 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;

4942 4943 4944
	gen_id = (unsigned int)atomic_inc_return(&cross_gen_id);
	xlock->hlock = *hlock;
	xlock->hlock.gen_id = gen_id;
4945
unlock:
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 5032 5033 5034 5035
	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;
5036
	unsigned int prev_hist_id = xhlock(cur).hist_id;
5037 5038 5039 5040 5041
	unsigned int i;

	if (!graph_lock())
		return;

5042 5043 5044
	if (xlock->nr_acquire) {
		for (i = 0; i < MAX_XHLOCKS_NR; i++) {
			struct hist_lock *xhlock = &xhlock(cur - i);
5045

5046 5047
			if (!xhlock_valid(xhlock))
				break;
5048

5049 5050
			if (before(xhlock->hlock.gen_id, xlock->hlock.gen_id))
				break;
5051

5052 5053
			if (!same_context_xhlock(xhlock))
				break;
5054

5055
			/*
5056 5057 5058 5059
			 * 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:
5060
			 */
5061
			if (unlikely(before(prev_hist_id, xhlock->hist_id)))
5062
				break;
5063

5064
			prev_hist_id = xhlock->hist_id;
5065

5066 5067 5068 5069 5070 5071 5072
			/*
			 * commit_xhlock() returns 0 with graph_lock already
			 * released if fail.
			 */
			if (!commit_xhlock(xlock, xhlock))
				return;
		}
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 5112 5113 5114 5115
	}

	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);

/*
5116 5117
 * Return: 0 - failure;
 *         1 - crosslock, done;
5118 5119 5120 5121
 *         2 - normal lock, continue to held_lock[] ops.
 */
static int lock_release_crosslock(struct lockdep_map *lock)
{
5122 5123 5124 5125 5126 5127 5128 5129
	if (cross_lock(lock)) {
		if (!graph_lock())
			return 0;
		((struct lockdep_map_cross *)lock)->xlock.nr_acquire--;
		graph_unlock();
		return 1;
	}
	return 2;
5130 5131 5132 5133
}

static void cross_init(struct lockdep_map *lock, int cross)
{
5134 5135 5136
	if (cross)
		((struct lockdep_map_cross *)lock)->xlock.nr_acquire = 0;

5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150
	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;
5151
	task->hist_id = 0;
5152

5153
	for (i = 0; i < XHLOCK_CTX_NR; i++) {
5154
		task->xhlock_idx_hist[i] = UINT_MAX;
5155 5156
		task->hist_id_save[i] = 0;
	}
5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171

	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