lockdep.c 96.9 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 9
 *  Copyright (C) 2006,2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
 *  Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
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 31 32 33 34 35 36 37 38 39 40
#include <linux/mutex.h>
#include <linux/sched.h>
#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>
41
#include <linux/utsname.h>
P
Peter Zijlstra 已提交
42
#include <linux/hash.h>
43
#include <linux/ftrace.h>
P
Peter Zijlstra 已提交
44
#include <linux/stringify.h>
45
#include <linux/bitops.h>
46
#include <linux/gfp.h>
P
Peter Zijlstra 已提交
47

I
Ingo Molnar 已提交
48 49 50 51
#include <asm/sections.h>

#include "lockdep_internals.h"

52
#define CREATE_TRACE_POINTS
53
#include <trace/events/lock.h>
54

P
Peter Zijlstra 已提交
55 56 57 58 59 60 61 62 63 64 65 66 67 68
#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

I
Ingo Molnar 已提交
69
/*
70 71
 * lockdep_lock: protects the lockdep graph, the hashes and the
 *               class/list/hash allocators.
I
Ingo Molnar 已提交
72 73 74
 *
 * This is one of the rare exceptions where it's justified
 * to use a raw spinlock - we really dont want the spinlock
75
 * code to recurse back into the lockdep code...
I
Ingo Molnar 已提交
76
 */
77
static arch_spinlock_t lockdep_lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
78 79 80

static int graph_lock(void)
{
81
	arch_spin_lock(&lockdep_lock);
82 83 84 85 86 87 88
	/*
	 * 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) {
89
		arch_spin_unlock(&lockdep_lock);
90 91
		return 0;
	}
92 93
	/* prevent any recursions within lockdep from causing deadlocks */
	current->lockdep_recursion++;
94 95 96 97 98
	return 1;
}

static inline int graph_unlock(void)
{
99
	if (debug_locks && !arch_spin_is_locked(&lockdep_lock))
100 101
		return DEBUG_LOCKS_WARN_ON(1);

102
	current->lockdep_recursion--;
103
	arch_spin_unlock(&lockdep_lock);
104 105 106 107 108 109 110 111 112 113 114
	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();

115
	arch_spin_unlock(&lockdep_lock);
116 117 118

	return ret;
}
I
Ingo Molnar 已提交
119 120 121 122

static int lockdep_initialized;

unsigned long nr_list_entries;
P
Peter Zijlstra 已提交
123
static struct lock_list list_entries[MAX_LOCKDEP_ENTRIES];
I
Ingo Molnar 已提交
124 125 126 127 128 129 130 131 132 133

/*
 * 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
134 135 136 137 138 139 140 141 142
static inline struct lock_class *hlock_class(struct held_lock *hlock)
{
	if (!hlock->class_idx) {
		DEBUG_LOCKS_WARN_ON(1);
		return NULL;
	}
	return lock_classes + hlock->class_idx - 1;
}

P
Peter Zijlstra 已提交
143
#ifdef CONFIG_LOCK_STAT
144 145
static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS],
		      cpu_lock_stats);
P
Peter Zijlstra 已提交
146

147 148
static inline u64 lockstat_clock(void)
{
149
	return local_clock();
150 151
}

P
Peter Zijlstra 已提交
152
static int lock_point(unsigned long points[], unsigned long ip)
P
Peter Zijlstra 已提交
153 154 155
{
	int i;

P
Peter Zijlstra 已提交
156 157 158
	for (i = 0; i < LOCKSTAT_POINTS; i++) {
		if (points[i] == 0) {
			points[i] = ip;
P
Peter Zijlstra 已提交
159 160
			break;
		}
P
Peter Zijlstra 已提交
161
		if (points[i] == ip)
P
Peter Zijlstra 已提交
162 163 164 165 166 167
			break;
	}

	return i;
}

168
static void lock_time_inc(struct lock_time *lt, u64 time)
P
Peter Zijlstra 已提交
169 170 171 172
{
	if (time > lt->max)
		lt->max = time;

173
	if (time < lt->min || !lt->nr)
P
Peter Zijlstra 已提交
174 175 176 177 178 179
		lt->min = time;

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

180 181
static inline void lock_time_add(struct lock_time *src, struct lock_time *dst)
{
182 183 184 185 186 187 188 189 190
	if (!src->nr)
		return;

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

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

191 192 193 194 195 196 197 198 199 200 201 202
	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 =
203
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
204 205 206 207

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

P
Peter Zijlstra 已提交
208 209 210
		for (i = 0; i < ARRAY_SIZE(stats.contending_point); i++)
			stats.contending_point[i] += pcs->contending_point[i];

211 212 213 214 215
		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
216 217 218

		for (i = 0; i < ARRAY_SIZE(stats.bounces); i++)
			stats.bounces[i] += pcs->bounces[i];
219 220 221 222 223 224 225 226 227 228 229
	}

	return stats;
}

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

	for_each_possible_cpu(cpu) {
		struct lock_class_stats *cpu_stats =
230
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
231 232 233 234

		memset(cpu_stats, 0, sizeof(struct lock_class_stats));
	}
	memset(class->contention_point, 0, sizeof(class->contention_point));
P
Peter Zijlstra 已提交
235
	memset(class->contending_point, 0, sizeof(class->contending_point));
236 237
}

P
Peter Zijlstra 已提交
238 239
static struct lock_class_stats *get_lock_stats(struct lock_class *class)
{
240
	return &get_cpu_var(cpu_lock_stats)[class - lock_classes];
P
Peter Zijlstra 已提交
241 242 243 244
}

static void put_lock_stats(struct lock_class_stats *stats)
{
245
	put_cpu_var(cpu_lock_stats);
P
Peter Zijlstra 已提交
246 247 248 249 250
}

static void lock_release_holdtime(struct held_lock *hlock)
{
	struct lock_class_stats *stats;
251
	u64 holdtime;
P
Peter Zijlstra 已提交
252 253 254 255

	if (!lock_stat)
		return;

256
	holdtime = lockstat_clock() - hlock->holdtime_stamp;
P
Peter Zijlstra 已提交
257

D
Dave Jones 已提交
258
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
259 260 261 262 263 264 265 266 267 268 269 270
	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
271 272 273 274 275 276 277 278 279 280 281 282
/*
 * 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
283
#define __classhashfn(key)	hash_long((unsigned long)key, CLASSHASH_BITS)
I
Ingo Molnar 已提交
284 285 286 287 288 289 290 291 292 293
#define classhashentry(key)	(classhash_table + __classhashfn((key)))

static struct list_head classhash_table[CLASSHASH_SIZE];

/*
 * 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
294
#define __chainhashfn(chain)	hash_long(chain, CHAINHASH_BITS)
I
Ingo Molnar 已提交
295 296 297 298 299 300 301 302 303 304 305
#define chainhashentry(chain)	(chainhash_table + __chainhashfn((chain)))

static struct list_head chainhash_table[CHAINHASH_SIZE];

/*
 * 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.
 */
#define iterate_chain_key(key1, key2) \
306 307
	(((key1) << MAX_LOCKDEP_KEYS_BITS) ^ \
	((key1) >> (64-MAX_LOCKDEP_KEYS_BITS)) ^ \
I
Ingo Molnar 已提交
308 309
	(key2))

310
void lockdep_off(void)
I
Ingo Molnar 已提交
311 312 313 314 315
{
	current->lockdep_recursion++;
}
EXPORT_SYMBOL(lockdep_off);

316
void lockdep_on(void)
I
Ingo Molnar 已提交
317 318 319 320 321 322 323 324 325 326
{
	current->lockdep_recursion--;
}
EXPORT_SYMBOL(lockdep_on);

/*
 * Debugging switches:
 */

#define VERBOSE			0
327
#define VERY_VERBOSE		0
I
Ingo Molnar 已提交
328 329 330 331

#if VERBOSE
# define HARDIRQ_VERBOSE	1
# define SOFTIRQ_VERBOSE	1
332
# define RECLAIM_VERBOSE	1
I
Ingo Molnar 已提交
333 334 335
#else
# define HARDIRQ_VERBOSE	0
# define SOFTIRQ_VERBOSE	0
336
# define RECLAIM_VERBOSE	0
I
Ingo Molnar 已提交
337 338
#endif

339
#if VERBOSE || HARDIRQ_VERBOSE || SOFTIRQ_VERBOSE || RECLAIM_VERBOSE
I
Ingo Molnar 已提交
340 341 342 343 344
/*
 * Quick filtering for interesting events:
 */
static int class_filter(struct lock_class *class)
{
A
Andi Kleen 已提交
345 346
#if 0
	/* Example */
I
Ingo Molnar 已提交
347
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
348
			!strcmp(class->name, "lockname"))
I
Ingo Molnar 已提交
349 350
		return 1;
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
351
			!strcmp(class->name, "&struct->lockfield"))
I
Ingo Molnar 已提交
352
		return 1;
A
Andi Kleen 已提交
353
#endif
354 355
	/* Filter everything else. 1 would be to allow everything else */
	return 0;
I
Ingo Molnar 已提交
356 357 358 359 360 361 362 363 364 365 366 367 368
}
#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
369
 * addresses. Protected by the graph_lock.
I
Ingo Molnar 已提交
370 371 372 373 374 375 376 377 378 379
 */
unsigned long nr_stack_trace_entries;
static unsigned long stack_trace[MAX_STACK_TRACE_ENTRIES];

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;

380 381
	trace->skip = 3;

C
Christoph Hellwig 已提交
382
	save_stack_trace(trace);
I
Ingo Molnar 已提交
383

P
Peter Zijlstra 已提交
384 385 386 387 388 389 390
	/*
	 * 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>
	 */
391 392
	if (trace->nr_entries != 0 &&
	    trace->entries[trace->nr_entries-1] == ULONG_MAX)
P
Peter Zijlstra 已提交
393 394
		trace->nr_entries--;

I
Ingo Molnar 已提交
395 396 397 398
	trace->max_entries = trace->nr_entries;

	nr_stack_trace_entries += trace->nr_entries;

P
Peter Zijlstra 已提交
399
	if (nr_stack_trace_entries >= MAX_STACK_TRACE_ENTRIES-1) {
400 401 402 403 404 405 406
		if (!debug_locks_off_graph_unlock())
			return 0;

		printk("BUG: MAX_STACK_TRACE_ENTRIES too low!\n");
		printk("turning off the locking correctness validator.\n");
		dump_stack();

I
Ingo Molnar 已提交
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
		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
/*
 * We cannot printk in early bootup code. Not even early_printk()
 * might work. So we mark any initialization errors and printk
 * about it later on, in lockdep_info().
 */
static int lockdep_init_error;
425 426 427 428 429
static unsigned long lockdep_init_trace_data[20];
static struct stack_trace lockdep_init_trace = {
	.max_entries = ARRAY_SIZE(lockdep_init_trace_data),
	.entries = lockdep_init_trace_data,
};
I
Ingo Molnar 已提交
430 431 432 433

/*
 * Various lockdep statistics:
 */
434
DEFINE_PER_CPU(struct lockdep_stats, lockdep_stats);
I
Ingo Molnar 已提交
435 436 437 438 439 440
#endif

/*
 * Locking printouts:
 */

P
Peter Zijlstra 已提交
441
#define __USAGE(__STATE)						\
P
Peter Zijlstra 已提交
442 443 444 445
	[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
446

I
Ingo Molnar 已提交
447 448
static const char *usage_str[] =
{
P
Peter Zijlstra 已提交
449 450 451 452
#define LOCKDEP_STATE(__STATE) __USAGE(__STATE)
#include "lockdep_states.h"
#undef LOCKDEP_STATE
	[LOCK_USED] = "INITIAL USE",
I
Ingo Molnar 已提交
453 454 455 456
};

const char * __get_key_name(struct lockdep_subclass_key *key, char *str)
{
A
Alexey Dobriyan 已提交
457
	return kallsyms_lookup((unsigned long)key, NULL, NULL, NULL, str);
I
Ingo Molnar 已提交
458 459
}

460
static inline unsigned long lock_flag(enum lock_usage_bit bit)
I
Ingo Molnar 已提交
461
{
462 463
	return 1UL << bit;
}
I
Ingo Molnar 已提交
464

465 466 467 468 469 470 471 472 473 474
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
475 476
	}

477 478
	return c;
}
479

P
Peter Zijlstra 已提交
480
void get_usage_chars(struct lock_class *class, char usage[LOCK_USAGE_CHARS])
481
{
P
Peter Zijlstra 已提交
482
	int i = 0;
483

P
Peter Zijlstra 已提交
484 485 486 487 488 489 490
#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
491 492
}

493 494 495 496 497 498 499 500 501 502 503 504
static int __print_lock_name(struct lock_class *class)
{
	char str[KSYM_NAME_LEN];
	const char *name;

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

	return printk("%s", name);
}

I
Ingo Molnar 已提交
505 506
static void print_lock_name(struct lock_class *class)
{
P
Peter Zijlstra 已提交
507
	char str[KSYM_NAME_LEN], usage[LOCK_USAGE_CHARS];
I
Ingo Molnar 已提交
508 509
	const char *name;

P
Peter Zijlstra 已提交
510
	get_usage_chars(class, usage);
I
Ingo Molnar 已提交
511 512 513 514 515 516 517 518 519 520 521 522

	name = class->name;
	if (!name) {
		name = __get_key_name(class->key, str);
		printk(" (%s", name);
	} else {
		printk(" (%s", name);
		if (class->name_version > 1)
			printk("#%d", class->name_version);
		if (class->subclass)
			printk("/%d", class->subclass);
	}
P
Peter Zijlstra 已提交
523
	printk("){%s}", usage);
I
Ingo Molnar 已提交
524 525 526 527 528
}

static void print_lockdep_cache(struct lockdep_map *lock)
{
	const char *name;
529
	char str[KSYM_NAME_LEN];
I
Ingo Molnar 已提交
530 531 532 533 534 535 536 537 538 539

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

	printk("%s", name);
}

static void print_lock(struct held_lock *hlock)
{
D
Dave Jones 已提交
540
	print_lock_name(hlock_class(hlock));
I
Ingo Molnar 已提交
541 542 543 544 545 546 547 548 549
	printk(", at: ");
	print_ip_sym(hlock->acquire_ip);
}

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

	if (!depth) {
550
		printk("no locks held by %s/%d.\n", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
551 552 553
		return;
	}
	printk("%d lock%s held by %s/%d:\n",
554
		depth, depth > 1 ? "s" : "", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
555 556 557 558 559 560 561

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

P
Peter Zijlstra 已提交
562 563 564 565 566 567 568 569 570 571 572 573 574 575 576
static void print_kernel_version(void)
{
	printk("%s %.*s\n", init_utsname()->release,
		(int)strcspn(init_utsname()->version, " "),
		init_utsname()->version);
}

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

I
Ingo Molnar 已提交
577
/*
P
Peter Zijlstra 已提交
578
 * Is this the address of a static object:
I
Ingo Molnar 已提交
579
 */
P
Peter Zijlstra 已提交
580
static int static_obj(void *obj)
I
Ingo Molnar 已提交
581
{
P
Peter Zijlstra 已提交
582 583 584 585
	unsigned long start = (unsigned long) &_stext,
		      end   = (unsigned long) &_end,
		      addr  = (unsigned long) obj;

I
Ingo Molnar 已提交
586
	/*
P
Peter Zijlstra 已提交
587
	 * static variable?
I
Ingo Molnar 已提交
588
	 */
P
Peter Zijlstra 已提交
589 590
	if ((addr >= start) && (addr < end))
		return 1;
I
Ingo Molnar 已提交
591

592 593 594
	if (arch_is_kernel_data(addr))
		return 1;

I
Ingo Molnar 已提交
595
	/*
596
	 * in-kernel percpu var?
I
Ingo Molnar 已提交
597
	 */
598 599
	if (is_kernel_percpu_address(addr))
		return 1;
I
Ingo Molnar 已提交
600

P
Peter Zijlstra 已提交
601
	/*
602
	 * module static or percpu var?
P
Peter Zijlstra 已提交
603
	 */
604
	return is_module_address(addr) || is_module_percpu_address(addr);
605 606
}

I
Ingo Molnar 已提交
607
/*
P
Peter Zijlstra 已提交
608 609
 * To make lock name printouts unique, we calculate a unique
 * class->name_version generation counter:
I
Ingo Molnar 已提交
610
 */
P
Peter Zijlstra 已提交
611
static int count_matching_names(struct lock_class *new_class)
I
Ingo Molnar 已提交
612
{
P
Peter Zijlstra 已提交
613 614
	struct lock_class *class;
	int count = 0;
I
Ingo Molnar 已提交
615

P
Peter Zijlstra 已提交
616
	if (!new_class->name)
I
Ingo Molnar 已提交
617 618
		return 0;

P
Peter Zijlstra 已提交
619 620 621 622 623 624
	list_for_each_entry(class, &all_lock_classes, lock_entry) {
		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
625

P
Peter Zijlstra 已提交
626
	return count + 1;
I
Ingo Molnar 已提交
627 628
}

P
Peter Zijlstra 已提交
629 630 631 632 633 634 635
/*
 * 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
636
{
P
Peter Zijlstra 已提交
637 638 639
	struct lockdep_subclass_key *key;
	struct list_head *hash_head;
	struct lock_class *class;
I
Ingo Molnar 已提交
640

P
Peter Zijlstra 已提交
641 642 643 644 645 646 647 648 649
#ifdef CONFIG_DEBUG_LOCKDEP
	/*
	 * If the architecture calls into lockdep before initializing
	 * the hashes then we'll warn about it later. (we cannot printk
	 * right now)
	 */
	if (unlikely(!lockdep_initialized)) {
		lockdep_init();
		lockdep_init_error = 1;
650
		save_stack_trace(&lockdep_init_trace);
P
Peter Zijlstra 已提交
651 652
	}
#endif
I
Ingo Molnar 已提交
653

654 655 656 657 658 659 660 661 662 663
	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
664 665 666 667 668 669
	/*
	 * Static locks do not have their class-keys yet - for them the key
	 * is the lock object itself:
	 */
	if (unlikely(!lock->key))
		lock->key = (void *)lock;
I
Ingo Molnar 已提交
670

P
Peter Zijlstra 已提交
671 672 673 674 675 676
	/*
	 * 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
677 678
	BUILD_BUG_ON(sizeof(struct lock_class_key) >
			sizeof(struct lockdep_map));
I
Ingo Molnar 已提交
679

P
Peter Zijlstra 已提交
680
	key = lock->key->subkeys + subclass;
681

P
Peter Zijlstra 已提交
682
	hash_head = classhashentry(key);
683

P
Peter Zijlstra 已提交
684 685 686 687
	/*
	 * We can walk the hash lockfree, because the hash only
	 * grows, and we are careful when adding entries to the end:
	 */
P
Peter Zijlstra 已提交
688 689 690
	list_for_each_entry(class, hash_head, hash_entry) {
		if (class->key == key) {
			WARN_ON_ONCE(class->name != lock->name);
P
Peter Zijlstra 已提交
691
			return class;
P
Peter Zijlstra 已提交
692 693
		}
	}
I
Ingo Molnar 已提交
694

P
Peter Zijlstra 已提交
695
	return NULL;
I
Ingo Molnar 已提交
696 697 698
}

/*
P
Peter Zijlstra 已提交
699 700 701
 * 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
702
 */
P
Peter Zijlstra 已提交
703 704
static inline struct lock_class *
register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
I
Ingo Molnar 已提交
705
{
P
Peter Zijlstra 已提交
706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
	struct lockdep_subclass_key *key;
	struct list_head *hash_head;
	struct lock_class *class;
	unsigned long flags;

	class = look_up_lock_class(lock, subclass);
	if (likely(class))
		return class;

	/*
	 * Debug-check: all keys must be persistent!
 	 */
	if (!static_obj(lock->key)) {
		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);

	raw_local_irq_save(flags);
	if (!graph_lock()) {
		raw_local_irq_restore(flags);
		return NULL;
	}
	/*
	 * We have to do the hash-walk again, to avoid races
	 * with another CPU:
	 */
	list_for_each_entry(class, hash_head, hash_entry)
		if (class->key == key)
			goto out_unlock_set;
	/*
	 * 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()) {
			raw_local_irq_restore(flags);
			return NULL;
		}
		raw_local_irq_restore(flags);

		printk("BUG: MAX_LOCKDEP_KEYS too low!\n");
		printk("turning off the locking correctness validator.\n");
756
		dump_stack();
P
Peter Zijlstra 已提交
757 758 759
		return NULL;
	}
	class = lock_classes + nr_lock_classes++;
760
	debug_atomic_inc(nr_unused_locks);
P
Peter Zijlstra 已提交
761 762 763 764 765 766 767 768 769 770 771 772
	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:
	 */
	list_add_tail_rcu(&class->hash_entry, hash_head);
773 774 775 776
	/*
	 * Add it to the global list of classes:
	 */
	list_add_tail_rcu(&class->lock_entry, &all_lock_classes);
P
Peter Zijlstra 已提交
777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798

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

		printk("\nnew class %p: %s", class->key, class->name);
		if (class->name_version > 1)
			printk("#%d", class->name_version);
		printk("\n");
		dump_stack();

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

	if (!subclass || force)
799 800 801
		lock->class_cache[0] = class;
	else if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		lock->class_cache[subclass] = class;
P
Peter Zijlstra 已提交
802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821

	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;

		printk("BUG: MAX_LOCKDEP_ENTRIES too low!\n");
		printk("turning off the locking correctness validator.\n");
822
		dump_stack();
P
Peter Zijlstra 已提交
823 824 825 826 827 828 829 830 831
		return NULL;
	}
	return list_entries + nr_list_entries++;
}

/*
 * Add a new dependency to the head of the list:
 */
static int add_lock_to_list(struct lock_class *class, struct lock_class *this,
Y
Yong Zhang 已提交
832 833
			    struct list_head *head, unsigned long ip,
			    int distance, struct stack_trace *trace)
P
Peter Zijlstra 已提交
834 835 836 837 838 839 840 841 842 843
{
	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;

844 845
	entry->class = this;
	entry->distance = distance;
Y
Yong Zhang 已提交
846
	entry->trace = *trace;
P
Peter Zijlstra 已提交
847 848 849 850 851 852 853 854 855 856 857 858
	/*
	 * Since we never remove from the dependency list, the list can
	 * be walked lockless by other CPUs, it's only allocation
	 * that must be protected by the spinlock. But this also means
	 * we must make new entries visible only once writes to the
	 * entry become visible - hence the RCU op:
	 */
	list_add_tail_rcu(&entry->entry, head);

	return 1;
}

P
Peter Zijlstra 已提交
859 860 861
/*
 * For good efficiency of modular, we use power of 2
 */
P
Peter Zijlstra 已提交
862 863 864
#define MAX_CIRCULAR_QUEUE_SIZE		4096UL
#define CQ_MASK				(MAX_CIRCULAR_QUEUE_SIZE-1)

P
Peter Zijlstra 已提交
865 866
/*
 * The circular_queue and helpers is used to implement the
P
Peter Zijlstra 已提交
867 868 869
 * 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
870
 */
P
Peter Zijlstra 已提交
871 872 873 874 875 876 877
struct circular_queue {
	unsigned long element[MAX_CIRCULAR_QUEUE_SIZE];
	unsigned int  front, rear;
};

static struct circular_queue lock_cq;

878
unsigned int max_bfs_queue_depth;
P
Peter Zijlstra 已提交
879

880 881
static unsigned int lockdep_dependency_gen_id;

P
Peter Zijlstra 已提交
882 883 884
static inline void __cq_init(struct circular_queue *cq)
{
	cq->front = cq->rear = 0;
885
	lockdep_dependency_gen_id++;
P
Peter Zijlstra 已提交
886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
}

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
927

P
Peter Zijlstra 已提交
928 929 930
	nr = lock - list_entries;
	WARN_ON(nr >= nr_list_entries);
	lock->parent = parent;
931
	lock->class->dep_gen_id = lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
932 933 934 935 936
}

static inline unsigned long lock_accessed(struct lock_list *lock)
{
	unsigned long nr;
P
Peter Zijlstra 已提交
937

P
Peter Zijlstra 已提交
938 939
	nr = lock - list_entries;
	WARN_ON(nr >= nr_list_entries);
940
	return lock->class->dep_gen_id == lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959
}

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;
}

960
static int __bfs(struct lock_list *source_entry,
P
Peter Zijlstra 已提交
961 962 963 964
		 void *data,
		 int (*match)(struct lock_list *entry, void *data),
		 struct lock_list **target_entry,
		 int forward)
965 966
{
	struct lock_list *entry;
967
	struct list_head *head;
968 969 970
	struct circular_queue *cq = &lock_cq;
	int ret = 1;

971
	if (match(source_entry, data)) {
972 973 974 975 976
		*target_entry = source_entry;
		ret = 0;
		goto exit;
	}

977 978 979 980 981 982 983 984 985
	if (forward)
		head = &source_entry->class->locks_after;
	else
		head = &source_entry->class->locks_before;

	if (list_empty(head))
		goto exit;

	__cq_init(cq);
986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
	__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;

		list_for_each_entry(entry, head, entry) {
			if (!lock_accessed(entry)) {
1005
				unsigned int cq_depth;
1006
				mark_lock_accessed(entry, lock);
1007
				if (match(entry, data)) {
1008 1009 1010 1011 1012 1013 1014 1015 1016
					*target_entry = entry;
					ret = 0;
					goto exit;
				}

				if (__cq_enqueue(cq, (unsigned long)entry)) {
					ret = -1;
					goto exit;
				}
1017 1018 1019
				cq_depth = __cq_get_elem_count(cq);
				if (max_bfs_queue_depth < cq_depth)
					max_bfs_queue_depth = cq_depth;
1020 1021 1022 1023 1024 1025 1026
			}
		}
	}
exit:
	return ret;
}

1027
static inline int __bfs_forwards(struct lock_list *src_entry,
1028 1029 1030
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1031
{
1032
	return __bfs(src_entry, data, match, target_entry, 1);
1033 1034 1035

}

1036
static inline int __bfs_backwards(struct lock_list *src_entry,
1037 1038 1039
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1040
{
1041
	return __bfs(src_entry, data, match, target_entry, 0);
1042 1043 1044

}

P
Peter Zijlstra 已提交
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
/*
 * 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
1056
print_circular_bug_entry(struct lock_list *target, int depth)
P
Peter Zijlstra 已提交
1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
{
	if (debug_locks_silent)
		return 0;
	printk("\n-> #%u", depth);
	print_lock_name(target->class);
	printk(":\n");
	print_stack_trace(&target->trace, 6);

	return 0;
}

1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
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);
		printk(" --> ");
		__print_lock_name(parent);
		printk(" --> ");
		__print_lock_name(target);
		printk("\n\n");
	}

	printk(" Possible unsafe locking scenario:\n\n");
	printk("       CPU0                    CPU1\n");
	printk("       ----                    ----\n");
	printk("  lock(");
	__print_lock_name(target);
	printk(");\n");
	printk("                               lock(");
	__print_lock_name(parent);
	printk(");\n");
	printk("                               lock(");
	__print_lock_name(target);
	printk(");\n");
	printk("  lock(");
	__print_lock_name(source);
	printk(");\n");
	printk("\n *** DEADLOCK ***\n\n");
}

P
Peter Zijlstra 已提交
1118 1119 1120 1121 1122
/*
 * When a circular dependency is detected, print the
 * header first:
 */
static noinline int
1123 1124 1125
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
1126 1127 1128
{
	struct task_struct *curr = current;

1129
	if (debug_locks_silent)
P
Peter Zijlstra 已提交
1130 1131 1132 1133 1134 1135 1136
		return 0;

	printk("\n=======================================================\n");
	printk(  "[ INFO: possible circular locking dependency detected ]\n");
	print_kernel_version();
	printk(  "-------------------------------------------------------\n");
	printk("%s/%d is trying to acquire lock:\n",
1137
		curr->comm, task_pid_nr(curr));
1138
	print_lock(check_src);
P
Peter Zijlstra 已提交
1139
	printk("\nbut task is already holding lock:\n");
1140
	print_lock(check_tgt);
P
Peter Zijlstra 已提交
1141 1142 1143 1144 1145 1146 1147 1148
	printk("\nwhich lock already depends on the new lock.\n\n");
	printk("\nthe existing dependency chain (in reverse order) is:\n");

	print_circular_bug_entry(entry, depth);

	return 0;
}

1149 1150 1151 1152 1153
static inline int class_equal(struct lock_list *entry, void *data)
{
	return entry->class == data;
}

1154 1155 1156 1157
static noinline int print_circular_bug(struct lock_list *this,
				struct lock_list *target,
				struct held_lock *check_src,
				struct held_lock *check_tgt)
P
Peter Zijlstra 已提交
1158 1159
{
	struct task_struct *curr = current;
1160
	struct lock_list *parent;
1161
	struct lock_list *first_parent;
1162
	int depth;
P
Peter Zijlstra 已提交
1163

1164
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
P
Peter Zijlstra 已提交
1165 1166
		return 0;

1167
	if (!save_trace(&this->trace))
P
Peter Zijlstra 已提交
1168 1169
		return 0;

1170 1171
	depth = get_lock_depth(target);

1172
	print_circular_bug_header(target, depth, check_src, check_tgt);
1173 1174

	parent = get_lock_parent(target);
1175
	first_parent = parent;
1176 1177 1178 1179 1180

	while (parent) {
		print_circular_bug_entry(parent, --depth);
		parent = get_lock_parent(parent);
	}
P
Peter Zijlstra 已提交
1181 1182

	printk("\nother info that might help us debug this:\n\n");
1183 1184 1185
	print_circular_lock_scenario(check_src, check_tgt,
				     first_parent);

P
Peter Zijlstra 已提交
1186 1187 1188 1189 1190 1191 1192 1193
	lockdep_print_held_locks(curr);

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

	return 0;
}

1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
static noinline int print_bfs_bug(int ret)
{
	if (!debug_locks_off_graph_unlock())
		return 0;

	WARN(1, "lockdep bfs error:%d\n", ret);

	return 0;
}

1204
static int noop_count(struct lock_list *entry, void *data)
1205
{
1206 1207 1208
	(*(unsigned long *)data)++;
	return 0;
}
1209

1210 1211 1212 1213
unsigned long __lockdep_count_forward_deps(struct lock_list *this)
{
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1214

1215
	__bfs_forwards(this, (void *)&count, noop_count, &target_entry);
1216

1217
	return count;
1218 1219 1220 1221
}
unsigned long lockdep_count_forward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1222 1223 1224 1225
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1226 1227

	local_irq_save(flags);
1228
	arch_spin_lock(&lockdep_lock);
1229
	ret = __lockdep_count_forward_deps(&this);
1230
	arch_spin_unlock(&lockdep_lock);
1231 1232 1233 1234 1235
	local_irq_restore(flags);

	return ret;
}

1236
unsigned long __lockdep_count_backward_deps(struct lock_list *this)
1237
{
1238 1239
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1240

1241
	__bfs_backwards(this, (void *)&count, noop_count, &target_entry);
1242

1243
	return count;
1244 1245 1246 1247 1248
}

unsigned long lockdep_count_backward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1249 1250 1251 1252
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1253 1254

	local_irq_save(flags);
1255
	arch_spin_lock(&lockdep_lock);
1256
	ret = __lockdep_count_backward_deps(&this);
1257
	arch_spin_unlock(&lockdep_lock);
1258 1259 1260 1261 1262
	local_irq_restore(flags);

	return ret;
}

P
Peter Zijlstra 已提交
1263 1264 1265 1266 1267
/*
 * 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
1268 1269
check_noncircular(struct lock_list *root, struct lock_class *target,
		struct lock_list **target_entry)
P
Peter Zijlstra 已提交
1270
{
1271
	int result;
P
Peter Zijlstra 已提交
1272

1273
	debug_atomic_inc(nr_cyclic_checks);
1274

1275
	result = __bfs_forwards(root, target, class_equal, target_entry);
I
Ingo Molnar 已提交
1276

1277 1278
	return result;
}
1279

1280
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
1281 1282 1283 1284 1285 1286
/*
 * 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.
 */

1287 1288 1289 1290 1291 1292 1293
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
1294 1295
/*
 * Find a node in the forwards-direction dependency sub-graph starting
1296
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1297
 *
1298 1299
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1300
 *
1301 1302
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1303
 */
1304 1305 1306
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
1307
{
1308
	int result;
I
Ingo Molnar 已提交
1309

1310
	debug_atomic_inc(nr_find_usage_forwards_checks);
I
Ingo Molnar 已提交
1311

1312 1313 1314
	result = __bfs_forwards(root, (void *)bit, usage_match, target_entry);

	return result;
I
Ingo Molnar 已提交
1315 1316 1317 1318
}

/*
 * Find a node in the backwards-direction dependency sub-graph starting
1319
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1320
 *
1321 1322
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1323
 *
1324 1325
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1326
 */
1327 1328 1329
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
1330
{
1331
	int result;
I
Ingo Molnar 已提交
1332

1333
	debug_atomic_inc(nr_find_usage_backwards_checks);
I
Ingo Molnar 已提交
1334

1335
	result = __bfs_backwards(root, (void *)bit, usage_match, target_entry);
D
Dave Jones 已提交
1336

1337
	return result;
I
Ingo Molnar 已提交
1338 1339
}

P
Peter Zijlstra 已提交
1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
static void print_lock_class_header(struct lock_class *class, int depth)
{
	int bit;

	printk("%*s->", depth, "");
	print_lock_name(class);
	printk(" ops: %lu", class->ops);
	printk(" {\n");

	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]);
			len += printk(" at:\n");
			print_stack_trace(class->usage_traces + bit, len);
		}
	}
	printk("%*s }\n", depth, "");

	printk("%*s ... key      at: ",depth,"");
	print_ip_sym((unsigned long)class->key);
}

/*
 * 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)) {
1384
			printk("lockdep:%s bad path found in chain graph\n", __func__);
P
Peter Zijlstra 已提交
1385 1386 1387 1388 1389 1390 1391 1392 1393
			break;
		}

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

	return;
}
1394

1395 1396 1397
static void
print_irq_lock_scenario(struct lock_list *safe_entry,
			struct lock_list *unsafe_entry,
1398 1399
			struct lock_class *prev_class,
			struct lock_class *next_class)
1400 1401 1402
{
	struct lock_class *safe_class = safe_entry->class;
	struct lock_class *unsafe_class = unsafe_entry->class;
1403
	struct lock_class *middle_class = prev_class;
1404 1405

	if (middle_class == safe_class)
1406
		middle_class = next_class;
1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450

	/*
	 * 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);
		printk(" --> ");
		__print_lock_name(middle_class);
		printk(" --> ");
		__print_lock_name(unsafe_class);
		printk("\n\n");
	}

	printk(" Possible interrupt unsafe locking scenario:\n\n");
	printk("       CPU0                    CPU1\n");
	printk("       ----                    ----\n");
	printk("  lock(");
	__print_lock_name(unsafe_class);
	printk(");\n");
	printk("                               local_irq_disable();\n");
	printk("                               lock(");
	__print_lock_name(safe_class);
	printk(");\n");
	printk("                               lock(");
	__print_lock_name(middle_class);
	printk(");\n");
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(safe_class);
	printk(");\n");
	printk("\n *** DEADLOCK ***\n\n");
}

I
Ingo Molnar 已提交
1451 1452
static int
print_bad_irq_dependency(struct task_struct *curr,
1453 1454 1455 1456
			 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
1457 1458 1459 1460 1461 1462
			 struct held_lock *prev,
			 struct held_lock *next,
			 enum lock_usage_bit bit1,
			 enum lock_usage_bit bit2,
			 const char *irqclass)
{
1463
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1464 1465 1466 1467 1468
		return 0;

	printk("\n======================================================\n");
	printk(  "[ INFO: %s-safe -> %s-unsafe lock order detected ]\n",
		irqclass, irqclass);
1469
	print_kernel_version();
I
Ingo Molnar 已提交
1470 1471
	printk(  "------------------------------------------------------\n");
	printk("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] is trying to acquire:\n",
1472
		curr->comm, task_pid_nr(curr),
I
Ingo Molnar 已提交
1473 1474 1475 1476 1477 1478 1479 1480 1481
		curr->hardirq_context, hardirq_count() >> HARDIRQ_SHIFT,
		curr->softirq_context, softirq_count() >> SOFTIRQ_SHIFT,
		curr->hardirqs_enabled,
		curr->softirqs_enabled);
	print_lock(next);

	printk("\nand this task is already holding:\n");
	print_lock(prev);
	printk("which would create a new lock dependency:\n");
D
Dave Jones 已提交
1482
	print_lock_name(hlock_class(prev));
I
Ingo Molnar 已提交
1483
	printk(" ->");
D
Dave Jones 已提交
1484
	print_lock_name(hlock_class(next));
I
Ingo Molnar 已提交
1485 1486 1487 1488
	printk("\n");

	printk("\nbut this new dependency connects a %s-irq-safe lock:\n",
		irqclass);
1489
	print_lock_name(backwards_entry->class);
I
Ingo Molnar 已提交
1490 1491
	printk("\n... which became %s-irq-safe at:\n", irqclass);

1492
	print_stack_trace(backwards_entry->class->usage_traces + bit1, 1);
I
Ingo Molnar 已提交
1493 1494

	printk("\nto a %s-irq-unsafe lock:\n", irqclass);
1495
	print_lock_name(forwards_entry->class);
I
Ingo Molnar 已提交
1496 1497 1498
	printk("\n... which became %s-irq-unsafe at:\n", irqclass);
	printk("...");

1499
	print_stack_trace(forwards_entry->class->usage_traces + bit2, 1);
I
Ingo Molnar 已提交
1500 1501

	printk("\nother info that might help us debug this:\n\n");
1502 1503
	print_irq_lock_scenario(backwards_entry, forwards_entry,
				hlock_class(prev), hlock_class(next));
1504

I
Ingo Molnar 已提交
1505 1506
	lockdep_print_held_locks(curr);

1507 1508 1509 1510 1511
	printk("\nthe dependencies between %s-irq-safe lock", irqclass);
	printk(" and the holding lock:\n");
	if (!save_trace(&prev_root->trace))
		return 0;
	print_shortest_lock_dependencies(backwards_entry, prev_root);
I
Ingo Molnar 已提交
1512

1513 1514 1515 1516 1517
	printk("\nthe dependencies between the lock to be acquired");
	printk(" and %s-irq-unsafe lock:\n", irqclass);
	if (!save_trace(&next_root->trace))
		return 0;
	print_shortest_lock_dependencies(forwards_entry, next_root);
I
Ingo Molnar 已提交
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530

	printk("\nstack backtrace:\n");
	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;
1531
	struct lock_list this, that;
1532
	struct lock_list *uninitialized_var(target_entry);
1533
	struct lock_list *uninitialized_var(target_entry1);
1534 1535 1536 1537 1538

	this.parent = NULL;

	this.class = hlock_class(prev);
	ret = find_usage_backwards(&this, bit_backwards, &target_entry);
P
Peter Zijlstra 已提交
1539 1540 1541 1542
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
1543

1544 1545 1546
	that.parent = NULL;
	that.class = hlock_class(next);
	ret = find_usage_forwards(&that, bit_forwards, &target_entry1);
P
Peter Zijlstra 已提交
1547 1548 1549 1550
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
1551

1552 1553 1554
	return print_bad_irq_dependency(curr, &this, &that,
			target_entry, target_entry1,
			prev, next,
I
Ingo Molnar 已提交
1555 1556 1557
			bit_backwards, bit_forwards, irqclass);
}

1558 1559
static const char *state_names[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1560
	__stringify(__STATE),
1561 1562 1563 1564 1565 1566
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static const char *state_rnames[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1567
	__stringify(__STATE)"-READ",
1568 1569 1570 1571 1572
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static inline const char *state_name(enum lock_usage_bit bit)
P
Peter Zijlstra 已提交
1573
{
1574 1575
	return (bit & 1) ? state_rnames[bit >> 2] : state_names[bit >> 2];
}
P
Peter Zijlstra 已提交
1576

1577 1578
static int exclusive_bit(int new_bit)
{
P
Peter Zijlstra 已提交
1579
	/*
1580 1581 1582 1583 1584 1585 1586 1587
	 * 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
1588
	 */
1589 1590 1591

	int state = new_bit & ~3;
	int dir = new_bit & 2;
P
Peter Zijlstra 已提交
1592 1593

	/*
1594
	 * keep state, bit flip the direction and strip read.
P
Peter Zijlstra 已提交
1595
	 */
1596 1597 1598 1599 1600 1601
	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
1602
	/*
1603 1604
	 * 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
1605 1606 1607
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1608 1609
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
P
Peter Zijlstra 已提交
1610 1611
		return 0;

1612 1613
	bit++; /* _READ */

1614
	/*
1615 1616
	 * Prove that the new dependency does not connect a hardirq-safe-read
	 * lock with a hardirq-unsafe lock - to achieve this we search
1617 1618 1619
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1620 1621
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
1622 1623
		return 0;

1624 1625 1626 1627 1628 1629 1630 1631 1632
	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))	\
1633
		return 0;
1634 1635
#include "lockdep_states.h"
#undef LOCKDEP_STATE
1636

P
Peter Zijlstra 已提交
1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665
	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
1666 1667
#endif

1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687
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);
	printk(");\n");
	printk("  lock(");
	__print_lock_name(next);
	printk(");\n");
	printk("\n *** DEADLOCK ***\n\n");
	printk(" May be due to missing lock nesting notation\n\n");
}

I
Ingo Molnar 已提交
1688 1689 1690 1691
static int
print_deadlock_bug(struct task_struct *curr, struct held_lock *prev,
		   struct held_lock *next)
{
1692
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1693 1694 1695 1696
		return 0;

	printk("\n=============================================\n");
	printk(  "[ INFO: possible recursive locking detected ]\n");
1697
	print_kernel_version();
I
Ingo Molnar 已提交
1698 1699
	printk(  "---------------------------------------------\n");
	printk("%s/%d is trying to acquire lock:\n",
1700
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
1701 1702 1703 1704 1705
	print_lock(next);
	printk("\nbut task is already holding lock:\n");
	print_lock(prev);

	printk("\nother info that might help us debug this:\n");
1706
	print_deadlock_scenario(next, prev);
I
Ingo Molnar 已提交
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
	lockdep_print_held_locks(curr);

	printk("\nstack backtrace:\n");
	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
1728
	struct held_lock *nest = NULL;
I
Ingo Molnar 已提交
1729 1730 1731 1732
	int i;

	for (i = 0; i < curr->lockdep_depth; i++) {
		prev = curr->held_locks + i;
P
Peter Zijlstra 已提交
1733 1734 1735 1736

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

D
Dave Jones 已提交
1737
		if (hlock_class(prev) != hlock_class(next))
I
Ingo Molnar 已提交
1738
			continue;
P
Peter Zijlstra 已提交
1739

I
Ingo Molnar 已提交
1740 1741
		/*
		 * Allow read-after-read recursion of the same
1742
		 * lock class (i.e. read_lock(lock)+read_lock(lock)):
I
Ingo Molnar 已提交
1743
		 */
1744
		if ((read == 2) && prev->read)
I
Ingo Molnar 已提交
1745
			return 2;
P
Peter Zijlstra 已提交
1746 1747 1748 1749 1750 1751 1752 1753

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

I
Ingo Molnar 已提交
1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782
		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,
Y
Yong Zhang 已提交
1783
	       struct held_lock *next, int distance, int trylock_loop)
I
Ingo Molnar 已提交
1784 1785 1786
{
	struct lock_list *entry;
	int ret;
1787 1788
	struct lock_list this;
	struct lock_list *uninitialized_var(target_entry);
Y
Yong Zhang 已提交
1789 1790 1791 1792 1793 1794 1795 1796
	/*
	 * Static variable, serialized by the graph_lock().
	 *
	 * We use this static variable to save the stack trace in case
	 * we call into this function multiple times due to encountering
	 * trylocks in the held lock stack.
	 */
	static struct stack_trace trace;
I
Ingo Molnar 已提交
1797 1798 1799 1800 1801 1802 1803 1804 1805 1806

	/*
	 * 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:
	 */
1807 1808 1809 1810 1811 1812 1813
	this.class = hlock_class(next);
	this.parent = NULL;
	ret = check_noncircular(&this, hlock_class(prev), &target_entry);
	if (unlikely(!ret))
		return print_circular_bug(&this, target_entry, next, prev);
	else if (unlikely(ret < 0))
		return print_bfs_bug(ret);
1814

P
Peter Zijlstra 已提交
1815
	if (!check_prev_add_irq(curr, prev, next))
I
Ingo Molnar 已提交
1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
		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
1836 1837
	list_for_each_entry(entry, &hlock_class(prev)->locks_after, entry) {
		if (entry->class == hlock_class(next)) {
1838 1839
			if (distance == 1)
				entry->distance = 1;
I
Ingo Molnar 已提交
1840
			return 2;
1841
		}
I
Ingo Molnar 已提交
1842 1843
	}

Y
Yong Zhang 已提交
1844 1845 1846
	if (!trylock_loop && !save_trace(&trace))
		return 0;

I
Ingo Molnar 已提交
1847 1848 1849 1850
	/*
	 * Ok, all validations passed, add the new lock
	 * to the previous lock's dependency list:
	 */
D
Dave Jones 已提交
1851 1852
	ret = add_lock_to_list(hlock_class(prev), hlock_class(next),
			       &hlock_class(prev)->locks_after,
Y
Yong Zhang 已提交
1853
			       next->acquire_ip, distance, &trace);
1854

I
Ingo Molnar 已提交
1855 1856
	if (!ret)
		return 0;
1857

D
Dave Jones 已提交
1858 1859
	ret = add_lock_to_list(hlock_class(next), hlock_class(prev),
			       &hlock_class(next)->locks_before,
Y
Yong Zhang 已提交
1860
			       next->acquire_ip, distance, &trace);
1861 1862
	if (!ret)
		return 0;
I
Ingo Molnar 已提交
1863 1864

	/*
P
Peter Zijlstra 已提交
1865 1866
	 * Debugging printouts:
	 */
D
Dave Jones 已提交
1867
	if (verbose(hlock_class(prev)) || verbose(hlock_class(next))) {
P
Peter Zijlstra 已提交
1868 1869
		graph_unlock();
		printk("\n new dependency: ");
D
Dave Jones 已提交
1870
		print_lock_name(hlock_class(prev));
P
Peter Zijlstra 已提交
1871
		printk(" => ");
D
Dave Jones 已提交
1872
		print_lock_name(hlock_class(next));
P
Peter Zijlstra 已提交
1873
		printk("\n");
I
Ingo Molnar 已提交
1874
		dump_stack();
P
Peter Zijlstra 已提交
1875
		return graph_lock();
I
Ingo Molnar 已提交
1876
	}
P
Peter Zijlstra 已提交
1877 1878
	return 1;
}
I
Ingo Molnar 已提交
1879

P
Peter Zijlstra 已提交
1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
/*
 * 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;
Y
Yong Zhang 已提交
1890
	int trylock_loop = 0;
P
Peter Zijlstra 已提交
1891
	struct held_lock *hlock;
1892

I
Ingo Molnar 已提交
1893
	/*
P
Peter Zijlstra 已提交
1894 1895 1896
	 * Debugging checks.
	 *
	 * Depth must not be zero for a non-head lock:
I
Ingo Molnar 已提交
1897
	 */
P
Peter Zijlstra 已提交
1898 1899
	if (!depth)
		goto out_bug;
I
Ingo Molnar 已提交
1900
	/*
P
Peter Zijlstra 已提交
1901 1902
	 * At least two relevant locks must exist for this
	 * to be a head:
I
Ingo Molnar 已提交
1903
	 */
P
Peter Zijlstra 已提交
1904 1905 1906
	if (curr->held_locks[depth].irq_context !=
			curr->held_locks[depth-1].irq_context)
		goto out_bug;
1907

P
Peter Zijlstra 已提交
1908 1909 1910 1911 1912 1913 1914 1915
	for (;;) {
		int distance = curr->lockdep_depth - depth + 1;
		hlock = curr->held_locks + depth-1;
		/*
		 * Only non-recursive-read entries get new dependencies
		 * added:
		 */
		if (hlock->read != 2) {
Y
Yong Zhang 已提交
1916 1917
			if (!check_prev_add(curr, hlock, next,
						distance, trylock_loop))
P
Peter Zijlstra 已提交
1918 1919 1920 1921 1922 1923 1924 1925 1926
				return 0;
			/*
			 * 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;
1927
		}
P
Peter Zijlstra 已提交
1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939
		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;
Y
Yong Zhang 已提交
1940
		trylock_loop = 1;
I
Ingo Molnar 已提交
1941
	}
P
Peter Zijlstra 已提交
1942 1943 1944 1945
	return 1;
out_bug:
	if (!debug_locks_off_graph_unlock())
		return 0;
I
Ingo Molnar 已提交
1946

P
Peter Zijlstra 已提交
1947
	WARN_ON(1);
I
Ingo Molnar 已提交
1948

P
Peter Zijlstra 已提交
1949
	return 0;
I
Ingo Molnar 已提交
1950 1951
}

P
Peter Zijlstra 已提交
1952
unsigned long nr_lock_chains;
1953
struct lock_chain lock_chains[MAX_LOCKDEP_CHAINS];
1954
int nr_chain_hlocks;
1955 1956 1957 1958 1959 1960
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
1961

I
Ingo Molnar 已提交
1962 1963
/*
 * Look up a dependency chain. If the key is not present yet then
1964 1965 1966
 * 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.)
I
Ingo Molnar 已提交
1967
 */
1968 1969 1970
static inline int lookup_chain_cache(struct task_struct *curr,
				     struct held_lock *hlock,
				     u64 chain_key)
I
Ingo Molnar 已提交
1971
{
D
Dave Jones 已提交
1972
	struct lock_class *class = hlock_class(hlock);
I
Ingo Molnar 已提交
1973 1974
	struct list_head *hash_head = chainhashentry(chain_key);
	struct lock_chain *chain;
1975
	struct held_lock *hlock_curr, *hlock_next;
1976
	int i, j;
I
Ingo Molnar 已提交
1977

1978 1979
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;
I
Ingo Molnar 已提交
1980 1981 1982 1983 1984 1985 1986
	/*
	 * We can walk it lock-free, because entries only get added
	 * to the hash:
	 */
	list_for_each_entry(chain, hash_head, entry) {
		if (chain->chain_key == chain_key) {
cache_hit:
1987
			debug_atomic_inc(chain_lookup_hits);
1988
			if (very_verbose(class))
1989 1990 1991 1992
				printk("\nhash chain already cached, key: "
					"%016Lx tail class: [%p] %s\n",
					(unsigned long long)chain_key,
					class->key, class->name);
I
Ingo Molnar 已提交
1993 1994 1995
			return 0;
		}
	}
1996
	if (very_verbose(class))
1997 1998
		printk("\nnew hash chain, key: %016Lx tail class: [%p] %s\n",
			(unsigned long long)chain_key, class->key, class->name);
I
Ingo Molnar 已提交
1999 2000 2001 2002
	/*
	 * Allocate a new chain entry from the static array, and add
	 * it to the hash:
	 */
2003 2004
	if (!graph_lock())
		return 0;
I
Ingo Molnar 已提交
2005 2006 2007 2008 2009
	/*
	 * We have to walk the chain again locked - to avoid duplicates:
	 */
	list_for_each_entry(chain, hash_head, entry) {
		if (chain->chain_key == chain_key) {
2010
			graph_unlock();
I
Ingo Molnar 已提交
2011 2012 2013 2014
			goto cache_hit;
		}
	}
	if (unlikely(nr_lock_chains >= MAX_LOCKDEP_CHAINS)) {
2015 2016 2017
		if (!debug_locks_off_graph_unlock())
			return 0;

I
Ingo Molnar 已提交
2018 2019
		printk("BUG: MAX_LOCKDEP_CHAINS too low!\n");
		printk("turning off the locking correctness validator.\n");
2020
		dump_stack();
I
Ingo Molnar 已提交
2021 2022 2023 2024
		return 0;
	}
	chain = lock_chains + nr_lock_chains++;
	chain->chain_key = chain_key;
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
	chain->irq_context = hlock->irq_context;
	/* Find the first held_lock of current chain */
	hlock_next = hlock;
	for (i = curr->lockdep_depth - 1; i >= 0; i--) {
		hlock_curr = curr->held_locks + i;
		if (hlock_curr->irq_context != hlock_next->irq_context)
			break;
		hlock_next = hlock;
	}
	i++;
	chain->depth = curr->lockdep_depth + 1 - i;
2036 2037 2038
	if (likely(nr_chain_hlocks + chain->depth <= MAX_LOCKDEP_CHAIN_HLOCKS)) {
		chain->base = nr_chain_hlocks;
		nr_chain_hlocks += chain->depth;
2039
		for (j = 0; j < chain->depth - 1; j++, i++) {
D
Dave Jones 已提交
2040
			int lock_id = curr->held_locks[i].class_idx - 1;
2041 2042 2043 2044
			chain_hlocks[chain->base + j] = lock_id;
		}
		chain_hlocks[chain->base + j] = class - lock_classes;
	}
I
Ingo Molnar 已提交
2045
	list_add_tail_rcu(&chain->entry, hash_head);
2046
	debug_atomic_inc(chain_lookup_misses);
P
Peter Zijlstra 已提交
2047 2048 2049 2050 2051 2052
	inc_chains();

	return 1;
}

static int validate_chain(struct task_struct *curr, struct lockdep_map *lock,
2053
		struct held_lock *hlock, int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065
{
	/*
	 * 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.
	 * (If lookup_chain_cache() returns with 1 it acquires
	 * graph_lock for us)
	 */
	if (!hlock->trylock && (hlock->check == 2) &&
2066
	    lookup_chain_cache(curr, hlock, chain_key)) {
P
Peter Zijlstra 已提交
2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101
		/*
		 * 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:
		 */
		if (!chain_head && ret != 2)
			if (!check_prevs_add(curr, hlock))
				return 0;
		graph_unlock();
	} else
		/* after lookup_chain_cache(): */
		if (unlikely(!debug_locks))
			return 0;
I
Ingo Molnar 已提交
2102 2103 2104

	return 1;
}
P
Peter Zijlstra 已提交
2105 2106 2107
#else
static inline int validate_chain(struct task_struct *curr,
	       	struct lockdep_map *lock, struct held_lock *hlock,
2108
		int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2109 2110 2111
{
	return 1;
}
2112
#endif
I
Ingo Molnar 已提交
2113 2114 2115 2116 2117

/*
 * We are building curr_chain_key incrementally, so double-check
 * it from scratch, to make sure that it's done correctly:
 */
2118
static void check_chain_key(struct task_struct *curr)
I
Ingo Molnar 已提交
2119 2120 2121 2122 2123 2124 2125 2126 2127 2128
{
#ifdef CONFIG_DEBUG_LOCKDEP
	struct held_lock *hlock, *prev_hlock = NULL;
	unsigned int i, id;
	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();
2129
			WARN(1, "hm#1, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2130 2131 2132 2133 2134
				curr->lockdep_depth, i,
				(unsigned long long)chain_key,
				(unsigned long long)hlock->prev_chain_key);
			return;
		}
D
Dave Jones 已提交
2135
		id = hlock->class_idx - 1;
2136 2137 2138
		if (DEBUG_LOCKS_WARN_ON(id >= MAX_LOCKDEP_KEYS))
			return;

I
Ingo Molnar 已提交
2139 2140 2141 2142 2143 2144 2145 2146
		if (prev_hlock && (prev_hlock->irq_context !=
							hlock->irq_context))
			chain_key = 0;
		chain_key = iterate_chain_key(chain_key, id);
		prev_hlock = hlock;
	}
	if (chain_key != curr->curr_chain_key) {
		debug_locks_off();
2147
		WARN(1, "hm#2, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2148 2149 2150 2151 2152 2153 2154
			curr->lockdep_depth, i,
			(unsigned long long)chain_key,
			(unsigned long long)curr->curr_chain_key);
	}
#endif
}

2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172
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);
	printk(");\n");
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(class);
	printk(");\n");
	printk("\n *** DEADLOCK ***\n\n");
}

P
Peter Zijlstra 已提交
2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
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;

	printk("\n=================================\n");
	printk(  "[ INFO: inconsistent lock state ]\n");
	print_kernel_version();
	printk(  "---------------------------------\n");

	printk("inconsistent {%s} -> {%s} usage.\n",
		usage_str[prev_bit], usage_str[new_bit]);

	printk("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] takes:\n",
2189
		curr->comm, task_pid_nr(curr),
P
Peter Zijlstra 已提交
2190 2191 2192 2193 2194 2195 2196
		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);

	printk("{%s} state was registered at:\n", usage_str[prev_bit]);
D
Dave Jones 已提交
2197
	print_stack_trace(hlock_class(this)->usage_traces + prev_bit, 1);
P
Peter Zijlstra 已提交
2198 2199 2200

	print_irqtrace_events(curr);
	printk("\nother info that might help us debug this:\n");
2201 2202
	print_usage_bug_scenario(this);

P
Peter Zijlstra 已提交
2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217
	lockdep_print_held_locks(curr);

	printk("\nstack backtrace:\n");
	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
2218
	if (unlikely(hlock_class(this)->usage_mask & (1 << bad_bit)))
P
Peter Zijlstra 已提交
2219 2220 2221 2222 2223 2224 2225
		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);

2226
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
2227 2228 2229 2230 2231

/*
 * print irq inversion bug:
 */
static int
2232 2233
print_irq_inversion_bug(struct task_struct *curr,
			struct lock_list *root, struct lock_list *other,
I
Ingo Molnar 已提交
2234 2235 2236
			struct held_lock *this, int forwards,
			const char *irqclass)
{
2237 2238 2239 2240
	struct lock_list *entry = other;
	struct lock_list *middle = NULL;
	int depth;

2241
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
2242 2243 2244 2245
		return 0;

	printk("\n=========================================================\n");
	printk(  "[ INFO: possible irq lock inversion dependency detected ]\n");
2246
	print_kernel_version();
I
Ingo Molnar 已提交
2247 2248
	printk(  "---------------------------------------------------------\n");
	printk("%s/%d just changed the state of lock:\n",
2249
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
2250 2251
	print_lock(this);
	if (forwards)
2252
		printk("but this lock took another, %s-unsafe lock in the past:\n", irqclass);
I
Ingo Molnar 已提交
2253
	else
2254
		printk("but this lock was taken by another, %s-safe lock in the past:\n", irqclass);
2255
	print_lock_name(other->class);
I
Ingo Molnar 已提交
2256 2257 2258
	printk("\n\nand interrupts could create inverse lock ordering between them.\n\n");

	printk("\nother info that might help us debug this:\n");
2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277

	/* Find a middle lock (if one exists) */
	depth = get_lock_depth(other);
	do {
		if (depth == 0 && (entry != root)) {
			printk("lockdep:%s bad path found in chain graph\n", __func__);
			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
2278 2279
	lockdep_print_held_locks(curr);

2280 2281 2282 2283
	printk("\nthe shortest dependencies between 2nd lock and 1st lock:\n");
	if (!save_trace(&root->trace))
		return 0;
	print_shortest_lock_dependencies(other, root);
I
Ingo Molnar 已提交
2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299

	printk("\nstack backtrace:\n");
	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;
2300 2301
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2302

2303 2304 2305
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_forwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2306 2307 2308 2309
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2310

2311
	return print_irq_inversion_bug(curr, &root, target_entry,
2312
					this, 1, irqclass);
I
Ingo Molnar 已提交
2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323
}

/*
 * 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;
2324 2325
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2326

2327 2328 2329
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_backwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2330 2331 2332 2333
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2334

2335
	return print_irq_inversion_bug(curr, &root, target_entry,
2336
					this, 0, irqclass);
I
Ingo Molnar 已提交
2337 2338
}

2339
void print_irqtrace_events(struct task_struct *curr)
I
Ingo Molnar 已提交
2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351
{
	printk("irq event stamp: %u\n", curr->irq_events);
	printk("hardirqs last  enabled at (%u): ", curr->hardirq_enable_event);
	print_ip_sym(curr->hardirq_enable_ip);
	printk("hardirqs last disabled at (%u): ", curr->hardirq_disable_event);
	print_ip_sym(curr->hardirq_disable_ip);
	printk("softirqs last  enabled at (%u): ", curr->softirq_enable_event);
	print_ip_sym(curr->softirq_enable_ip);
	printk("softirqs last disabled at (%u): ", curr->softirq_disable_event);
	print_ip_sym(curr->softirq_disable_ip);
}

2352
static int HARDIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2353
{
P
Peter Zijlstra 已提交
2354 2355 2356
#if HARDIRQ_VERBOSE
	return class_filter(class);
#endif
I
Ingo Molnar 已提交
2357 2358 2359
	return 0;
}

2360
static int SOFTIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2361
{
P
Peter Zijlstra 已提交
2362 2363 2364 2365
#if SOFTIRQ_VERBOSE
	return class_filter(class);
#endif
	return 0;
I
Ingo Molnar 已提交
2366 2367
}

2368
static int RECLAIM_FS_verbose(struct lock_class *class)
2369 2370 2371 2372 2373 2374 2375
{
#if RECLAIM_VERBOSE
	return class_filter(class);
#endif
	return 0;
}

I
Ingo Molnar 已提交
2376 2377
#define STRICT_READ_CHECKS	1

2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
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);
}

2391 2392 2393
typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
			     enum lock_usage_bit bit, const char *name);

2394
static int
2395 2396
mark_lock_irq(struct task_struct *curr, struct held_lock *this,
		enum lock_usage_bit new_bit)
2397
{
2398
	int excl_bit = exclusive_bit(new_bit);
2399
	int read = new_bit & 1;
2400 2401
	int dir = new_bit & 2;

2402 2403 2404 2405 2406 2407 2408
	/*
	 * 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.
	 */
2409 2410
	check_usage_f usage = dir ?
		check_usage_backwards : check_usage_forwards;
2411

2412 2413 2414 2415
	/*
	 * Validate that this particular lock does not have conflicting
	 * usage states.
	 */
2416 2417
	if (!valid_state(curr, this, new_bit, excl_bit))
		return 0;
2418

2419 2420 2421 2422 2423
	/*
	 * Validate that the lock dependencies don't have conflicting usage
	 * states.
	 */
	if ((!read || !dir || STRICT_READ_CHECKS) &&
2424
			!usage(curr, this, excl_bit, state_name(new_bit & ~1)))
2425
		return 0;
2426

2427 2428 2429 2430 2431 2432 2433 2434
	/*
	 * Check for read in write conflicts
	 */
	if (!read) {
		if (!valid_state(curr, this, new_bit, excl_bit + 1))
			return 0;

		if (STRICT_READ_CHECKS &&
2435 2436
			!usage(curr, this, excl_bit + 1,
				state_name(new_bit + 1)))
2437 2438
			return 0;
	}
2439

2440
	if (state_verbose(new_bit, hlock_class(this)))
2441 2442 2443 2444 2445
		return 2;

	return 1;
}

2446
enum mark_type {
2447 2448 2449
#define LOCKDEP_STATE(__STATE)	__STATE,
#include "lockdep_states.h"
#undef LOCKDEP_STATE
2450 2451
};

I
Ingo Molnar 已提交
2452 2453 2454
/*
 * Mark all held locks with a usage bit:
 */
2455
static int
2456
mark_held_locks(struct task_struct *curr, enum mark_type mark)
I
Ingo Molnar 已提交
2457 2458 2459 2460 2461 2462 2463 2464
{
	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;

2465 2466 2467 2468 2469
		usage_bit = 2 + (mark << 2); /* ENABLED */
		if (hlock->read)
			usage_bit += 1; /* READ */

		BUG_ON(usage_bit >= LOCK_USAGE_STATES);
2470

P
Peter Zijlstra 已提交
2471
		if (hlock_class(hlock)->key == __lockdep_no_validate__.subkeys)
2472 2473
			continue;

2474
		if (!mark_lock(curr, hlock, usage_bit))
I
Ingo Molnar 已提交
2475 2476 2477 2478 2479 2480 2481 2482 2483
			return 0;
	}

	return 1;
}

/*
 * Hardirqs will be enabled:
 */
2484
static void __trace_hardirqs_on_caller(unsigned long ip)
I
Ingo Molnar 已提交
2485 2486 2487 2488 2489 2490 2491 2492 2493 2494
{
	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:
	 */
2495
	if (!mark_held_locks(curr, HARDIRQ))
I
Ingo Molnar 已提交
2496 2497 2498 2499 2500 2501 2502
		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)
2503
		if (!mark_held_locks(curr, SOFTIRQ))
I
Ingo Molnar 已提交
2504 2505
			return;

P
Peter Zijlstra 已提交
2506 2507
	curr->hardirq_enable_ip = ip;
	curr->hardirq_enable_event = ++curr->irq_events;
2508
	debug_atomic_inc(hardirqs_on_events);
P
Peter Zijlstra 已提交
2509
}
2510 2511 2512 2513 2514 2515 2516 2517

void trace_hardirqs_on_caller(unsigned long ip)
{
	time_hardirqs_on(CALLER_ADDR0, ip);

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

2518 2519 2520 2521 2522 2523 2524 2525 2526 2527
	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;
	}

2528 2529 2530
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

2531 2532 2533 2534 2535 2536
	if (DEBUG_LOCKS_WARN_ON(unlikely(early_boot_irqs_disabled)))
		return;

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

2537 2538 2539 2540
	current->lockdep_recursion = 1;
	__trace_hardirqs_on_caller(ip);
	current->lockdep_recursion = 0;
}
2541
EXPORT_SYMBOL(trace_hardirqs_on_caller);
P
Peter Zijlstra 已提交
2542

2543
void trace_hardirqs_on(void)
2544 2545 2546
{
	trace_hardirqs_on_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2547 2548 2549 2550 2551
EXPORT_SYMBOL(trace_hardirqs_on);

/*
 * Hardirqs were disabled:
 */
2552
void trace_hardirqs_off_caller(unsigned long ip)
P
Peter Zijlstra 已提交
2553 2554 2555
{
	struct task_struct *curr = current;

2556
	time_hardirqs_off(CALLER_ADDR0, ip);
2557

P
Peter Zijlstra 已提交
2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568
	if (unlikely(!debug_locks || current->lockdep_recursion))
		return;

	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->hardirqs_enabled) {
		/*
		 * We have done an ON -> OFF transition:
		 */
		curr->hardirqs_enabled = 0;
2569
		curr->hardirq_disable_ip = ip;
P
Peter Zijlstra 已提交
2570
		curr->hardirq_disable_event = ++curr->irq_events;
2571
		debug_atomic_inc(hardirqs_off_events);
P
Peter Zijlstra 已提交
2572
	} else
2573
		debug_atomic_inc(redundant_hardirqs_off);
P
Peter Zijlstra 已提交
2574
}
2575
EXPORT_SYMBOL(trace_hardirqs_off_caller);
P
Peter Zijlstra 已提交
2576

2577
void trace_hardirqs_off(void)
2578 2579 2580
{
	trace_hardirqs_off_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2581 2582 2583 2584 2585 2586 2587 2588 2589
EXPORT_SYMBOL(trace_hardirqs_off);

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

2590
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
2591 2592 2593 2594 2595 2596
		return;

	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->softirqs_enabled) {
2597
		debug_atomic_inc(redundant_softirqs_on);
P
Peter Zijlstra 已提交
2598 2599 2600
		return;
	}

2601
	current->lockdep_recursion = 1;
P
Peter Zijlstra 已提交
2602 2603 2604 2605 2606 2607
	/*
	 * We'll do an OFF -> ON transition:
	 */
	curr->softirqs_enabled = 1;
	curr->softirq_enable_ip = ip;
	curr->softirq_enable_event = ++curr->irq_events;
2608
	debug_atomic_inc(softirqs_on_events);
P
Peter Zijlstra 已提交
2609 2610 2611 2612 2613 2614
	/*
	 * 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)
2615
		mark_held_locks(curr, SOFTIRQ);
2616
	current->lockdep_recursion = 0;
P
Peter Zijlstra 已提交
2617 2618 2619 2620 2621 2622 2623 2624 2625
}

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

2626
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638
		return;

	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;
2639
		debug_atomic_inc(softirqs_off_events);
P
Peter Zijlstra 已提交
2640 2641
		DEBUG_LOCKS_WARN_ON(!softirq_count());
	} else
2642
		debug_atomic_inc(redundant_softirqs_off);
P
Peter Zijlstra 已提交
2643 2644
}

2645
static void __lockdep_trace_alloc(gfp_t gfp_mask, unsigned long flags)
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663
{
	struct task_struct *curr = current;

	if (unlikely(!debug_locks))
		return;

	/* no reclaim without waiting on it */
	if (!(gfp_mask & __GFP_WAIT))
		return;

	/* this guy won't enter reclaim */
	if ((curr->flags & PF_MEMALLOC) && !(gfp_mask & __GFP_NOMEMALLOC))
		return;

	/* We're only interested __GFP_FS allocations for now */
	if (!(gfp_mask & __GFP_FS))
		return;

2664
	if (DEBUG_LOCKS_WARN_ON(irqs_disabled_flags(flags)))
2665 2666 2667 2668 2669
		return;

	mark_held_locks(curr, RECLAIM_FS);
}

2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686
static void check_flags(unsigned long flags);

void lockdep_trace_alloc(gfp_t gfp_mask)
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
	__lockdep_trace_alloc(gfp_mask, flags);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}

P
Peter Zijlstra 已提交
2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714
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
2715
					LOCK_ENABLED_HARDIRQ_READ))
P
Peter Zijlstra 已提交
2716 2717 2718
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
2719
						LOCK_ENABLED_SOFTIRQ_READ))
P
Peter Zijlstra 已提交
2720 2721 2722
					return 0;
		} else {
			if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
2723
					LOCK_ENABLED_HARDIRQ))
P
Peter Zijlstra 已提交
2724 2725 2726
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
2727
						LOCK_ENABLED_SOFTIRQ))
P
Peter Zijlstra 已提交
2728 2729 2730 2731
					return 0;
		}
	}

2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
	/*
	 * We reuse the irq context infrastructure more broadly as a general
	 * context checking code. This tests GFP_FS recursion (a lock taken
	 * during reclaim for a GFP_FS allocation is held over a GFP_FS
	 * allocation).
	 */
	if (!hlock->trylock && (curr->lockdep_reclaim_gfp & __GFP_FS)) {
		if (hlock->read) {
			if (!mark_lock(curr, hlock, LOCK_USED_IN_RECLAIM_FS_READ))
					return 0;
		} else {
			if (!mark_lock(curr, hlock, LOCK_USED_IN_RECLAIM_FS))
					return 0;
		}
	}

P
Peter Zijlstra 已提交
2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773
	return 1;
}

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:
	 */
	hlock->irq_context = 2*(curr->hardirq_context ? 1 : 0) +
				curr->softirq_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
2774 2775
}

P
Peter Zijlstra 已提交
2776
#else
I
Ingo Molnar 已提交
2777

P
Peter Zijlstra 已提交
2778 2779 2780
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
2781
{
P
Peter Zijlstra 已提交
2782 2783 2784
	WARN_ON(1);
	return 1;
}
I
Ingo Molnar 已提交
2785

P
Peter Zijlstra 已提交
2786 2787 2788 2789 2790
static inline int mark_irqflags(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 1;
}
I
Ingo Molnar 已提交
2791

P
Peter Zijlstra 已提交
2792 2793 2794 2795
static inline int separate_irq_context(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 0;
I
Ingo Molnar 已提交
2796 2797
}

2798 2799 2800 2801
void lockdep_trace_alloc(gfp_t gfp_mask)
{
}

P
Peter Zijlstra 已提交
2802
#endif
I
Ingo Molnar 已提交
2803 2804

/*
P
Peter Zijlstra 已提交
2805
 * Mark a lock with a usage bit, and validate the state transition:
I
Ingo Molnar 已提交
2806
 */
2807
static int mark_lock(struct task_struct *curr, struct held_lock *this,
2808
			     enum lock_usage_bit new_bit)
I
Ingo Molnar 已提交
2809
{
P
Peter Zijlstra 已提交
2810
	unsigned int new_mask = 1 << new_bit, ret = 1;
I
Ingo Molnar 已提交
2811 2812

	/*
P
Peter Zijlstra 已提交
2813 2814
	 * If already set then do not dirty the cacheline,
	 * nor do any checks:
I
Ingo Molnar 已提交
2815
	 */
D
Dave Jones 已提交
2816
	if (likely(hlock_class(this)->usage_mask & new_mask))
P
Peter Zijlstra 已提交
2817 2818 2819 2820
		return 1;

	if (!graph_lock())
		return 0;
I
Ingo Molnar 已提交
2821
	/*
L
Lucas De Marchi 已提交
2822
	 * Make sure we didn't race:
I
Ingo Molnar 已提交
2823
	 */
D
Dave Jones 已提交
2824
	if (unlikely(hlock_class(this)->usage_mask & new_mask)) {
P
Peter Zijlstra 已提交
2825 2826 2827
		graph_unlock();
		return 1;
	}
I
Ingo Molnar 已提交
2828

D
Dave Jones 已提交
2829
	hlock_class(this)->usage_mask |= new_mask;
I
Ingo Molnar 已提交
2830

D
Dave Jones 已提交
2831
	if (!save_trace(hlock_class(this)->usage_traces + new_bit))
P
Peter Zijlstra 已提交
2832
		return 0;
I
Ingo Molnar 已提交
2833

P
Peter Zijlstra 已提交
2834
	switch (new_bit) {
P
Peter Zijlstra 已提交
2835 2836 2837 2838 2839 2840 2841
#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
2842 2843 2844 2845 2846
		ret = mark_lock_irq(curr, this, new_bit);
		if (!ret)
			return 0;
		break;
	case LOCK_USED:
2847
		debug_atomic_dec(nr_unused_locks);
P
Peter Zijlstra 已提交
2848 2849 2850 2851 2852 2853 2854
		break;
	default:
		if (!debug_locks_off_graph_unlock())
			return 0;
		WARN_ON(1);
		return 0;
	}
I
Ingo Molnar 已提交
2855

P
Peter Zijlstra 已提交
2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869
	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
2870 2871 2872 2873 2874

/*
 * Initialize a lock instance's lock-class mapping info:
 */
void lockdep_init_map(struct lockdep_map *lock, const char *name,
2875
		      struct lock_class_key *key, int subclass)
I
Ingo Molnar 已提交
2876
{
2877
	memset(lock, 0, sizeof(*lock));
2878

2879 2880 2881 2882 2883 2884
#ifdef CONFIG_LOCK_STAT
	lock->cpu = raw_smp_processor_id();
#endif

	if (DEBUG_LOCKS_WARN_ON(!name)) {
		lock->name = "NULL";
I
Ingo Molnar 已提交
2885
		return;
2886 2887 2888
	}

	lock->name = name;
I
Ingo Molnar 已提交
2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900

	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);
		DEBUG_LOCKS_WARN_ON(1);
		return;
	}
	lock->key = key;
2901 2902 2903 2904

	if (unlikely(!debug_locks))
		return;

2905 2906
	if (subclass)
		register_lock_class(lock, subclass, 1);
I
Ingo Molnar 已提交
2907 2908 2909
}
EXPORT_SYMBOL_GPL(lockdep_init_map);

2910 2911
struct lock_class_key __lockdep_no_validate__;

I
Ingo Molnar 已提交
2912 2913 2914 2915 2916 2917
/*
 * 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,
2918 2919
			  struct lockdep_map *nest_lock, unsigned long ip,
			  int references)
I
Ingo Molnar 已提交
2920 2921
{
	struct task_struct *curr = current;
2922
	struct lock_class *class = NULL;
I
Ingo Molnar 已提交
2923 2924 2925
	struct held_lock *hlock;
	unsigned int depth, id;
	int chain_head = 0;
2926
	int class_idx;
I
Ingo Molnar 已提交
2927 2928
	u64 chain_key;

P
Peter Zijlstra 已提交
2929 2930 2931
	if (!prove_locking)
		check = 1;

I
Ingo Molnar 已提交
2932 2933 2934 2935 2936 2937
	if (unlikely(!debug_locks))
		return 0;

	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;

2938 2939 2940
	if (lock->key == &__lockdep_no_validate__)
		check = 1;

2941 2942
	if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		class = lock->class_cache[subclass];
2943
	/*
2944
	 * Not cached?
2945
	 */
I
Ingo Molnar 已提交
2946
	if (unlikely(!class)) {
2947
		class = register_lock_class(lock, subclass, 0);
I
Ingo Molnar 已提交
2948 2949 2950
		if (!class)
			return 0;
	}
2951
	atomic_inc((atomic_t *)&class->ops);
I
Ingo Molnar 已提交
2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968
	if (very_verbose(class)) {
		printk("\nacquire class [%p] %s", class->key, class->name);
		if (class->name_version > 1)
			printk("#%d", class->name_version);
		printk("\n");
		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;
	if (DEBUG_LOCKS_WARN_ON(depth >= MAX_LOCK_DEPTH))
		return 0;

2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982
	class_idx = class - lock_classes + 1;

	if (depth) {
		hlock = curr->held_locks + depth - 1;
		if (hlock->class_idx == class_idx && nest_lock) {
			if (hlock->references)
				hlock->references++;
			else
				hlock->references = 2;

			return 1;
		}
	}

I
Ingo Molnar 已提交
2983
	hlock = curr->held_locks + depth;
D
Dave Jones 已提交
2984 2985
	if (DEBUG_LOCKS_WARN_ON(!class))
		return 0;
2986
	hlock->class_idx = class_idx;
I
Ingo Molnar 已提交
2987 2988
	hlock->acquire_ip = ip;
	hlock->instance = lock;
P
Peter Zijlstra 已提交
2989
	hlock->nest_lock = nest_lock;
I
Ingo Molnar 已提交
2990 2991 2992
	hlock->trylock = trylock;
	hlock->read = read;
	hlock->check = check;
2993
	hlock->hardirqs_off = !!hardirqs_off;
2994
	hlock->references = references;
P
Peter Zijlstra 已提交
2995 2996
#ifdef CONFIG_LOCK_STAT
	hlock->waittime_stamp = 0;
2997
	hlock->holdtime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
2998
#endif
I
Ingo Molnar 已提交
2999

P
Peter Zijlstra 已提交
3000 3001 3002
	if (check == 2 && !mark_irqflags(curr, hlock))
		return 0;

I
Ingo Molnar 已提交
3003
	/* mark it as used: */
3004
	if (!mark_lock(curr, hlock, LOCK_USED))
I
Ingo Molnar 已提交
3005
		return 0;
P
Peter Zijlstra 已提交
3006

I
Ingo Molnar 已提交
3007
	/*
3008
	 * Calculate the chain hash: it's the combined hash of all the
I
Ingo Molnar 已提交
3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028
	 * 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.
	 */
	id = class - lock_classes;
	if (DEBUG_LOCKS_WARN_ON(id >= MAX_LOCKDEP_KEYS))
		return 0;

	chain_key = curr->curr_chain_key;
	if (!depth) {
		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
3029 3030 3031
	if (separate_irq_context(curr, hlock)) {
		chain_key = 0;
		chain_head = 1;
I
Ingo Molnar 已提交
3032 3033 3034
	}
	chain_key = iterate_chain_key(chain_key, id);

3035
	if (!validate_chain(curr, lock, hlock, chain_head, chain_key))
P
Peter Zijlstra 已提交
3036
		return 0;
3037

3038
	curr->curr_chain_key = chain_key;
I
Ingo Molnar 已提交
3039 3040
	curr->lockdep_depth++;
	check_chain_key(curr);
3041 3042 3043 3044
#ifdef CONFIG_DEBUG_LOCKDEP
	if (unlikely(!debug_locks))
		return 0;
#endif
I
Ingo Molnar 已提交
3045 3046 3047 3048
	if (unlikely(curr->lockdep_depth >= MAX_LOCK_DEPTH)) {
		debug_locks_off();
		printk("BUG: MAX_LOCK_DEPTH too low!\n");
		printk("turning off the locking correctness validator.\n");
3049
		dump_stack();
I
Ingo Molnar 已提交
3050 3051
		return 0;
	}
3052

I
Ingo Molnar 已提交
3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071
	if (unlikely(curr->lockdep_depth > max_lockdep_depth))
		max_lockdep_depth = curr->lockdep_depth;

	return 1;
}

static int
print_unlock_inbalance_bug(struct task_struct *curr, struct lockdep_map *lock,
			   unsigned long ip)
{
	if (!debug_locks_off())
		return 0;
	if (debug_locks_silent)
		return 0;

	printk("\n=====================================\n");
	printk(  "[ BUG: bad unlock balance detected! ]\n");
	printk(  "-------------------------------------\n");
	printk("%s/%d is trying to release lock (",
3072
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102
	print_lockdep_cache(lock);
	printk(") at:\n");
	print_ip_sym(ip);
	printk("but there are no more locks to release!\n");
	printk("\nother info that might help us debug this:\n");
	lockdep_print_held_locks(curr);

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

	return 0;
}

/*
 * Common debugging checks for both nested and non-nested unlock:
 */
static int check_unlock(struct task_struct *curr, struct lockdep_map *lock,
			unsigned long ip)
{
	if (unlikely(!debug_locks))
		return 0;
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;

	if (curr->lockdep_depth <= 0)
		return print_unlock_inbalance_bug(curr, lock, ip);

	return 1;
}

3103 3104 3105 3106 3107 3108
static int match_held_lock(struct held_lock *hlock, struct lockdep_map *lock)
{
	if (hlock->instance == lock)
		return 1;

	if (hlock->references) {
3109
		struct lock_class *class = lock->class_cache[0];
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126

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

		if (DEBUG_LOCKS_WARN_ON(!class))
			return 0;

		if (DEBUG_LOCKS_WARN_ON(!hlock->nest_lock))
			return 0;

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

	return 0;
}

3127
static int
3128 3129 3130
__lock_set_class(struct lockdep_map *lock, const char *name,
		 struct lock_class_key *key, unsigned int subclass,
		 unsigned long ip)
3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149
{
	struct task_struct *curr = current;
	struct held_lock *hlock, *prev_hlock;
	struct lock_class *class;
	unsigned int depth;
	int i;

	depth = curr->lockdep_depth;
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return 0;

	prev_hlock = NULL;
	for (i = depth-1; i >= 0; i--) {
		hlock = curr->held_locks + i;
		/*
		 * We must not cross into another context:
		 */
		if (prev_hlock && prev_hlock->irq_context != hlock->irq_context)
			break;
3150
		if (match_held_lock(hlock, lock))
3151 3152 3153 3154 3155 3156
			goto found_it;
		prev_hlock = hlock;
	}
	return print_unlock_inbalance_bug(curr, lock, ip);

found_it:
3157
	lockdep_init_map(lock, name, key, 0);
3158
	class = register_lock_class(lock, subclass, 0);
D
Dave Jones 已提交
3159
	hlock->class_idx = class - lock_classes + 1;
3160 3161 3162 3163 3164 3165 3166

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

	for (; i < depth; i++) {
		hlock = curr->held_locks + i;
		if (!__lock_acquire(hlock->instance,
D
Dave Jones 已提交
3167
			hlock_class(hlock)->subclass, hlock->trylock,
3168
				hlock->read, hlock->check, hlock->hardirqs_off,
3169 3170
				hlock->nest_lock, hlock->acquire_ip,
				hlock->references))
3171 3172 3173 3174 3175 3176 3177 3178
			return 0;
	}

	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth))
		return 0;
	return 1;
}

I
Ingo Molnar 已提交
3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208
/*
 * Remove the lock to the list of currently held locks in a
 * potentially non-nested (out of order) manner. This is a
 * relatively rare operation, as all the unlock APIs default
 * to nested mode (which uses lock_release()):
 */
static int
lock_release_non_nested(struct task_struct *curr,
			struct lockdep_map *lock, unsigned long ip)
{
	struct held_lock *hlock, *prev_hlock;
	unsigned int depth;
	int i;

	/*
	 * Check whether the lock exists in the current stack
	 * of held locks:
	 */
	depth = curr->lockdep_depth;
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return 0;

	prev_hlock = NULL;
	for (i = depth-1; i >= 0; i--) {
		hlock = curr->held_locks + i;
		/*
		 * We must not cross into another context:
		 */
		if (prev_hlock && prev_hlock->irq_context != hlock->irq_context)
			break;
3209
		if (match_held_lock(hlock, lock))
I
Ingo Molnar 已提交
3210 3211 3212 3213 3214 3215
			goto found_it;
		prev_hlock = hlock;
	}
	return print_unlock_inbalance_bug(curr, lock, ip);

found_it:
3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229
	if (hlock->instance == lock)
		lock_release_holdtime(hlock);

	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
3230

I
Ingo Molnar 已提交
3231 3232 3233 3234 3235
	/*
	 * 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:
	 */
3236

I
Ingo Molnar 已提交
3237 3238 3239 3240 3241 3242
	curr->lockdep_depth = i;
	curr->curr_chain_key = hlock->prev_chain_key;

	for (i++; i < depth; i++) {
		hlock = curr->held_locks + i;
		if (!__lock_acquire(hlock->instance,
D
Dave Jones 已提交
3243
			hlock_class(hlock)->subclass, hlock->trylock,
I
Ingo Molnar 已提交
3244
				hlock->read, hlock->check, hlock->hardirqs_off,
3245 3246
				hlock->nest_lock, hlock->acquire_ip,
				hlock->references))
I
Ingo Molnar 已提交
3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275
			return 0;
	}

	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth - 1))
		return 0;
	return 1;
}

/*
 * 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()). This is done for unlocks that nest
 * perfectly. (i.e. the current top of the lock-stack is unlocked)
 */
static int lock_release_nested(struct task_struct *curr,
			       struct lockdep_map *lock, unsigned long ip)
{
	struct held_lock *hlock;
	unsigned int depth;

	/*
	 * Pop off the top of the lock stack:
	 */
	depth = curr->lockdep_depth - 1;
	hlock = curr->held_locks + depth;

	/*
	 * Is the unlock non-nested:
	 */
3276
	if (hlock->instance != lock || hlock->references)
I
Ingo Molnar 已提交
3277 3278 3279 3280 3281 3282 3283 3284
		return lock_release_non_nested(curr, lock, ip);
	curr->lockdep_depth--;

	if (DEBUG_LOCKS_WARN_ON(!depth && (hlock->prev_chain_key != 0)))
		return 0;

	curr->curr_chain_key = hlock->prev_chain_key;

P
Peter Zijlstra 已提交
3285 3286
	lock_release_holdtime(hlock);

I
Ingo Molnar 已提交
3287 3288
#ifdef CONFIG_DEBUG_LOCKDEP
	hlock->prev_chain_key = 0;
D
Dave Jones 已提交
3289
	hlock->class_idx = 0;
I
Ingo Molnar 已提交
3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320
	hlock->acquire_ip = 0;
	hlock->irq_context = 0;
#endif
	return 1;
}

/*
 * 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()). This is done for unlocks that nest
 * perfectly. (i.e. the current top of the lock-stack is unlocked)
 */
static void
__lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
{
	struct task_struct *curr = current;

	if (!check_unlock(curr, lock, ip))
		return;

	if (nested) {
		if (!lock_release_nested(curr, lock, ip))
			return;
	} else {
		if (!lock_release_non_nested(curr, lock, ip))
			return;
	}

	check_chain_key(curr);
}

3321 3322 3323 3324 3325 3326
static int __lock_is_held(struct lockdep_map *lock)
{
	struct task_struct *curr = current;
	int i;

	for (i = 0; i < curr->lockdep_depth; i++) {
3327 3328 3329
		struct held_lock *hlock = curr->held_locks + i;

		if (match_held_lock(hlock, lock))
3330 3331 3332 3333 3334 3335
			return 1;
	}

	return 0;
}

I
Ingo Molnar 已提交
3336 3337 3338
/*
 * Check whether we follow the irq-flags state precisely:
 */
3339
static void check_flags(unsigned long flags)
I
Ingo Molnar 已提交
3340
{
3341 3342
#if defined(CONFIG_PROVE_LOCKING) && defined(CONFIG_DEBUG_LOCKDEP) && \
    defined(CONFIG_TRACE_IRQFLAGS)
I
Ingo Molnar 已提交
3343 3344 3345
	if (!debug_locks)
		return;

3346 3347 3348 3349 3350 3351 3352 3353 3354
	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
3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372

	/*
	 * 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()) {
		if (softirq_count())
			DEBUG_LOCKS_WARN_ON(current->softirqs_enabled);
		else
			DEBUG_LOCKS_WARN_ON(!current->softirqs_enabled);
	}

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

3373 3374 3375
void lock_set_class(struct lockdep_map *lock, const char *name,
		    struct lock_class_key *key, unsigned int subclass,
		    unsigned long ip)
3376 3377 3378 3379 3380 3381 3382 3383 3384
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	current->lockdep_recursion = 1;
	check_flags(flags);
3385
	if (__lock_set_class(lock, name, key, subclass, ip))
3386 3387 3388 3389
		check_chain_key(current);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
3390
EXPORT_SYMBOL_GPL(lock_set_class);
3391

I
Ingo Molnar 已提交
3392 3393 3394 3395
/*
 * We are not always called with irqs disabled - do that here,
 * and also avoid lockdep recursion:
 */
3396
void lock_acquire(struct lockdep_map *lock, unsigned int subclass,
P
Peter Zijlstra 已提交
3397 3398
			  int trylock, int read, int check,
			  struct lockdep_map *nest_lock, unsigned long ip)
I
Ingo Molnar 已提交
3399 3400 3401 3402 3403 3404 3405 3406 3407 3408
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
3409
	trace_lock_acquire(lock, subclass, trylock, read, check, nest_lock, ip);
I
Ingo Molnar 已提交
3410
	__lock_acquire(lock, subclass, trylock, read, check,
3411
		       irqs_disabled_flags(flags), nest_lock, ip, 0);
I
Ingo Molnar 已提交
3412 3413 3414 3415 3416
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquire);

3417
void lock_release(struct lockdep_map *lock, int nested,
3418
			  unsigned long ip)
I
Ingo Molnar 已提交
3419 3420 3421 3422 3423 3424 3425 3426 3427
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
3428
	trace_lock_release(lock, ip);
I
Ingo Molnar 已提交
3429 3430 3431 3432 3433 3434
	__lock_release(lock, nested, ip);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_release);

3435 3436 3437 3438 3439 3440
int lock_is_held(struct lockdep_map *lock)
{
	unsigned long flags;
	int ret = 0;

	if (unlikely(current->lockdep_recursion))
3441
		return 1; /* avoid false negative lockdep_assert_held() */
3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
	ret = __lock_is_held(lock);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);

	return ret;
}
EXPORT_SYMBOL_GPL(lock_is_held);

3455 3456 3457 3458 3459 3460 3461 3462 3463 3464
void lockdep_set_current_reclaim_state(gfp_t gfp_mask)
{
	current->lockdep_reclaim_gfp = gfp_mask;
}

void lockdep_clear_current_reclaim_state(void)
{
	current->lockdep_reclaim_gfp = 0;
}

P
Peter Zijlstra 已提交
3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478
#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;

	printk("\n=================================\n");
	printk(  "[ BUG: bad contention detected! ]\n");
	printk(  "---------------------------------\n");
	printk("%s/%d is trying to contend lock (",
3479
		curr->comm, task_pid_nr(curr));
P
Peter Zijlstra 已提交
3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499
	print_lockdep_cache(lock);
	printk(") at:\n");
	print_ip_sym(ip);
	printk("but there are no locks held!\n");
	printk("\nother info that might help us debug this:\n");
	lockdep_print_held_locks(curr);

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

	return 0;
}

static void
__lock_contended(struct lockdep_map *lock, unsigned long ip)
{
	struct task_struct *curr = current;
	struct held_lock *hlock, *prev_hlock;
	struct lock_class_stats *stats;
	unsigned int depth;
P
Peter Zijlstra 已提交
3500
	int i, contention_point, contending_point;
P
Peter Zijlstra 已提交
3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513

	depth = curr->lockdep_depth;
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return;

	prev_hlock = NULL;
	for (i = depth-1; i >= 0; i--) {
		hlock = curr->held_locks + i;
		/*
		 * We must not cross into another context:
		 */
		if (prev_hlock && prev_hlock->irq_context != hlock->irq_context)
			break;
3514
		if (match_held_lock(hlock, lock))
P
Peter Zijlstra 已提交
3515 3516 3517 3518 3519 3520 3521
			goto found_it;
		prev_hlock = hlock;
	}
	print_lock_contention_bug(curr, lock, ip);
	return;

found_it:
3522 3523 3524
	if (hlock->instance != lock)
		return;

3525
	hlock->waittime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
3526

P
Peter Zijlstra 已提交
3527 3528 3529
	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
3530

D
Dave Jones 已提交
3531
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
3532 3533 3534 3535
	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
3536 3537
	if (lock->cpu != smp_processor_id())
		stats->bounces[bounce_contended + !!hlock->read]++;
P
Peter Zijlstra 已提交
3538 3539 3540 3541
	put_lock_stats(stats);
}

static void
P
Peter Zijlstra 已提交
3542
__lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
3543 3544 3545 3546 3547
{
	struct task_struct *curr = current;
	struct held_lock *hlock, *prev_hlock;
	struct lock_class_stats *stats;
	unsigned int depth;
3548
	u64 now, waittime = 0;
P
Peter Zijlstra 已提交
3549
	int i, cpu;
P
Peter Zijlstra 已提交
3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562

	depth = curr->lockdep_depth;
	if (DEBUG_LOCKS_WARN_ON(!depth))
		return;

	prev_hlock = NULL;
	for (i = depth-1; i >= 0; i--) {
		hlock = curr->held_locks + i;
		/*
		 * We must not cross into another context:
		 */
		if (prev_hlock && prev_hlock->irq_context != hlock->irq_context)
			break;
3563
		if (match_held_lock(hlock, lock))
P
Peter Zijlstra 已提交
3564 3565 3566 3567 3568 3569 3570
			goto found_it;
		prev_hlock = hlock;
	}
	print_lock_contention_bug(curr, lock, _RET_IP_);
	return;

found_it:
3571 3572 3573
	if (hlock->instance != lock)
		return;

P
Peter Zijlstra 已提交
3574 3575
	cpu = smp_processor_id();
	if (hlock->waittime_stamp) {
3576
		now = lockstat_clock();
P
Peter Zijlstra 已提交
3577 3578 3579
		waittime = now - hlock->waittime_stamp;
		hlock->holdtime_stamp = now;
	}
P
Peter Zijlstra 已提交
3580

3581
	trace_lock_acquired(lock, ip);
3582

D
Dave Jones 已提交
3583
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
3584 3585 3586 3587 3588 3589 3590 3591
	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
3592
	put_lock_stats(stats);
P
Peter Zijlstra 已提交
3593 3594

	lock->cpu = cpu;
P
Peter Zijlstra 已提交
3595
	lock->ip = ip;
P
Peter Zijlstra 已提交
3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610
}

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;
3611
	trace_lock_contended(lock, ip);
P
Peter Zijlstra 已提交
3612 3613 3614 3615 3616 3617
	__lock_contended(lock, ip);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_contended);

P
Peter Zijlstra 已提交
3618
void lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630
{
	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
3631
	__lock_acquired(lock, ip);
P
Peter Zijlstra 已提交
3632 3633 3634 3635 3636 3637
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquired);
#endif

I
Ingo Molnar 已提交
3638 3639 3640 3641 3642 3643 3644 3645
/*
 * Used by the testsuite, sanitize the validator state
 * after a simulated failure:
 */

void lockdep_reset(void)
{
	unsigned long flags;
3646
	int i;
I
Ingo Molnar 已提交
3647 3648 3649 3650 3651 3652 3653 3654 3655 3656

	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;
3657 3658
	for (i = 0; i < CHAINHASH_SIZE; i++)
		INIT_LIST_HEAD(chainhash_table + i);
I
Ingo Molnar 已提交
3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679
	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:
	 */
	list_del_rcu(&class->hash_entry);
	list_del_rcu(&class->lock_entry);

3680
	class->key = NULL;
I
Ingo Molnar 已提交
3681 3682
}

3683
static inline int within(const void *addr, void *start, unsigned long size)
I
Ingo Molnar 已提交
3684 3685 3686 3687 3688 3689 3690 3691 3692 3693
{
	return addr >= start && addr < start + size;
}

void lockdep_free_key_range(void *start, unsigned long size)
{
	struct lock_class *class, *next;
	struct list_head *head;
	unsigned long flags;
	int i;
3694
	int locked;
I
Ingo Molnar 已提交
3695 3696

	raw_local_irq_save(flags);
3697
	locked = graph_lock();
I
Ingo Molnar 已提交
3698 3699 3700 3701 3702 3703 3704 3705

	/*
	 * Unhash all classes that were created by this module:
	 */
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
		if (list_empty(head))
			continue;
3706
		list_for_each_entry_safe(class, next, head, hash_entry) {
I
Ingo Molnar 已提交
3707 3708
			if (within(class->key, start, size))
				zap_class(class);
3709 3710 3711
			else if (within(class->name, start, size))
				zap_class(class);
		}
I
Ingo Molnar 已提交
3712 3713
	}

3714 3715
	if (locked)
		graph_unlock();
I
Ingo Molnar 已提交
3716 3717 3718 3719 3720
	raw_local_irq_restore(flags);
}

void lockdep_reset_lock(struct lockdep_map *lock)
{
3721
	struct lock_class *class, *next;
I
Ingo Molnar 已提交
3722 3723 3724
	struct list_head *head;
	unsigned long flags;
	int i, j;
3725
	int locked;
I
Ingo Molnar 已提交
3726 3727 3728 3729

	raw_local_irq_save(flags);

	/*
3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742
	 * 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);
		if (class)
			zap_class(class);
	}
	/*
	 * Debug check: in the end all mapped classes should
	 * be gone.
I
Ingo Molnar 已提交
3743
	 */
3744
	locked = graph_lock();
I
Ingo Molnar 已提交
3745 3746 3747 3748 3749
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
		if (list_empty(head))
			continue;
		list_for_each_entry_safe(class, next, head, hash_entry) {
3750 3751 3752 3753 3754 3755
			int match = 0;

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

			if (unlikely(match)) {
3756 3757
				if (debug_locks_off_graph_unlock())
					WARN_ON(1);
3758
				goto out_restore;
I
Ingo Molnar 已提交
3759 3760 3761
			}
		}
	}
3762 3763
	if (locked)
		graph_unlock();
3764 3765

out_restore:
I
Ingo Molnar 已提交
3766 3767 3768
	raw_local_irq_restore(flags);
}

3769
void lockdep_init(void)
I
Ingo Molnar 已提交
3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794
{
	int i;

	/*
	 * Some architectures have their own start_kernel()
	 * code which calls lockdep_init(), while we also
	 * call lockdep_init() from the start_kernel() itself,
	 * and we want to initialize the hashes only once:
	 */
	if (lockdep_initialized)
		return;

	for (i = 0; i < CLASSHASH_SIZE; i++)
		INIT_LIST_HEAD(classhash_table + i);

	for (i = 0; i < CHAINHASH_SIZE; i++)
		INIT_LIST_HEAD(chainhash_table + i);

	lockdep_initialized = 1;
}

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

3795
	printk("... MAX_LOCKDEP_SUBCLASSES:  %lu\n", MAX_LOCKDEP_SUBCLASSES);
I
Ingo Molnar 已提交
3796 3797
	printk("... MAX_LOCK_DEPTH:          %lu\n", MAX_LOCK_DEPTH);
	printk("... MAX_LOCKDEP_KEYS:        %lu\n", MAX_LOCKDEP_KEYS);
3798
	printk("... CLASSHASH_SIZE:          %lu\n", CLASSHASH_SIZE);
I
Ingo Molnar 已提交
3799 3800 3801 3802 3803 3804 3805 3806 3807
	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
3808
		sizeof(struct list_head) * CHAINHASH_SIZE
3809
#ifdef CONFIG_PROVE_LOCKING
3810
		+ sizeof(struct circular_queue)
3811
#endif
M
Ming Lei 已提交
3812
		) / 1024
3813
		);
I
Ingo Molnar 已提交
3814 3815 3816 3817 3818

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

#ifdef CONFIG_DEBUG_LOCKDEP
3819 3820 3821 3822 3823
	if (lockdep_init_error) {
		printk("WARNING: lockdep init error! Arch code didn't call lockdep_init() early enough?\n");
		printk("Call stack leading to lockdep invocation was:\n");
		print_stack_trace(&lockdep_init_trace, 0);
	}
I
Ingo Molnar 已提交
3824 3825 3826 3827 3828
#endif
}

static void
print_freed_lock_bug(struct task_struct *curr, const void *mem_from,
3829
		     const void *mem_to, struct held_lock *hlock)
I
Ingo Molnar 已提交
3830 3831 3832 3833 3834 3835 3836 3837 3838 3839
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

	printk("\n=========================\n");
	printk(  "[ BUG: held lock freed! ]\n");
	printk(  "-------------------------\n");
	printk("%s/%d is freeing memory %p-%p, with a lock still held there!\n",
3840
		curr->comm, task_pid_nr(curr), mem_from, mem_to-1);
3841
	print_lock(hlock);
I
Ingo Molnar 已提交
3842 3843 3844 3845 3846 3847
	lockdep_print_held_locks(curr);

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

O
Oleg Nesterov 已提交
3848 3849 3850 3851 3852 3853 3854
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
3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873
/*
 * 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
3874 3875
		if (not_in_range(mem_from, mem_len, hlock->instance,
					sizeof(*hlock->instance)))
I
Ingo Molnar 已提交
3876 3877
			continue;

O
Oleg Nesterov 已提交
3878
		print_freed_lock_bug(curr, mem_from, mem_from + mem_len, hlock);
I
Ingo Molnar 已提交
3879 3880 3881 3882
		break;
	}
	local_irq_restore(flags);
}
3883
EXPORT_SYMBOL_GPL(debug_check_no_locks_freed);
I
Ingo Molnar 已提交
3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895

static void print_held_locks_bug(struct task_struct *curr)
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

	printk("\n=====================================\n");
	printk(  "[ BUG: lock held at task exit time! ]\n");
	printk(  "-------------------------------------\n");
	printk("%s/%d is exiting with locks still held!\n",
3896
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914
	lockdep_print_held_locks(curr);

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

void debug_check_no_locks_held(struct task_struct *task)
{
	if (unlikely(task->lockdep_depth > 0))
		print_held_locks_bug(task);
}

void debug_show_all_locks(void)
{
	struct task_struct *g, *p;
	int count = 10;
	int unlock = 1;

3915 3916 3917 3918
	if (unlikely(!debug_locks)) {
		printk("INFO: lockdep is turned off.\n");
		return;
	}
I
Ingo Molnar 已提交
3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938
	printk("\nShowing all locks held in the system:\n");

	/*
	 * 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)
			printk("hm, tasklist_lock locked, retrying... ");
		if (count) {
			count--;
			printk(" #%d", 10-count);
			mdelay(200);
			goto retry;
		}
		printk(" ignoring it.\n");
		unlock = 0;
3939 3940 3941
	} else {
		if (count != 10)
			printk(KERN_CONT " locked it.\n");
I
Ingo Molnar 已提交
3942 3943 3944
	}

	do_each_thread(g, p) {
I
Ingo Molnar 已提交
3945 3946 3947 3948 3949 3950 3951
		/*
		 * 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
3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966
		if (p->lockdep_depth)
			lockdep_print_held_locks(p);
		if (!unlock)
			if (read_trylock(&tasklist_lock))
				unlock = 1;
	} while_each_thread(g, p);

	printk("\n");
	printk("=============================================\n\n");

	if (unlock)
		read_unlock(&tasklist_lock);
}
EXPORT_SYMBOL_GPL(debug_show_all_locks);

3967 3968 3969 3970
/*
 * Careful: only use this function if you are sure that
 * the task cannot run in parallel!
 */
3971
void debug_show_held_locks(struct task_struct *task)
I
Ingo Molnar 已提交
3972
{
3973 3974 3975 3976
	if (unlikely(!debug_locks)) {
		printk("INFO: lockdep is turned off.\n");
		return;
	}
I
Ingo Molnar 已提交
3977 3978 3979
	lockdep_print_held_locks(task);
}
EXPORT_SYMBOL_GPL(debug_show_held_locks);
P
Peter Zijlstra 已提交
3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995

void lockdep_sys_exit(void)
{
	struct task_struct *curr = current;

	if (unlikely(curr->lockdep_depth)) {
		if (!debug_locks_off())
			return;
		printk("\n================================================\n");
		printk(  "[ BUG: lock held when returning to user space! ]\n");
		printk(  "------------------------------------------------\n");
		printk("%s/%d is leaving the kernel with locks still held!\n",
				curr->comm, curr->pid);
		lockdep_print_held_locks(curr);
	}
}
3996 3997 3998 3999 4000

void lockdep_rcu_dereference(const char *file, const int line)
{
	struct task_struct *curr = current;

4001
#ifndef CONFIG_PROVE_RCU_REPEATEDLY
4002 4003
	if (!debug_locks_off())
		return;
4004 4005
#endif /* #ifdef CONFIG_PROVE_RCU_REPEATEDLY */
	/* Note: the following can be executed concurrently, so be careful. */
4006 4007 4008
	printk("\n===================================================\n");
	printk(  "[ INFO: suspicious rcu_dereference_check() usage. ]\n");
	printk(  "---------------------------------------------------\n");
4009 4010 4011
	printk("%s:%d invoked rcu_dereference_check() without protection!\n",
			file, line);
	printk("\nother info that might help us debug this:\n\n");
4012
	printk("\nrcu_scheduler_active = %d, debug_locks = %d\n", rcu_scheduler_active, debug_locks);
4013 4014 4015 4016 4017
	lockdep_print_held_locks(curr);
	printk("\nstack backtrace:\n");
	dump_stack();
}
EXPORT_SYMBOL_GPL(lockdep_rcu_dereference);