lockdep.c 108.8 KB
Newer Older
I
Ingo Molnar 已提交
1 2 3 4 5 6 7
/*
 * kernel/lockdep.c
 *
 * Runtime locking correctness validator
 *
 * Started by Ingo Molnar:
 *
P
Peter Zijlstra 已提交
8
 *  Copyright (C) 2006,2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
9
 *  Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra
I
Ingo Molnar 已提交
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
 *
 * this code maps all the lock dependencies as they occur in a live kernel
 * and will warn about the following classes of locking bugs:
 *
 * - lock inversion scenarios
 * - circular lock dependencies
 * - hardirq/softirq safe/unsafe locking bugs
 *
 * Bugs are reported even if the current locking scenario does not cause
 * any deadlock at this point.
 *
 * I.e. if anytime in the past two locks were taken in a different order,
 * even if it happened for another task, even if those were different
 * locks (but of the same class as this lock), this code will detect it.
 *
 * Thanks to Arjan van de Ven for coming up with the initial idea of
 * mapping lock dependencies runtime.
 */
28
#define DISABLE_BRANCH_PROFILING
I
Ingo Molnar 已提交
29 30
#include <linux/mutex.h>
#include <linux/sched.h>
31
#include <linux/sched/clock.h>
32
#include <linux/sched/task.h>
I
Ingo Molnar 已提交
33 34 35 36 37 38 39 40 41 42
#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>
43
#include <linux/utsname.h>
P
Peter Zijlstra 已提交
44
#include <linux/hash.h>
45
#include <linux/ftrace.h>
P
Peter Zijlstra 已提交
46
#include <linux/stringify.h>
47
#include <linux/bitops.h>
48
#include <linux/gfp.h>
49
#include <linux/kmemcheck.h>
50
#include <linux/random.h>
51
#include <linux/jhash.h>
P
Peter Zijlstra 已提交
52

I
Ingo Molnar 已提交
53 54 55 56
#include <asm/sections.h>

#include "lockdep_internals.h"

57
#define CREATE_TRACE_POINTS
58
#include <trace/events/lock.h>
59

P
Peter Zijlstra 已提交
60 61 62 63 64 65 66 67 68 69 70 71 72 73
#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
74
/*
75 76
 * lockdep_lock: protects the lockdep graph, the hashes and the
 *               class/list/hash allocators.
I
Ingo Molnar 已提交
77 78 79
 *
 * This is one of the rare exceptions where it's justified
 * to use a raw spinlock - we really dont want the spinlock
80
 * code to recurse back into the lockdep code...
I
Ingo Molnar 已提交
81
 */
82
static arch_spinlock_t lockdep_lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
83 84 85

static int graph_lock(void)
{
86
	arch_spin_lock(&lockdep_lock);
87 88 89 90 91 92 93
	/*
	 * 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) {
94
		arch_spin_unlock(&lockdep_lock);
95 96
		return 0;
	}
97 98
	/* prevent any recursions within lockdep from causing deadlocks */
	current->lockdep_recursion++;
99 100 101 102 103
	return 1;
}

static inline int graph_unlock(void)
{
P
Peter Zijlstra 已提交
104 105 106 107 108
	if (debug_locks && !arch_spin_is_locked(&lockdep_lock)) {
		/*
		 * The lockdep graph lock isn't locked while we expect it to
		 * be, we're confused now, bye!
		 */
109
		return DEBUG_LOCKS_WARN_ON(1);
P
Peter Zijlstra 已提交
110
	}
111

112
	current->lockdep_recursion--;
113
	arch_spin_unlock(&lockdep_lock);
114 115 116 117 118 119 120 121 122 123 124
	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();

125
	arch_spin_unlock(&lockdep_lock);
126 127 128

	return ret;
}
I
Ingo Molnar 已提交
129 130

unsigned long nr_list_entries;
P
Peter Zijlstra 已提交
131
static struct lock_list list_entries[MAX_LOCKDEP_ENTRIES];
I
Ingo Molnar 已提交
132 133 134 135 136 137 138 139 140 141

/*
 * 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
142 143 144
static inline struct lock_class *hlock_class(struct held_lock *hlock)
{
	if (!hlock->class_idx) {
P
Peter Zijlstra 已提交
145 146 147
		/*
		 * Someone passed in garbage, we give up.
		 */
D
Dave Jones 已提交
148 149 150 151 152 153
		DEBUG_LOCKS_WARN_ON(1);
		return NULL;
	}
	return lock_classes + hlock->class_idx - 1;
}

P
Peter Zijlstra 已提交
154
#ifdef CONFIG_LOCK_STAT
155
static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
P
Peter Zijlstra 已提交
156

157 158
static inline u64 lockstat_clock(void)
{
159
	return local_clock();
160 161
}

P
Peter Zijlstra 已提交
162
static int lock_point(unsigned long points[], unsigned long ip)
P
Peter Zijlstra 已提交
163 164 165
{
	int i;

P
Peter Zijlstra 已提交
166 167 168
	for (i = 0; i < LOCKSTAT_POINTS; i++) {
		if (points[i] == 0) {
			points[i] = ip;
P
Peter Zijlstra 已提交
169 170
			break;
		}
P
Peter Zijlstra 已提交
171
		if (points[i] == ip)
P
Peter Zijlstra 已提交
172 173 174 175 176 177
			break;
	}

	return i;
}

178
static void lock_time_inc(struct lock_time *lt, u64 time)
P
Peter Zijlstra 已提交
179 180 181 182
{
	if (time > lt->max)
		lt->max = time;

183
	if (time < lt->min || !lt->nr)
P
Peter Zijlstra 已提交
184 185 186 187 188 189
		lt->min = time;

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

190 191
static inline void lock_time_add(struct lock_time *src, struct lock_time *dst)
{
192 193 194 195 196 197 198 199 200
	if (!src->nr)
		return;

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

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

201 202 203 204 205 206 207 208 209 210 211 212
	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 =
213
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
214 215 216 217

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

P
Peter Zijlstra 已提交
218 219 220
		for (i = 0; i < ARRAY_SIZE(stats.contending_point); i++)
			stats.contending_point[i] += pcs->contending_point[i];

221 222 223 224 225
		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
226 227 228

		for (i = 0; i < ARRAY_SIZE(stats.bounces); i++)
			stats.bounces[i] += pcs->bounces[i];
229 230 231 232 233 234 235 236 237 238 239
	}

	return stats;
}

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

	for_each_possible_cpu(cpu) {
		struct lock_class_stats *cpu_stats =
240
			&per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
241 242 243 244

		memset(cpu_stats, 0, sizeof(struct lock_class_stats));
	}
	memset(class->contention_point, 0, sizeof(class->contention_point));
P
Peter Zijlstra 已提交
245
	memset(class->contending_point, 0, sizeof(class->contending_point));
246 247
}

P
Peter Zijlstra 已提交
248 249
static struct lock_class_stats *get_lock_stats(struct lock_class *class)
{
250
	return &get_cpu_var(cpu_lock_stats)[class - lock_classes];
P
Peter Zijlstra 已提交
251 252 253 254
}

static void put_lock_stats(struct lock_class_stats *stats)
{
255
	put_cpu_var(cpu_lock_stats);
P
Peter Zijlstra 已提交
256 257 258 259 260
}

static void lock_release_holdtime(struct held_lock *hlock)
{
	struct lock_class_stats *stats;
261
	u64 holdtime;
P
Peter Zijlstra 已提交
262 263 264 265

	if (!lock_stat)
		return;

266
	holdtime = lockstat_clock() - hlock->holdtime_stamp;
P
Peter Zijlstra 已提交
267

D
Dave Jones 已提交
268
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
269 270 271 272 273 274 275 276 277 278 279 280
	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
281 282 283 284 285 286 287 288 289 290 291 292
/*
 * 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
293
#define __classhashfn(key)	hash_long((unsigned long)key, CLASSHASH_BITS)
I
Ingo Molnar 已提交
294 295
#define classhashentry(key)	(classhash_table + __classhashfn((key)))

296
static struct hlist_head classhash_table[CLASSHASH_SIZE];
I
Ingo Molnar 已提交
297 298 299 300 301 302 303

/*
 * 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
304
#define __chainhashfn(chain)	hash_long(chain, CHAINHASH_BITS)
I
Ingo Molnar 已提交
305 306
#define chainhashentry(chain)	(chainhash_table + __chainhashfn((chain)))

307
static struct hlist_head chainhash_table[CHAINHASH_SIZE];
I
Ingo Molnar 已提交
308 309 310 311 312 313 314

/*
 * 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.
 */
315 316 317 318 319 320 321 322
static inline u64 iterate_chain_key(u64 key, u32 idx)
{
	u32 k0 = key, k1 = key >> 32;

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

	return k0 | (u64)k1 << 32;
}
I
Ingo Molnar 已提交
323

324
void lockdep_off(void)
I
Ingo Molnar 已提交
325 326 327 328 329
{
	current->lockdep_recursion++;
}
EXPORT_SYMBOL(lockdep_off);

330
void lockdep_on(void)
I
Ingo Molnar 已提交
331 332 333 334 335 336 337 338 339 340
{
	current->lockdep_recursion--;
}
EXPORT_SYMBOL(lockdep_on);

/*
 * Debugging switches:
 */

#define VERBOSE			0
341
#define VERY_VERBOSE		0
I
Ingo Molnar 已提交
342 343 344 345

#if VERBOSE
# define HARDIRQ_VERBOSE	1
# define SOFTIRQ_VERBOSE	1
346
# define RECLAIM_VERBOSE	1
I
Ingo Molnar 已提交
347 348 349
#else
# define HARDIRQ_VERBOSE	0
# define SOFTIRQ_VERBOSE	0
350
# define RECLAIM_VERBOSE	0
I
Ingo Molnar 已提交
351 352
#endif

353
#if VERBOSE || HARDIRQ_VERBOSE || SOFTIRQ_VERBOSE || RECLAIM_VERBOSE
I
Ingo Molnar 已提交
354 355 356 357 358
/*
 * Quick filtering for interesting events:
 */
static int class_filter(struct lock_class *class)
{
A
Andi Kleen 已提交
359 360
#if 0
	/* Example */
I
Ingo Molnar 已提交
361
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
362
			!strcmp(class->name, "lockname"))
I
Ingo Molnar 已提交
363 364
		return 1;
	if (class->name_version == 1 &&
A
Andi Kleen 已提交
365
			!strcmp(class->name, "&struct->lockfield"))
I
Ingo Molnar 已提交
366
		return 1;
A
Andi Kleen 已提交
367
#endif
368 369
	/* Filter everything else. 1 would be to allow everything else */
	return 0;
I
Ingo Molnar 已提交
370 371 372 373 374 375 376 377 378 379 380 381 382
}
#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
383
 * addresses. Protected by the graph_lock.
I
Ingo Molnar 已提交
384 385 386 387
 */
unsigned long nr_stack_trace_entries;
static unsigned long stack_trace[MAX_STACK_TRACE_ENTRIES];

388 389 390 391
static void print_lockdep_off(const char *bug_msg)
{
	printk(KERN_DEBUG "%s\n", bug_msg);
	printk(KERN_DEBUG "turning off the locking correctness validator.\n");
392
#ifdef CONFIG_LOCK_STAT
393
	printk(KERN_DEBUG "Please attach the output of /proc/lock_stat to the bug report\n");
394
#endif
395 396
}

I
Ingo Molnar 已提交
397 398 399 400 401 402
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;

403 404
	trace->skip = 3;

C
Christoph Hellwig 已提交
405
	save_stack_trace(trace);
I
Ingo Molnar 已提交
406

P
Peter Zijlstra 已提交
407 408 409 410 411 412 413
	/*
	 * 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>
	 */
414 415
	if (trace->nr_entries != 0 &&
	    trace->entries[trace->nr_entries-1] == ULONG_MAX)
P
Peter Zijlstra 已提交
416 417
		trace->nr_entries--;

I
Ingo Molnar 已提交
418 419 420 421
	trace->max_entries = trace->nr_entries;

	nr_stack_trace_entries += trace->nr_entries;

P
Peter Zijlstra 已提交
422
	if (nr_stack_trace_entries >= MAX_STACK_TRACE_ENTRIES-1) {
423 424 425
		if (!debug_locks_off_graph_unlock())
			return 0;

426
		print_lockdep_off("BUG: MAX_STACK_TRACE_ENTRIES too low!");
427 428
		dump_stack();

I
Ingo Molnar 已提交
429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
		return 0;
	}

	return 1;
}

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

#ifdef CONFIG_DEBUG_LOCKDEP
/*
 * Various lockdep statistics:
 */
444
DEFINE_PER_CPU(struct lockdep_stats, lockdep_stats);
I
Ingo Molnar 已提交
445 446 447 448 449 450
#endif

/*
 * Locking printouts:
 */

P
Peter Zijlstra 已提交
451
#define __USAGE(__STATE)						\
P
Peter Zijlstra 已提交
452 453 454 455
	[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
456

I
Ingo Molnar 已提交
457 458
static const char *usage_str[] =
{
P
Peter Zijlstra 已提交
459 460 461 462
#define LOCKDEP_STATE(__STATE) __USAGE(__STATE)
#include "lockdep_states.h"
#undef LOCKDEP_STATE
	[LOCK_USED] = "INITIAL USE",
I
Ingo Molnar 已提交
463 464 465 466
};

const char * __get_key_name(struct lockdep_subclass_key *key, char *str)
{
A
Alexey Dobriyan 已提交
467
	return kallsyms_lookup((unsigned long)key, NULL, NULL, NULL, str);
I
Ingo Molnar 已提交
468 469
}

470
static inline unsigned long lock_flag(enum lock_usage_bit bit)
I
Ingo Molnar 已提交
471
{
472 473
	return 1UL << bit;
}
I
Ingo Molnar 已提交
474

475 476 477 478 479 480 481 482 483 484
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
485 486
	}

487 488
	return c;
}
489

P
Peter Zijlstra 已提交
490
void get_usage_chars(struct lock_class *class, char usage[LOCK_USAGE_CHARS])
491
{
P
Peter Zijlstra 已提交
492
	int i = 0;
493

P
Peter Zijlstra 已提交
494 495 496 497 498 499 500
#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
501 502
}

503
static void __print_lock_name(struct lock_class *class)
504 505 506 507
{
	char str[KSYM_NAME_LEN];
	const char *name;

I
Ingo Molnar 已提交
508 509 510
	name = class->name;
	if (!name) {
		name = __get_key_name(class->key, str);
D
Dmitry Vyukov 已提交
511
		printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
512
	} else {
D
Dmitry Vyukov 已提交
513
		printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
514
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
515
			printk(KERN_CONT "#%d", class->name_version);
I
Ingo Molnar 已提交
516
		if (class->subclass)
D
Dmitry Vyukov 已提交
517
			printk(KERN_CONT "/%d", class->subclass);
I
Ingo Molnar 已提交
518
	}
519 520 521 522 523 524 525 526
}

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

	get_usage_chars(class, usage);

D
Dmitry Vyukov 已提交
527
	printk(KERN_CONT " (");
528
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
529
	printk(KERN_CONT "){%s}", usage);
I
Ingo Molnar 已提交
530 531 532 533 534
}

static void print_lockdep_cache(struct lockdep_map *lock)
{
	const char *name;
535
	char str[KSYM_NAME_LEN];
I
Ingo Molnar 已提交
536 537 538 539 540

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

D
Dmitry Vyukov 已提交
541
	printk(KERN_CONT "%s", name);
I
Ingo Molnar 已提交
542 543 544 545
}

static void print_lock(struct held_lock *hlock)
{
546 547 548 549 550 551 552 553 554 555
	/*
	 * We can be called locklessly through debug_show_all_locks() so be
	 * extra careful, the hlock might have been released and cleared.
	 */
	unsigned int class_idx = hlock->class_idx;

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

	if (!class_idx || (class_idx - 1) >= MAX_LOCKDEP_KEYS) {
D
Dmitry Vyukov 已提交
556
		printk(KERN_CONT "<RELEASED>\n");
557 558 559 560
		return;
	}

	print_lock_name(lock_classes + class_idx - 1);
D
Dmitry Vyukov 已提交
561 562
	printk(KERN_CONT ", at: [<%p>] %pS\n",
		(void *)hlock->acquire_ip, (void *)hlock->acquire_ip);
I
Ingo Molnar 已提交
563 564 565 566 567 568 569
}

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

	if (!depth) {
570
		printk("no locks held by %s/%d.\n", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
571 572 573
		return;
	}
	printk("%d lock%s held by %s/%d:\n",
574
		depth, depth > 1 ? "s" : "", curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
575 576 577 578 579 580 581

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

582
static void print_kernel_ident(void)
P
Peter Zijlstra 已提交
583
{
584
	printk("%s %.*s %s\n", init_utsname()->release,
P
Peter Zijlstra 已提交
585
		(int)strcspn(init_utsname()->version, " "),
586 587
		init_utsname()->version,
		print_tainted());
P
Peter Zijlstra 已提交
588 589 590 591 592 593 594 595 596 597
}

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

I
Ingo Molnar 已提交
598
/*
P
Peter Zijlstra 已提交
599
 * Is this the address of a static object:
I
Ingo Molnar 已提交
600
 */
601
#ifdef __KERNEL__
P
Peter Zijlstra 已提交
602
static int static_obj(void *obj)
I
Ingo Molnar 已提交
603
{
P
Peter Zijlstra 已提交
604 605 606 607
	unsigned long start = (unsigned long) &_stext,
		      end   = (unsigned long) &_end,
		      addr  = (unsigned long) obj;

I
Ingo Molnar 已提交
608
	/*
P
Peter Zijlstra 已提交
609
	 * static variable?
I
Ingo Molnar 已提交
610
	 */
P
Peter Zijlstra 已提交
611 612
	if ((addr >= start) && (addr < end))
		return 1;
I
Ingo Molnar 已提交
613

614 615 616
	if (arch_is_kernel_data(addr))
		return 1;

I
Ingo Molnar 已提交
617
	/*
618
	 * in-kernel percpu var?
I
Ingo Molnar 已提交
619
	 */
620 621
	if (is_kernel_percpu_address(addr))
		return 1;
I
Ingo Molnar 已提交
622

P
Peter Zijlstra 已提交
623
	/*
624
	 * module static or percpu var?
P
Peter Zijlstra 已提交
625
	 */
626
	return is_module_address(addr) || is_module_percpu_address(addr);
627
}
628
#endif
629

I
Ingo Molnar 已提交
630
/*
P
Peter Zijlstra 已提交
631 632
 * To make lock name printouts unique, we calculate a unique
 * class->name_version generation counter:
I
Ingo Molnar 已提交
633
 */
P
Peter Zijlstra 已提交
634
static int count_matching_names(struct lock_class *new_class)
I
Ingo Molnar 已提交
635
{
P
Peter Zijlstra 已提交
636 637
	struct lock_class *class;
	int count = 0;
I
Ingo Molnar 已提交
638

P
Peter Zijlstra 已提交
639
	if (!new_class->name)
I
Ingo Molnar 已提交
640 641
		return 0;

642
	list_for_each_entry_rcu(class, &all_lock_classes, lock_entry) {
P
Peter Zijlstra 已提交
643 644 645 646 647
		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
648

P
Peter Zijlstra 已提交
649
	return count + 1;
I
Ingo Molnar 已提交
650 651
}

P
Peter Zijlstra 已提交
652 653 654 655 656 657 658
/*
 * 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
659
{
P
Peter Zijlstra 已提交
660
	struct lockdep_subclass_key *key;
661
	struct hlist_head *hash_head;
P
Peter Zijlstra 已提交
662
	struct lock_class *class;
I
Ingo Molnar 已提交
663

664 665 666 667 668 669 670 671 672 673
	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
674 675 676 677 678 679
	/*
	 * 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
680

P
Peter Zijlstra 已提交
681 682 683 684 685 686
	/*
	 * 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
687 688
	BUILD_BUG_ON(sizeof(struct lock_class_key) >
			sizeof(struct lockdep_map));
I
Ingo Molnar 已提交
689

P
Peter Zijlstra 已提交
690
	key = lock->key->subkeys + subclass;
691

P
Peter Zijlstra 已提交
692
	hash_head = classhashentry(key);
693

P
Peter Zijlstra 已提交
694
	/*
695
	 * We do an RCU walk of the hash, see lockdep_free_key_range().
P
Peter Zijlstra 已提交
696
	 */
697 698 699
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return NULL;

700
	hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
P
Peter Zijlstra 已提交
701
		if (class->key == key) {
P
Peter Zijlstra 已提交
702 703 704 705
			/*
			 * Huh! same key, different name? Did someone trample
			 * on some memory? We're most confused.
			 */
P
Peter Zijlstra 已提交
706
			WARN_ON_ONCE(class->name != lock->name);
P
Peter Zijlstra 已提交
707
			return class;
P
Peter Zijlstra 已提交
708 709
		}
	}
I
Ingo Molnar 已提交
710

P
Peter Zijlstra 已提交
711
	return NULL;
I
Ingo Molnar 已提交
712 713 714
}

/*
P
Peter Zijlstra 已提交
715 716 717
 * 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
718
 */
719
static struct lock_class *
P
Peter Zijlstra 已提交
720
register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
I
Ingo Molnar 已提交
721
{
P
Peter Zijlstra 已提交
722
	struct lockdep_subclass_key *key;
723
	struct hlist_head *hash_head;
P
Peter Zijlstra 已提交
724
	struct lock_class *class;
725 726

	DEBUG_LOCKS_WARN_ON(!irqs_disabled());
P
Peter Zijlstra 已提交
727 728 729

	class = look_up_lock_class(lock, subclass);
	if (likely(class))
730
		goto out_set_class_cache;
P
Peter Zijlstra 已提交
731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754

	/*
	 * 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);

	if (!graph_lock()) {
		return NULL;
	}
	/*
	 * We have to do the hash-walk again, to avoid races
	 * with another CPU:
	 */
755
	hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
P
Peter Zijlstra 已提交
756 757
		if (class->key == key)
			goto out_unlock_set;
758 759
	}

P
Peter Zijlstra 已提交
760 761 762 763 764 765 766 767 768
	/*
	 * Allocate a new key from the static array, and add it to
	 * the hash:
	 */
	if (nr_lock_classes >= MAX_LOCKDEP_KEYS) {
		if (!debug_locks_off_graph_unlock()) {
			return NULL;
		}

769
		print_lockdep_off("BUG: MAX_LOCKDEP_KEYS too low!");
770
		dump_stack();
P
Peter Zijlstra 已提交
771 772 773
		return NULL;
	}
	class = lock_classes + nr_lock_classes++;
774
	debug_atomic_inc(nr_unused_locks);
P
Peter Zijlstra 已提交
775 776 777 778 779 780 781 782 783 784 785
	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:
	 */
786
	hlist_add_head_rcu(&class->hash_entry, hash_head);
787 788 789 790
	/*
	 * Add it to the global list of classes:
	 */
	list_add_tail_rcu(&class->lock_entry, &all_lock_classes);
P
Peter Zijlstra 已提交
791 792 793 794 795 796

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

		printk("\nnew class %p: %s", class->key, class->name);
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
797 798
			printk(KERN_CONT "#%d", class->name_version);
		printk(KERN_CONT "\n");
P
Peter Zijlstra 已提交
799 800 801 802 803 804 805 806 807
		dump_stack();

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

808
out_set_class_cache:
P
Peter Zijlstra 已提交
809
	if (!subclass || force)
810 811 812
		lock->class_cache[0] = class;
	else if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		lock->class_cache[subclass] = class;
P
Peter Zijlstra 已提交
813

P
Peter Zijlstra 已提交
814 815 816 817
	/*
	 * Hash collision, did we smoke some? We found a class with a matching
	 * hash but the subclass -- which is hashed in -- didn't match.
	 */
P
Peter Zijlstra 已提交
818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834
	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;

835
		print_lockdep_off("BUG: MAX_LOCKDEP_ENTRIES too low!");
836
		dump_stack();
P
Peter Zijlstra 已提交
837 838 839 840 841 842 843 844
		return NULL;
	}
	return list_entries + nr_list_entries++;
}

/*
 * Add a new dependency to the head of the list:
 */
845 846 847
static int add_lock_to_list(struct lock_class *this, struct list_head *head,
			    unsigned long ip, int distance,
			    struct stack_trace *trace)
P
Peter Zijlstra 已提交
848 849 850 851 852 853 854 855 856 857
{
	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;

858 859
	entry->class = this;
	entry->distance = distance;
Y
Yong Zhang 已提交
860
	entry->trace = *trace;
P
Peter Zijlstra 已提交
861
	/*
862 863 864
	 * Both allocation and removal are done under the graph lock; but
	 * iteration is under RCU-sched; see look_up_lock_class() and
	 * lockdep_free_key_range().
P
Peter Zijlstra 已提交
865 866 867 868 869 870
	 */
	list_add_tail_rcu(&entry->entry, head);

	return 1;
}

P
Peter Zijlstra 已提交
871 872 873
/*
 * For good efficiency of modular, we use power of 2
 */
P
Peter Zijlstra 已提交
874 875 876
#define MAX_CIRCULAR_QUEUE_SIZE		4096UL
#define CQ_MASK				(MAX_CIRCULAR_QUEUE_SIZE-1)

P
Peter Zijlstra 已提交
877 878
/*
 * The circular_queue and helpers is used to implement the
P
Peter Zijlstra 已提交
879 880 881
 * 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
882
 */
P
Peter Zijlstra 已提交
883 884 885 886 887 888 889
struct circular_queue {
	unsigned long element[MAX_CIRCULAR_QUEUE_SIZE];
	unsigned int  front, rear;
};

static struct circular_queue lock_cq;

890
unsigned int max_bfs_queue_depth;
P
Peter Zijlstra 已提交
891

892 893
static unsigned int lockdep_dependency_gen_id;

P
Peter Zijlstra 已提交
894 895 896
static inline void __cq_init(struct circular_queue *cq)
{
	cq->front = cq->rear = 0;
897
	lockdep_dependency_gen_id++;
P
Peter Zijlstra 已提交
898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938
}

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
939

P
Peter Zijlstra 已提交
940
	nr = lock - list_entries;
P
Peter Zijlstra 已提交
941
	WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
P
Peter Zijlstra 已提交
942
	lock->parent = parent;
943
	lock->class->dep_gen_id = lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
944 945 946 947 948
}

static inline unsigned long lock_accessed(struct lock_list *lock)
{
	unsigned long nr;
P
Peter Zijlstra 已提交
949

P
Peter Zijlstra 已提交
950
	nr = lock - list_entries;
P
Peter Zijlstra 已提交
951
	WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
952
	return lock->class->dep_gen_id == lockdep_dependency_gen_id;
P
Peter Zijlstra 已提交
953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971
}

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;
}

972
static int __bfs(struct lock_list *source_entry,
P
Peter Zijlstra 已提交
973 974 975 976
		 void *data,
		 int (*match)(struct lock_list *entry, void *data),
		 struct lock_list **target_entry,
		 int forward)
977 978
{
	struct lock_list *entry;
979
	struct list_head *head;
980 981 982
	struct circular_queue *cq = &lock_cq;
	int ret = 1;

983
	if (match(source_entry, data)) {
984 985 986 987 988
		*target_entry = source_entry;
		ret = 0;
		goto exit;
	}

989 990 991 992 993 994 995 996 997
	if (forward)
		head = &source_entry->class->locks_after;
	else
		head = &source_entry->class->locks_before;

	if (list_empty(head))
		goto exit;

	__cq_init(cq);
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
	__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;

1015 1016 1017
		DEBUG_LOCKS_WARN_ON(!irqs_disabled());

		list_for_each_entry_rcu(entry, head, entry) {
1018
			if (!lock_accessed(entry)) {
1019
				unsigned int cq_depth;
1020
				mark_lock_accessed(entry, lock);
1021
				if (match(entry, data)) {
1022 1023 1024 1025 1026 1027 1028 1029 1030
					*target_entry = entry;
					ret = 0;
					goto exit;
				}

				if (__cq_enqueue(cq, (unsigned long)entry)) {
					ret = -1;
					goto exit;
				}
1031 1032 1033
				cq_depth = __cq_get_elem_count(cq);
				if (max_bfs_queue_depth < cq_depth)
					max_bfs_queue_depth = cq_depth;
1034 1035 1036 1037 1038 1039 1040
			}
		}
	}
exit:
	return ret;
}

1041
static inline int __bfs_forwards(struct lock_list *src_entry,
1042 1043 1044
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1045
{
1046
	return __bfs(src_entry, data, match, target_entry, 1);
1047 1048 1049

}

1050
static inline int __bfs_backwards(struct lock_list *src_entry,
1051 1052 1053
			void *data,
			int (*match)(struct lock_list *entry, void *data),
			struct lock_list **target_entry)
1054
{
1055
	return __bfs(src_entry, data, match, target_entry, 0);
1056 1057 1058

}

P
Peter Zijlstra 已提交
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069
/*
 * 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
1070
print_circular_bug_entry(struct lock_list *target, int depth)
P
Peter Zijlstra 已提交
1071 1072 1073 1074 1075
{
	if (debug_locks_silent)
		return 0;
	printk("\n-> #%u", depth);
	print_lock_name(target->class);
D
Dmitry Vyukov 已提交
1076
	printk(KERN_CONT ":\n");
P
Peter Zijlstra 已提交
1077 1078 1079 1080 1081
	print_stack_trace(&target->trace, 6);

	return 0;
}

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
static void
print_circular_lock_scenario(struct held_lock *src,
			     struct held_lock *tgt,
			     struct lock_list *prt)
{
	struct lock_class *source = hlock_class(src);
	struct lock_class *target = hlock_class(tgt);
	struct lock_class *parent = prt->class;

	/*
	 * A direct locking problem where unsafe_class lock is taken
	 * directly by safe_class lock, then all we need to show
	 * is the deadlock scenario, as it is obvious that the
	 * unsafe lock is taken under the safe lock.
	 *
	 * But if there is a chain instead, where the safe lock takes
	 * an intermediate lock (middle_class) where this lock is
	 * not the same as the safe lock, then the lock chain is
	 * used to describe the problem. Otherwise we would need
	 * to show a different CPU case for each link in the chain
	 * from the safe_class lock to the unsafe_class lock.
	 */
	if (parent != source) {
		printk("Chain exists of:\n  ");
		__print_lock_name(source);
D
Dmitry Vyukov 已提交
1107
		printk(KERN_CONT " --> ");
1108
		__print_lock_name(parent);
D
Dmitry Vyukov 已提交
1109
		printk(KERN_CONT " --> ");
1110
		__print_lock_name(target);
D
Dmitry Vyukov 已提交
1111
		printk(KERN_CONT "\n\n");
1112 1113 1114 1115 1116 1117 1118
	}

	printk(" Possible unsafe locking scenario:\n\n");
	printk("       CPU0                    CPU1\n");
	printk("       ----                    ----\n");
	printk("  lock(");
	__print_lock_name(target);
D
Dmitry Vyukov 已提交
1119
	printk(KERN_CONT ");\n");
1120 1121
	printk("                               lock(");
	__print_lock_name(parent);
D
Dmitry Vyukov 已提交
1122
	printk(KERN_CONT ");\n");
1123 1124
	printk("                               lock(");
	__print_lock_name(target);
D
Dmitry Vyukov 已提交
1125
	printk(KERN_CONT ");\n");
1126 1127
	printk("  lock(");
	__print_lock_name(source);
D
Dmitry Vyukov 已提交
1128
	printk(KERN_CONT ");\n");
1129 1130 1131
	printk("\n *** DEADLOCK ***\n\n");
}

P
Peter Zijlstra 已提交
1132 1133 1134 1135 1136
/*
 * When a circular dependency is detected, print the
 * header first:
 */
static noinline int
1137 1138 1139
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
1140 1141 1142
{
	struct task_struct *curr = current;

1143
	if (debug_locks_silent)
P
Peter Zijlstra 已提交
1144 1145
		return 0;

1146 1147 1148
	printk("\n");
	printk("======================================================\n");
	printk("[ INFO: possible circular locking dependency detected ]\n");
1149
	print_kernel_ident();
1150
	printk("-------------------------------------------------------\n");
P
Peter Zijlstra 已提交
1151
	printk("%s/%d is trying to acquire lock:\n",
1152
		curr->comm, task_pid_nr(curr));
1153
	print_lock(check_src);
P
Peter Zijlstra 已提交
1154
	printk("\nbut task is already holding lock:\n");
1155
	print_lock(check_tgt);
P
Peter Zijlstra 已提交
1156 1157 1158 1159 1160 1161 1162 1163
	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;
}

1164 1165 1166 1167 1168
static inline int class_equal(struct lock_list *entry, void *data)
{
	return entry->class == data;
}

1169 1170 1171 1172
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
1173 1174
{
	struct task_struct *curr = current;
1175
	struct lock_list *parent;
1176
	struct lock_list *first_parent;
1177
	int depth;
P
Peter Zijlstra 已提交
1178

1179
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
P
Peter Zijlstra 已提交
1180 1181
		return 0;

1182
	if (!save_trace(&this->trace))
P
Peter Zijlstra 已提交
1183 1184
		return 0;

1185 1186
	depth = get_lock_depth(target);

1187
	print_circular_bug_header(target, depth, check_src, check_tgt);
1188 1189

	parent = get_lock_parent(target);
1190
	first_parent = parent;
1191 1192 1193 1194 1195

	while (parent) {
		print_circular_bug_entry(parent, --depth);
		parent = get_lock_parent(parent);
	}
P
Peter Zijlstra 已提交
1196 1197

	printk("\nother info that might help us debug this:\n\n");
1198 1199 1200
	print_circular_lock_scenario(check_src, check_tgt,
				     first_parent);

P
Peter Zijlstra 已提交
1201 1202 1203 1204 1205 1206 1207 1208
	lockdep_print_held_locks(curr);

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

	return 0;
}

1209 1210 1211 1212 1213
static noinline int print_bfs_bug(int ret)
{
	if (!debug_locks_off_graph_unlock())
		return 0;

P
Peter Zijlstra 已提交
1214 1215 1216
	/*
	 * Breadth-first-search failed, graph got corrupted?
	 */
1217 1218 1219 1220 1221
	WARN(1, "lockdep bfs error:%d\n", ret);

	return 0;
}

1222
static int noop_count(struct lock_list *entry, void *data)
1223
{
1224 1225 1226
	(*(unsigned long *)data)++;
	return 0;
}
1227

1228
static unsigned long __lockdep_count_forward_deps(struct lock_list *this)
1229 1230 1231
{
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1232

1233
	__bfs_forwards(this, (void *)&count, noop_count, &target_entry);
1234

1235
	return count;
1236 1237 1238 1239
}
unsigned long lockdep_count_forward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1240 1241 1242 1243
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1244 1245

	local_irq_save(flags);
1246
	arch_spin_lock(&lockdep_lock);
1247
	ret = __lockdep_count_forward_deps(&this);
1248
	arch_spin_unlock(&lockdep_lock);
1249 1250 1251 1252 1253
	local_irq_restore(flags);

	return ret;
}

1254
static unsigned long __lockdep_count_backward_deps(struct lock_list *this)
1255
{
1256 1257
	unsigned long  count = 0;
	struct lock_list *uninitialized_var(target_entry);
1258

1259
	__bfs_backwards(this, (void *)&count, noop_count, &target_entry);
1260

1261
	return count;
1262 1263 1264 1265 1266
}

unsigned long lockdep_count_backward_deps(struct lock_class *class)
{
	unsigned long ret, flags;
1267 1268 1269 1270
	struct lock_list this;

	this.parent = NULL;
	this.class = class;
1271 1272

	local_irq_save(flags);
1273
	arch_spin_lock(&lockdep_lock);
1274
	ret = __lockdep_count_backward_deps(&this);
1275
	arch_spin_unlock(&lockdep_lock);
1276 1277 1278 1279 1280
	local_irq_restore(flags);

	return ret;
}

P
Peter Zijlstra 已提交
1281 1282 1283 1284 1285
/*
 * 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
1286 1287
check_noncircular(struct lock_list *root, struct lock_class *target,
		struct lock_list **target_entry)
P
Peter Zijlstra 已提交
1288
{
1289
	int result;
P
Peter Zijlstra 已提交
1290

1291
	debug_atomic_inc(nr_cyclic_checks);
1292

1293
	result = __bfs_forwards(root, target, class_equal, target_entry);
I
Ingo Molnar 已提交
1294

1295 1296
	return result;
}
1297

1298
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
1299 1300 1301 1302 1303 1304
/*
 * 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.
 */

1305 1306 1307 1308 1309 1310 1311
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
1312 1313
/*
 * Find a node in the forwards-direction dependency sub-graph starting
1314
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1315
 *
1316 1317
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1318
 *
1319 1320
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1321
 */
1322 1323 1324
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
1325
{
1326
	int result;
I
Ingo Molnar 已提交
1327

1328
	debug_atomic_inc(nr_find_usage_forwards_checks);
I
Ingo Molnar 已提交
1329

1330 1331 1332
	result = __bfs_forwards(root, (void *)bit, usage_match, target_entry);

	return result;
I
Ingo Molnar 已提交
1333 1334 1335 1336
}

/*
 * Find a node in the backwards-direction dependency sub-graph starting
1337
 * at @root->class that matches @bit.
I
Ingo Molnar 已提交
1338
 *
1339 1340
 * Return 0 if such a node exists in the subgraph, and put that node
 * into *@target_entry.
I
Ingo Molnar 已提交
1341
 *
1342 1343
 * Return 1 otherwise and keep *@target_entry unchanged.
 * Return <0 on error.
I
Ingo Molnar 已提交
1344
 */
1345 1346 1347
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
1348
{
1349
	int result;
I
Ingo Molnar 已提交
1350

1351
	debug_atomic_inc(nr_find_usage_backwards_checks);
I
Ingo Molnar 已提交
1352

1353
	result = __bfs_backwards(root, (void *)bit, usage_match, target_entry);
D
Dave Jones 已提交
1354

1355
	return result;
I
Ingo Molnar 已提交
1356 1357
}

P
Peter Zijlstra 已提交
1358 1359 1360 1361 1362 1363
static void print_lock_class_header(struct lock_class *class, int depth)
{
	int bit;

	printk("%*s->", depth, "");
	print_lock_name(class);
D
Dmitry Vyukov 已提交
1364 1365
	printk(KERN_CONT " ops: %lu", class->ops);
	printk(KERN_CONT " {\n");
P
Peter Zijlstra 已提交
1366 1367 1368 1369 1370 1371

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

			len += printk("%*s   %s", depth, "", usage_str[bit]);
D
Dmitry Vyukov 已提交
1372
			len += printk(KERN_CONT " at:\n");
P
Peter Zijlstra 已提交
1373 1374 1375 1376 1377
			print_stack_trace(class->usage_traces + bit, len);
		}
	}
	printk("%*s }\n", depth, "");

D
Dmitry Vyukov 已提交
1378 1379
	printk("%*s ... key      at: [<%p>] %pS\n",
		depth, "", class->key, class->key);
P
Peter Zijlstra 已提交
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
}

/*
 * 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)) {
1402
			printk("lockdep:%s bad path found in chain graph\n", __func__);
P
Peter Zijlstra 已提交
1403 1404 1405 1406 1407 1408 1409 1410 1411
			break;
		}

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

	return;
}
1412

1413 1414 1415
static void
print_irq_lock_scenario(struct lock_list *safe_entry,
			struct lock_list *unsafe_entry,
1416 1417
			struct lock_class *prev_class,
			struct lock_class *next_class)
1418 1419 1420
{
	struct lock_class *safe_class = safe_entry->class;
	struct lock_class *unsafe_class = unsafe_entry->class;
1421
	struct lock_class *middle_class = prev_class;
1422 1423

	if (middle_class == safe_class)
1424
		middle_class = next_class;
1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441

	/*
	 * A direct locking problem where unsafe_class lock is taken
	 * directly by safe_class lock, then all we need to show
	 * is the deadlock scenario, as it is obvious that the
	 * unsafe lock is taken under the safe lock.
	 *
	 * But if there is a chain instead, where the safe lock takes
	 * an intermediate lock (middle_class) where this lock is
	 * not the same as the safe lock, then the lock chain is
	 * used to describe the problem. Otherwise we would need
	 * to show a different CPU case for each link in the chain
	 * from the safe_class lock to the unsafe_class lock.
	 */
	if (middle_class != unsafe_class) {
		printk("Chain exists of:\n  ");
		__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1442
		printk(KERN_CONT " --> ");
1443
		__print_lock_name(middle_class);
D
Dmitry Vyukov 已提交
1444
		printk(KERN_CONT " --> ");
1445
		__print_lock_name(unsafe_class);
D
Dmitry Vyukov 已提交
1446
		printk(KERN_CONT "\n\n");
1447 1448 1449 1450 1451 1452 1453
	}

	printk(" Possible interrupt unsafe locking scenario:\n\n");
	printk("       CPU0                    CPU1\n");
	printk("       ----                    ----\n");
	printk("  lock(");
	__print_lock_name(unsafe_class);
D
Dmitry Vyukov 已提交
1454
	printk(KERN_CONT ");\n");
1455 1456 1457
	printk("                               local_irq_disable();\n");
	printk("                               lock(");
	__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1458
	printk(KERN_CONT ");\n");
1459 1460
	printk("                               lock(");
	__print_lock_name(middle_class);
D
Dmitry Vyukov 已提交
1461
	printk(KERN_CONT ");\n");
1462 1463 1464
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(safe_class);
D
Dmitry Vyukov 已提交
1465
	printk(KERN_CONT ");\n");
1466 1467 1468
	printk("\n *** DEADLOCK ***\n\n");
}

I
Ingo Molnar 已提交
1469 1470
static int
print_bad_irq_dependency(struct task_struct *curr,
1471 1472 1473 1474
			 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
1475 1476 1477 1478 1479 1480
			 struct held_lock *prev,
			 struct held_lock *next,
			 enum lock_usage_bit bit1,
			 enum lock_usage_bit bit2,
			 const char *irqclass)
{
1481
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1482 1483
		return 0;

1484 1485 1486
	printk("\n");
	printk("======================================================\n");
	printk("[ INFO: %s-safe -> %s-unsafe lock order detected ]\n",
I
Ingo Molnar 已提交
1487
		irqclass, irqclass);
1488
	print_kernel_ident();
1489
	printk("------------------------------------------------------\n");
I
Ingo Molnar 已提交
1490
	printk("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] is trying to acquire:\n",
1491
		curr->comm, task_pid_nr(curr),
I
Ingo Molnar 已提交
1492 1493 1494 1495 1496 1497 1498 1499 1500
		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
1501
	print_lock_name(hlock_class(prev));
D
Dmitry Vyukov 已提交
1502
	printk(KERN_CONT " ->");
D
Dave Jones 已提交
1503
	print_lock_name(hlock_class(next));
D
Dmitry Vyukov 已提交
1504
	printk(KERN_CONT "\n");
I
Ingo Molnar 已提交
1505 1506 1507

	printk("\nbut this new dependency connects a %s-irq-safe lock:\n",
		irqclass);
1508
	print_lock_name(backwards_entry->class);
I
Ingo Molnar 已提交
1509 1510
	printk("\n... which became %s-irq-safe at:\n", irqclass);

1511
	print_stack_trace(backwards_entry->class->usage_traces + bit1, 1);
I
Ingo Molnar 已提交
1512 1513

	printk("\nto a %s-irq-unsafe lock:\n", irqclass);
1514
	print_lock_name(forwards_entry->class);
I
Ingo Molnar 已提交
1515 1516 1517
	printk("\n... which became %s-irq-unsafe at:\n", irqclass);
	printk("...");

1518
	print_stack_trace(forwards_entry->class->usage_traces + bit2, 1);
I
Ingo Molnar 已提交
1519 1520

	printk("\nother info that might help us debug this:\n\n");
1521 1522
	print_irq_lock_scenario(backwards_entry, forwards_entry,
				hlock_class(prev), hlock_class(next));
1523

I
Ingo Molnar 已提交
1524 1525
	lockdep_print_held_locks(curr);

D
Dmitry Vyukov 已提交
1526
	printk("\nthe dependencies between %s-irq-safe lock and the holding lock:\n", irqclass);
1527 1528 1529
	if (!save_trace(&prev_root->trace))
		return 0;
	print_shortest_lock_dependencies(backwards_entry, prev_root);
I
Ingo Molnar 已提交
1530

1531 1532 1533 1534 1535
	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
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548

	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;
1549
	struct lock_list this, that;
1550
	struct lock_list *uninitialized_var(target_entry);
1551
	struct lock_list *uninitialized_var(target_entry1);
1552 1553 1554 1555 1556

	this.parent = NULL;

	this.class = hlock_class(prev);
	ret = find_usage_backwards(&this, bit_backwards, &target_entry);
P
Peter Zijlstra 已提交
1557 1558 1559 1560
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
1561

1562 1563 1564
	that.parent = NULL;
	that.class = hlock_class(next);
	ret = find_usage_forwards(&that, bit_forwards, &target_entry1);
P
Peter Zijlstra 已提交
1565 1566 1567 1568
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
1569

1570 1571 1572
	return print_bad_irq_dependency(curr, &this, &that,
			target_entry, target_entry1,
			prev, next,
I
Ingo Molnar 已提交
1573 1574 1575
			bit_backwards, bit_forwards, irqclass);
}

1576 1577
static const char *state_names[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1578
	__stringify(__STATE),
1579 1580 1581 1582 1583 1584
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static const char *state_rnames[] = {
#define LOCKDEP_STATE(__STATE) \
P
Peter Zijlstra 已提交
1585
	__stringify(__STATE)"-READ",
1586 1587 1588 1589 1590
#include "lockdep_states.h"
#undef LOCKDEP_STATE
};

static inline const char *state_name(enum lock_usage_bit bit)
P
Peter Zijlstra 已提交
1591
{
1592 1593
	return (bit & 1) ? state_rnames[bit >> 2] : state_names[bit >> 2];
}
P
Peter Zijlstra 已提交
1594

1595 1596
static int exclusive_bit(int new_bit)
{
P
Peter Zijlstra 已提交
1597
	/*
1598 1599 1600 1601 1602 1603 1604 1605
	 * 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
1606
	 */
1607 1608 1609

	int state = new_bit & ~3;
	int dir = new_bit & 2;
P
Peter Zijlstra 已提交
1610 1611

	/*
1612
	 * keep state, bit flip the direction and strip read.
P
Peter Zijlstra 已提交
1613
	 */
1614 1615 1616 1617 1618 1619
	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
1620
	/*
1621 1622
	 * 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
1623 1624 1625
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1626 1627
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
P
Peter Zijlstra 已提交
1628 1629
		return 0;

1630 1631
	bit++; /* _READ */

1632
	/*
1633 1634
	 * Prove that the new dependency does not connect a hardirq-safe-read
	 * lock with a hardirq-unsafe lock - to achieve this we search
1635 1636 1637
	 * the backwards-subgraph starting at <prev>, and the
	 * forwards-subgraph starting at <next>:
	 */
1638 1639
	if (!check_usage(curr, prev, next, bit,
			   exclusive_bit(bit), state_name(bit)))
1640 1641
		return 0;

1642 1643 1644 1645 1646 1647 1648 1649 1650
	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))	\
1651
		return 0;
1652 1653
#include "lockdep_states.h"
#undef LOCKDEP_STATE
1654

P
Peter Zijlstra 已提交
1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683
	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
1684 1685
#endif

1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
static void
print_deadlock_scenario(struct held_lock *nxt,
			     struct held_lock *prv)
{
	struct lock_class *next = hlock_class(nxt);
	struct lock_class *prev = hlock_class(prv);

	printk(" Possible unsafe locking scenario:\n\n");
	printk("       CPU0\n");
	printk("       ----\n");
	printk("  lock(");
	__print_lock_name(prev);
D
Dmitry Vyukov 已提交
1698
	printk(KERN_CONT ");\n");
1699 1700
	printk("  lock(");
	__print_lock_name(next);
D
Dmitry Vyukov 已提交
1701
	printk(KERN_CONT ");\n");
1702 1703 1704 1705
	printk("\n *** DEADLOCK ***\n\n");
	printk(" May be due to missing lock nesting notation\n\n");
}

I
Ingo Molnar 已提交
1706 1707 1708 1709
static int
print_deadlock_bug(struct task_struct *curr, struct held_lock *prev,
		   struct held_lock *next)
{
1710
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
1711 1712
		return 0;

1713 1714 1715
	printk("\n");
	printk("=============================================\n");
	printk("[ INFO: possible recursive locking detected ]\n");
1716
	print_kernel_ident();
1717
	printk("---------------------------------------------\n");
I
Ingo Molnar 已提交
1718
	printk("%s/%d is trying to acquire lock:\n",
1719
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
1720 1721 1722 1723 1724
	print_lock(next);
	printk("\nbut task is already holding lock:\n");
	print_lock(prev);

	printk("\nother info that might help us debug this:\n");
1725
	print_deadlock_scenario(next, prev);
I
Ingo Molnar 已提交
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
	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
1747
	struct held_lock *nest = NULL;
I
Ingo Molnar 已提交
1748 1749 1750 1751
	int i;

	for (i = 0; i < curr->lockdep_depth; i++) {
		prev = curr->held_locks + i;
P
Peter Zijlstra 已提交
1752 1753 1754 1755

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

D
Dave Jones 已提交
1756
		if (hlock_class(prev) != hlock_class(next))
I
Ingo Molnar 已提交
1757
			continue;
P
Peter Zijlstra 已提交
1758

I
Ingo Molnar 已提交
1759 1760
		/*
		 * Allow read-after-read recursion of the same
1761
		 * lock class (i.e. read_lock(lock)+read_lock(lock)):
I
Ingo Molnar 已提交
1762
		 */
1763
		if ((read == 2) && prev->read)
I
Ingo Molnar 已提交
1764
			return 2;
P
Peter Zijlstra 已提交
1765 1766 1767 1768 1769 1770 1771 1772

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

I
Ingo Molnar 已提交
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801
		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,
1802
	       struct held_lock *next, int distance, int *stack_saved)
I
Ingo Molnar 已提交
1803 1804 1805
{
	struct lock_list *entry;
	int ret;
1806 1807
	struct lock_list this;
	struct lock_list *uninitialized_var(target_entry);
Y
Yong Zhang 已提交
1808 1809 1810 1811 1812 1813 1814 1815
	/*
	 * 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
1816 1817 1818 1819 1820 1821 1822 1823 1824 1825

	/*
	 * 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:
	 */
1826 1827 1828 1829 1830 1831 1832
	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);
1833

P
Peter Zijlstra 已提交
1834
	if (!check_prev_add_irq(curr, prev, next))
I
Ingo Molnar 已提交
1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854
		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
1855 1856
	list_for_each_entry(entry, &hlock_class(prev)->locks_after, entry) {
		if (entry->class == hlock_class(next)) {
1857 1858
			if (distance == 1)
				entry->distance = 1;
I
Ingo Molnar 已提交
1859
			return 2;
1860
		}
I
Ingo Molnar 已提交
1861 1862
	}

1863 1864 1865 1866 1867
	if (!*stack_saved) {
		if (!save_trace(&trace))
			return 0;
		*stack_saved = 1;
	}
Y
Yong Zhang 已提交
1868

I
Ingo Molnar 已提交
1869 1870 1871 1872
	/*
	 * Ok, all validations passed, add the new lock
	 * to the previous lock's dependency list:
	 */
1873
	ret = add_lock_to_list(hlock_class(next),
D
Dave Jones 已提交
1874
			       &hlock_class(prev)->locks_after,
Y
Yong Zhang 已提交
1875
			       next->acquire_ip, distance, &trace);
1876

I
Ingo Molnar 已提交
1877 1878
	if (!ret)
		return 0;
1879

1880
	ret = add_lock_to_list(hlock_class(prev),
D
Dave Jones 已提交
1881
			       &hlock_class(next)->locks_before,
Y
Yong Zhang 已提交
1882
			       next->acquire_ip, distance, &trace);
1883 1884
	if (!ret)
		return 0;
I
Ingo Molnar 已提交
1885 1886

	/*
P
Peter Zijlstra 已提交
1887 1888
	 * Debugging printouts:
	 */
D
Dave Jones 已提交
1889
	if (verbose(hlock_class(prev)) || verbose(hlock_class(next))) {
1890 1891
		/* We drop graph lock, so another thread can overwrite trace. */
		*stack_saved = 0;
P
Peter Zijlstra 已提交
1892 1893
		graph_unlock();
		printk("\n new dependency: ");
D
Dave Jones 已提交
1894
		print_lock_name(hlock_class(prev));
D
Dmitry Vyukov 已提交
1895
		printk(KERN_CONT " => ");
D
Dave Jones 已提交
1896
		print_lock_name(hlock_class(next));
D
Dmitry Vyukov 已提交
1897
		printk(KERN_CONT "\n");
I
Ingo Molnar 已提交
1898
		dump_stack();
P
Peter Zijlstra 已提交
1899
		return graph_lock();
I
Ingo Molnar 已提交
1900
	}
P
Peter Zijlstra 已提交
1901 1902
	return 1;
}
I
Ingo Molnar 已提交
1903

P
Peter Zijlstra 已提交
1904 1905 1906 1907 1908 1909 1910 1911 1912 1913
/*
 * 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;
1914
	int stack_saved = 0;
P
Peter Zijlstra 已提交
1915
	struct held_lock *hlock;
1916

I
Ingo Molnar 已提交
1917
	/*
P
Peter Zijlstra 已提交
1918 1919 1920
	 * Debugging checks.
	 *
	 * Depth must not be zero for a non-head lock:
I
Ingo Molnar 已提交
1921
	 */
P
Peter Zijlstra 已提交
1922 1923
	if (!depth)
		goto out_bug;
I
Ingo Molnar 已提交
1924
	/*
P
Peter Zijlstra 已提交
1925 1926
	 * At least two relevant locks must exist for this
	 * to be a head:
I
Ingo Molnar 已提交
1927
	 */
P
Peter Zijlstra 已提交
1928 1929 1930
	if (curr->held_locks[depth].irq_context !=
			curr->held_locks[depth-1].irq_context)
		goto out_bug;
1931

P
Peter Zijlstra 已提交
1932 1933
	for (;;) {
		int distance = curr->lockdep_depth - depth + 1;
1934
		hlock = curr->held_locks + depth - 1;
P
Peter Zijlstra 已提交
1935 1936 1937 1938
		/*
		 * Only non-recursive-read entries get new dependencies
		 * added:
		 */
1939
		if (hlock->read != 2 && hlock->check) {
Y
Yong Zhang 已提交
1940
			if (!check_prev_add(curr, hlock, next,
1941
						distance, &stack_saved))
P
Peter Zijlstra 已提交
1942 1943 1944 1945 1946 1947 1948 1949 1950
				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;
1951
		}
P
Peter Zijlstra 已提交
1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963
		depth--;
		/*
		 * End of lock-stack?
		 */
		if (!depth)
			break;
		/*
		 * Stop the search if we cross into another context:
		 */
		if (curr->held_locks[depth].irq_context !=
				curr->held_locks[depth-1].irq_context)
			break;
I
Ingo Molnar 已提交
1964
	}
P
Peter Zijlstra 已提交
1965 1966 1967 1968
	return 1;
out_bug:
	if (!debug_locks_off_graph_unlock())
		return 0;
I
Ingo Molnar 已提交
1969

P
Peter Zijlstra 已提交
1970 1971 1972 1973 1974
	/*
	 * Clearly we all shouldn't be here, but since we made it we
	 * can reliable say we messed up our state. See the above two
	 * gotos for reasons why we could possibly end up here.
	 */
P
Peter Zijlstra 已提交
1975
	WARN_ON(1);
I
Ingo Molnar 已提交
1976

P
Peter Zijlstra 已提交
1977
	return 0;
I
Ingo Molnar 已提交
1978 1979
}

P
Peter Zijlstra 已提交
1980
unsigned long nr_lock_chains;
1981
struct lock_chain lock_chains[MAX_LOCKDEP_CHAINS];
1982
int nr_chain_hlocks;
1983 1984 1985 1986 1987 1988
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
1989

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
/*
 * Returns the index of the first held_lock of the current chain
 */
static inline int get_first_held_lock(struct task_struct *curr,
					struct held_lock *hlock)
{
	int i;
	struct held_lock *hlock_curr;

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

	}

	return ++i;
}

2009
#ifdef CONFIG_DEBUG_LOCKDEP
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079
/*
 * Returns the next chain_key iteration
 */
static u64 print_chain_key_iteration(int class_idx, u64 chain_key)
{
	u64 new_chain_key = iterate_chain_key(chain_key, class_idx);

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

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

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

		print_lock(hlock);
	}

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

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

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

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

static void print_collision(struct task_struct *curr,
			struct held_lock *hlock_next,
			struct lock_chain *chain)
{
	printk("\n");
	printk("======================\n");
	printk("[chain_key collision ]\n");
	print_kernel_ident();
	printk("----------------------\n");
	printk("%s/%d: ", current->comm, task_pid_nr(current));
	printk("Hash chain already cached but the contents don't match!\n");

	printk("Held locks:");
	print_chain_keys_held_locks(curr, hlock_next);

	printk("Locks in cached chain:");
	print_chain_keys_chain(chain);

	printk("\nstack backtrace:\n");
	dump_stack();
}
2080
#endif
2081

2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
/*
 * Checks whether the chain and the current held locks are consistent
 * in depth and also in content. If they are not it most likely means
 * that there was a collision during the calculation of the chain_key.
 * Returns: 0 not passed, 1 passed
 */
static int check_no_collision(struct task_struct *curr,
			struct held_lock *hlock,
			struct lock_chain *chain)
{
#ifdef CONFIG_DEBUG_LOCKDEP
	int i, j, id;

	i = get_first_held_lock(curr, hlock);

2097 2098
	if (DEBUG_LOCKS_WARN_ON(chain->depth != curr->lockdep_depth - (i - 1))) {
		print_collision(curr, hlock, chain);
2099
		return 0;
2100
	}
2101 2102 2103 2104

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

2105 2106
		if (DEBUG_LOCKS_WARN_ON(chain_hlocks[chain->base + j] != id)) {
			print_collision(curr, hlock, chain);
2107
			return 0;
2108
		}
2109 2110 2111 2112 2113
	}
#endif
	return 1;
}

I
Ingo Molnar 已提交
2114 2115
/*
 * Look up a dependency chain. If the key is not present yet then
2116 2117 2118
 * 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
2119
 */
2120 2121 2122
static inline int lookup_chain_cache(struct task_struct *curr,
				     struct held_lock *hlock,
				     u64 chain_key)
I
Ingo Molnar 已提交
2123
{
D
Dave Jones 已提交
2124
	struct lock_class *class = hlock_class(hlock);
2125
	struct hlist_head *hash_head = chainhashentry(chain_key);
I
Ingo Molnar 已提交
2126
	struct lock_chain *chain;
2127
	int i, j;
I
Ingo Molnar 已提交
2128

P
Peter Zijlstra 已提交
2129 2130 2131 2132 2133
	/*
	 * We might need to take the graph lock, ensure we've got IRQs
	 * disabled to make this an IRQ-safe lock.. for recursion reasons
	 * lockdep won't complain about its own locking errors.
	 */
2134 2135
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;
I
Ingo Molnar 已提交
2136 2137 2138 2139
	/*
	 * We can walk it lock-free, because entries only get added
	 * to the hash:
	 */
2140
	hlist_for_each_entry_rcu(chain, hash_head, entry) {
I
Ingo Molnar 已提交
2141 2142
		if (chain->chain_key == chain_key) {
cache_hit:
2143
			debug_atomic_inc(chain_lookup_hits);
2144 2145 2146
			if (!check_no_collision(curr, hlock, chain))
				return 0;

2147
			if (very_verbose(class))
2148 2149 2150 2151
				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
2152 2153 2154
			return 0;
		}
	}
2155
	if (very_verbose(class))
2156 2157
		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
2158 2159 2160 2161
	/*
	 * Allocate a new chain entry from the static array, and add
	 * it to the hash:
	 */
2162 2163
	if (!graph_lock())
		return 0;
I
Ingo Molnar 已提交
2164 2165 2166
	/*
	 * We have to walk the chain again locked - to avoid duplicates:
	 */
2167
	hlist_for_each_entry(chain, hash_head, entry) {
I
Ingo Molnar 已提交
2168
		if (chain->chain_key == chain_key) {
2169
			graph_unlock();
I
Ingo Molnar 已提交
2170 2171 2172 2173
			goto cache_hit;
		}
	}
	if (unlikely(nr_lock_chains >= MAX_LOCKDEP_CHAINS)) {
2174 2175 2176
		if (!debug_locks_off_graph_unlock())
			return 0;

2177
		print_lockdep_off("BUG: MAX_LOCKDEP_CHAINS too low!");
2178
		dump_stack();
I
Ingo Molnar 已提交
2179 2180 2181 2182
		return 0;
	}
	chain = lock_chains + nr_lock_chains++;
	chain->chain_key = chain_key;
2183
	chain->irq_context = hlock->irq_context;
2184
	i = get_first_held_lock(curr, hlock);
2185
	chain->depth = curr->lockdep_depth + 1 - i;
2186 2187 2188 2189 2190

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

2191 2192
	if (likely(nr_chain_hlocks + chain->depth <= MAX_LOCKDEP_CHAIN_HLOCKS)) {
		chain->base = nr_chain_hlocks;
2193
		for (j = 0; j < chain->depth - 1; j++, i++) {
D
Dave Jones 已提交
2194
			int lock_id = curr->held_locks[i].class_idx - 1;
2195 2196 2197 2198
			chain_hlocks[chain->base + j] = lock_id;
		}
		chain_hlocks[chain->base + j] = class - lock_classes;
	}
2199 2200 2201 2202 2203 2204 2205 2206 2207

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

#ifdef CONFIG_DEBUG_LOCKDEP
	/*
	 * Important for check_no_collision().
	 */
	if (unlikely(nr_chain_hlocks > MAX_LOCKDEP_CHAIN_HLOCKS)) {
2208
		if (!debug_locks_off_graph_unlock())
2209 2210 2211 2212 2213 2214 2215 2216
			return 0;

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

2217
	hlist_add_head_rcu(&chain->entry, hash_head);
2218
	debug_atomic_inc(chain_lookup_misses);
P
Peter Zijlstra 已提交
2219 2220 2221 2222 2223 2224
	inc_chains();

	return 1;
}

static int validate_chain(struct task_struct *curr, struct lockdep_map *lock,
2225
		struct held_lock *hlock, int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236
{
	/*
	 * 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)
	 */
2237
	if (!hlock->trylock && hlock->check &&
2238
	    lookup_chain_cache(curr, hlock, chain_key)) {
P
Peter Zijlstra 已提交
2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273
		/*
		 * 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
2274 2275 2276

	return 1;
}
P
Peter Zijlstra 已提交
2277 2278 2279
#else
static inline int validate_chain(struct task_struct *curr,
	       	struct lockdep_map *lock, struct held_lock *hlock,
2280
		int chain_head, u64 chain_key)
P
Peter Zijlstra 已提交
2281 2282 2283
{
	return 1;
}
2284
#endif
I
Ingo Molnar 已提交
2285 2286 2287 2288 2289

/*
 * We are building curr_chain_key incrementally, so double-check
 * it from scratch, to make sure that it's done correctly:
 */
2290
static void check_chain_key(struct task_struct *curr)
I
Ingo Molnar 已提交
2291 2292 2293
{
#ifdef CONFIG_DEBUG_LOCKDEP
	struct held_lock *hlock, *prev_hlock = NULL;
2294
	unsigned int i;
I
Ingo Molnar 已提交
2295 2296 2297 2298 2299 2300
	u64 chain_key = 0;

	for (i = 0; i < curr->lockdep_depth; i++) {
		hlock = curr->held_locks + i;
		if (chain_key != hlock->prev_chain_key) {
			debug_locks_off();
P
Peter Zijlstra 已提交
2301 2302 2303 2304
			/*
			 * We got mighty confused, our chain keys don't match
			 * with what we expect, someone trample on our task state?
			 */
2305
			WARN(1, "hm#1, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2306 2307 2308 2309 2310
				curr->lockdep_depth, i,
				(unsigned long long)chain_key,
				(unsigned long long)hlock->prev_chain_key);
			return;
		}
P
Peter Zijlstra 已提交
2311 2312 2313
		/*
		 * Whoops ran out of static storage again?
		 */
2314
		if (DEBUG_LOCKS_WARN_ON(hlock->class_idx > MAX_LOCKDEP_KEYS))
2315 2316
			return;

I
Ingo Molnar 已提交
2317 2318 2319
		if (prev_hlock && (prev_hlock->irq_context !=
							hlock->irq_context))
			chain_key = 0;
2320
		chain_key = iterate_chain_key(chain_key, hlock->class_idx);
I
Ingo Molnar 已提交
2321 2322 2323 2324
		prev_hlock = hlock;
	}
	if (chain_key != curr->curr_chain_key) {
		debug_locks_off();
P
Peter Zijlstra 已提交
2325 2326 2327 2328
		/*
		 * More smoking hash instead of calculating it, damn see these
		 * numbers float.. I bet that a pink elephant stepped on my memory.
		 */
2329
		WARN(1, "hm#2, depth: %u [%u], %016Lx != %016Lx\n",
I
Ingo Molnar 已提交
2330 2331 2332 2333 2334 2335 2336
			curr->lockdep_depth, i,
			(unsigned long long)chain_key,
			(unsigned long long)curr->curr_chain_key);
	}
#endif
}

2337 2338 2339 2340 2341 2342 2343 2344 2345 2346
static void
print_usage_bug_scenario(struct held_lock *lock)
{
	struct lock_class *class = hlock_class(lock);

	printk(" Possible unsafe locking scenario:\n\n");
	printk("       CPU0\n");
	printk("       ----\n");
	printk("  lock(");
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
2347
	printk(KERN_CONT ");\n");
2348 2349 2350
	printk("  <Interrupt>\n");
	printk("    lock(");
	__print_lock_name(class);
D
Dmitry Vyukov 已提交
2351
	printk(KERN_CONT ");\n");
2352 2353 2354
	printk("\n *** DEADLOCK ***\n\n");
}

P
Peter Zijlstra 已提交
2355 2356 2357 2358 2359 2360 2361
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;

2362 2363 2364
	printk("\n");
	printk("=================================\n");
	printk("[ INFO: inconsistent lock state ]\n");
2365
	print_kernel_ident();
2366
	printk("---------------------------------\n");
P
Peter Zijlstra 已提交
2367 2368 2369 2370 2371

	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",
2372
		curr->comm, task_pid_nr(curr),
P
Peter Zijlstra 已提交
2373 2374 2375 2376 2377 2378 2379
		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
2380
	print_stack_trace(hlock_class(this)->usage_traces + prev_bit, 1);
P
Peter Zijlstra 已提交
2381 2382 2383

	print_irqtrace_events(curr);
	printk("\nother info that might help us debug this:\n");
2384 2385
	print_usage_bug_scenario(this);

P
Peter Zijlstra 已提交
2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400
	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
2401
	if (unlikely(hlock_class(this)->usage_mask & (1 << bad_bit)))
P
Peter Zijlstra 已提交
2402 2403 2404 2405 2406 2407 2408
		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);

2409
#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
I
Ingo Molnar 已提交
2410 2411 2412 2413 2414

/*
 * print irq inversion bug:
 */
static int
2415 2416
print_irq_inversion_bug(struct task_struct *curr,
			struct lock_list *root, struct lock_list *other,
I
Ingo Molnar 已提交
2417 2418 2419
			struct held_lock *this, int forwards,
			const char *irqclass)
{
2420 2421 2422 2423
	struct lock_list *entry = other;
	struct lock_list *middle = NULL;
	int depth;

2424
	if (!debug_locks_off_graph_unlock() || debug_locks_silent)
I
Ingo Molnar 已提交
2425 2426
		return 0;

2427 2428 2429
	printk("\n");
	printk("=========================================================\n");
	printk("[ INFO: possible irq lock inversion dependency detected ]\n");
2430
	print_kernel_ident();
2431
	printk("---------------------------------------------------------\n");
I
Ingo Molnar 已提交
2432
	printk("%s/%d just changed the state of lock:\n",
2433
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
2434 2435
	print_lock(this);
	if (forwards)
2436
		printk("but this lock took another, %s-unsafe lock in the past:\n", irqclass);
I
Ingo Molnar 已提交
2437
	else
2438
		printk("but this lock was taken by another, %s-safe lock in the past:\n", irqclass);
2439
	print_lock_name(other->class);
I
Ingo Molnar 已提交
2440 2441 2442
	printk("\n\nand interrupts could create inverse lock ordering between them.\n\n");

	printk("\nother info that might help us debug this:\n");
2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461

	/* 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
2462 2463
	lockdep_print_held_locks(curr);

2464 2465 2466 2467
	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
2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483

	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;
2484 2485
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2486

2487 2488 2489
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_forwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2490 2491 2492 2493
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2494

2495
	return print_irq_inversion_bug(curr, &root, target_entry,
2496
					this, 1, irqclass);
I
Ingo Molnar 已提交
2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507
}

/*
 * 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;
2508 2509
	struct lock_list root;
	struct lock_list *uninitialized_var(target_entry);
I
Ingo Molnar 已提交
2510

2511 2512 2513
	root.parent = NULL;
	root.class = hlock_class(this);
	ret = find_usage_backwards(&root, bit, &target_entry);
P
Peter Zijlstra 已提交
2514 2515 2516 2517
	if (ret < 0)
		return print_bfs_bug(ret);
	if (ret == 1)
		return ret;
I
Ingo Molnar 已提交
2518

2519
	return print_irq_inversion_bug(curr, &root, target_entry,
2520
					this, 0, irqclass);
I
Ingo Molnar 已提交
2521 2522
}

2523
void print_irqtrace_events(struct task_struct *curr)
I
Ingo Molnar 已提交
2524 2525
{
	printk("irq event stamp: %u\n", curr->irq_events);
D
Dmitry Vyukov 已提交
2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537
	printk("hardirqs last  enabled at (%u): [<%p>] %pS\n",
		curr->hardirq_enable_event, (void *)curr->hardirq_enable_ip,
		(void *)curr->hardirq_enable_ip);
	printk("hardirqs last disabled at (%u): [<%p>] %pS\n",
		curr->hardirq_disable_event, (void *)curr->hardirq_disable_ip,
		(void *)curr->hardirq_disable_ip);
	printk("softirqs last  enabled at (%u): [<%p>] %pS\n",
		curr->softirq_enable_event, (void *)curr->softirq_enable_ip,
		(void *)curr->softirq_enable_ip);
	printk("softirqs last disabled at (%u): [<%p>] %pS\n",
		curr->softirq_disable_event, (void *)curr->softirq_disable_ip,
		(void *)curr->softirq_disable_ip);
I
Ingo Molnar 已提交
2538 2539
}

2540
static int HARDIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2541
{
P
Peter Zijlstra 已提交
2542 2543 2544
#if HARDIRQ_VERBOSE
	return class_filter(class);
#endif
I
Ingo Molnar 已提交
2545 2546 2547
	return 0;
}

2548
static int SOFTIRQ_verbose(struct lock_class *class)
I
Ingo Molnar 已提交
2549
{
P
Peter Zijlstra 已提交
2550 2551 2552 2553
#if SOFTIRQ_VERBOSE
	return class_filter(class);
#endif
	return 0;
I
Ingo Molnar 已提交
2554 2555
}

2556
static int RECLAIM_FS_verbose(struct lock_class *class)
2557 2558 2559 2560 2561 2562 2563
{
#if RECLAIM_VERBOSE
	return class_filter(class);
#endif
	return 0;
}

I
Ingo Molnar 已提交
2564 2565
#define STRICT_READ_CHECKS	1

2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578
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);
}

2579 2580 2581
typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
			     enum lock_usage_bit bit, const char *name);

2582
static int
2583 2584
mark_lock_irq(struct task_struct *curr, struct held_lock *this,
		enum lock_usage_bit new_bit)
2585
{
2586
	int excl_bit = exclusive_bit(new_bit);
2587
	int read = new_bit & 1;
2588 2589
	int dir = new_bit & 2;

2590 2591 2592 2593 2594 2595 2596
	/*
	 * 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.
	 */
2597 2598
	check_usage_f usage = dir ?
		check_usage_backwards : check_usage_forwards;
2599

2600 2601 2602 2603
	/*
	 * Validate that this particular lock does not have conflicting
	 * usage states.
	 */
2604 2605
	if (!valid_state(curr, this, new_bit, excl_bit))
		return 0;
2606

2607 2608 2609 2610 2611
	/*
	 * Validate that the lock dependencies don't have conflicting usage
	 * states.
	 */
	if ((!read || !dir || STRICT_READ_CHECKS) &&
2612
			!usage(curr, this, excl_bit, state_name(new_bit & ~1)))
2613
		return 0;
2614

2615 2616 2617 2618 2619 2620 2621 2622
	/*
	 * Check for read in write conflicts
	 */
	if (!read) {
		if (!valid_state(curr, this, new_bit, excl_bit + 1))
			return 0;

		if (STRICT_READ_CHECKS &&
2623 2624
			!usage(curr, this, excl_bit + 1,
				state_name(new_bit + 1)))
2625 2626
			return 0;
	}
2627

2628
	if (state_verbose(new_bit, hlock_class(this)))
2629 2630 2631 2632 2633
		return 2;

	return 1;
}

2634
enum mark_type {
2635 2636 2637
#define LOCKDEP_STATE(__STATE)	__STATE,
#include "lockdep_states.h"
#undef LOCKDEP_STATE
2638 2639
};

I
Ingo Molnar 已提交
2640 2641 2642
/*
 * Mark all held locks with a usage bit:
 */
2643
static int
2644
mark_held_locks(struct task_struct *curr, enum mark_type mark)
I
Ingo Molnar 已提交
2645 2646 2647 2648 2649 2650 2651 2652
{
	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;

2653 2654 2655 2656 2657
		usage_bit = 2 + (mark << 2); /* ENABLED */
		if (hlock->read)
			usage_bit += 1; /* READ */

		BUG_ON(usage_bit >= LOCK_USAGE_STATES);
2658

2659
		if (!hlock->check)
2660 2661
			continue;

2662
		if (!mark_lock(curr, hlock, usage_bit))
I
Ingo Molnar 已提交
2663 2664 2665 2666 2667 2668 2669 2670 2671
			return 0;
	}

	return 1;
}

/*
 * Hardirqs will be enabled:
 */
2672
static void __trace_hardirqs_on_caller(unsigned long ip)
I
Ingo Molnar 已提交
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682
{
	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:
	 */
2683
	if (!mark_held_locks(curr, HARDIRQ))
I
Ingo Molnar 已提交
2684 2685 2686 2687 2688 2689 2690
		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)
2691
		if (!mark_held_locks(curr, SOFTIRQ))
I
Ingo Molnar 已提交
2692 2693
			return;

P
Peter Zijlstra 已提交
2694 2695
	curr->hardirq_enable_ip = ip;
	curr->hardirq_enable_event = ++curr->irq_events;
2696
	debug_atomic_inc(hardirqs_on_events);
P
Peter Zijlstra 已提交
2697
}
2698

2699
__visible void trace_hardirqs_on_caller(unsigned long ip)
2700 2701 2702 2703 2704 2705
{
	time_hardirqs_on(CALLER_ADDR0, ip);

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

2706 2707 2708 2709 2710 2711 2712 2713 2714 2715
	if (unlikely(current->hardirqs_enabled)) {
		/*
		 * Neither irq nor preemption are disabled here
		 * so this is racy by nature but losing one hit
		 * in a stat is not a big deal.
		 */
		__debug_atomic_inc(redundant_hardirqs_on);
		return;
	}

P
Peter Zijlstra 已提交
2716 2717 2718 2719 2720
	/*
	 * We're enabling irqs and according to our state above irqs weren't
	 * already enabled, yet we find the hardware thinks they are in fact
	 * enabled.. someone messed up their IRQ state tracing.
	 */
2721 2722 2723
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

P
Peter Zijlstra 已提交
2724 2725 2726
	/*
	 * See the fine text that goes along with this variable definition.
	 */
2727 2728 2729
	if (DEBUG_LOCKS_WARN_ON(unlikely(early_boot_irqs_disabled)))
		return;

P
Peter Zijlstra 已提交
2730 2731 2732 2733
	/*
	 * Can't allow enabling interrupts while in an interrupt handler,
	 * that's general bad form and such. Recursion, limited stack etc..
	 */
2734 2735 2736
	if (DEBUG_LOCKS_WARN_ON(current->hardirq_context))
		return;

2737 2738 2739 2740
	current->lockdep_recursion = 1;
	__trace_hardirqs_on_caller(ip);
	current->lockdep_recursion = 0;
}
2741
EXPORT_SYMBOL(trace_hardirqs_on_caller);
P
Peter Zijlstra 已提交
2742

2743
void trace_hardirqs_on(void)
2744 2745 2746
{
	trace_hardirqs_on_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2747 2748 2749 2750 2751
EXPORT_SYMBOL(trace_hardirqs_on);

/*
 * Hardirqs were disabled:
 */
2752
__visible void trace_hardirqs_off_caller(unsigned long ip)
P
Peter Zijlstra 已提交
2753 2754 2755
{
	struct task_struct *curr = current;

2756
	time_hardirqs_off(CALLER_ADDR0, ip);
2757

P
Peter Zijlstra 已提交
2758 2759 2760
	if (unlikely(!debug_locks || current->lockdep_recursion))
		return;

P
Peter Zijlstra 已提交
2761 2762 2763 2764
	/*
	 * So we're supposed to get called after you mask local IRQs, but for
	 * some reason the hardware doesn't quite think you did a proper job.
	 */
P
Peter Zijlstra 已提交
2765 2766 2767 2768 2769 2770 2771 2772
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->hardirqs_enabled) {
		/*
		 * We have done an ON -> OFF transition:
		 */
		curr->hardirqs_enabled = 0;
2773
		curr->hardirq_disable_ip = ip;
P
Peter Zijlstra 已提交
2774
		curr->hardirq_disable_event = ++curr->irq_events;
2775
		debug_atomic_inc(hardirqs_off_events);
P
Peter Zijlstra 已提交
2776
	} else
2777
		debug_atomic_inc(redundant_hardirqs_off);
P
Peter Zijlstra 已提交
2778
}
2779
EXPORT_SYMBOL(trace_hardirqs_off_caller);
P
Peter Zijlstra 已提交
2780

2781
void trace_hardirqs_off(void)
2782 2783 2784
{
	trace_hardirqs_off_caller(CALLER_ADDR0);
}
P
Peter Zijlstra 已提交
2785 2786 2787 2788 2789 2790 2791 2792 2793
EXPORT_SYMBOL(trace_hardirqs_off);

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

2794
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
2795 2796
		return;

P
Peter Zijlstra 已提交
2797 2798 2799 2800
	/*
	 * We fancy IRQs being disabled here, see softirq.c, avoids
	 * funny state and nesting things.
	 */
P
Peter Zijlstra 已提交
2801 2802 2803 2804
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return;

	if (curr->softirqs_enabled) {
2805
		debug_atomic_inc(redundant_softirqs_on);
P
Peter Zijlstra 已提交
2806 2807 2808
		return;
	}

2809
	current->lockdep_recursion = 1;
P
Peter Zijlstra 已提交
2810 2811 2812 2813 2814 2815
	/*
	 * We'll do an OFF -> ON transition:
	 */
	curr->softirqs_enabled = 1;
	curr->softirq_enable_ip = ip;
	curr->softirq_enable_event = ++curr->irq_events;
2816
	debug_atomic_inc(softirqs_on_events);
P
Peter Zijlstra 已提交
2817 2818 2819 2820 2821 2822
	/*
	 * 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)
2823
		mark_held_locks(curr, SOFTIRQ);
2824
	current->lockdep_recursion = 0;
P
Peter Zijlstra 已提交
2825 2826 2827 2828 2829 2830 2831 2832 2833
}

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

2834
	if (unlikely(!debug_locks || current->lockdep_recursion))
P
Peter Zijlstra 已提交
2835 2836
		return;

P
Peter Zijlstra 已提交
2837 2838 2839
	/*
	 * We fancy IRQs being disabled here, see softirq.c
	 */
P
Peter Zijlstra 已提交
2840 2841 2842 2843 2844 2845 2846 2847 2848 2849
	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;
2850
		debug_atomic_inc(softirqs_off_events);
P
Peter Zijlstra 已提交
2851 2852 2853
		/*
		 * Whoops, we wanted softirqs off, so why aren't they?
		 */
P
Peter Zijlstra 已提交
2854 2855
		DEBUG_LOCKS_WARN_ON(!softirq_count());
	} else
2856
		debug_atomic_inc(redundant_softirqs_off);
P
Peter Zijlstra 已提交
2857 2858
}

2859
static void __lockdep_trace_alloc(gfp_t gfp_mask, unsigned long flags)
2860 2861 2862 2863 2864 2865 2866
{
	struct task_struct *curr = current;

	if (unlikely(!debug_locks))
		return;

	/* no reclaim without waiting on it */
2867
	if (!(gfp_mask & __GFP_DIRECT_RECLAIM))
2868 2869 2870 2871 2872 2873 2874 2875 2876 2877
		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;

P
Peter Zijlstra 已提交
2878 2879 2880
	/*
	 * Oi! Can't be having __GFP_FS allocations with IRQs disabled.
	 */
2881
	if (DEBUG_LOCKS_WARN_ON(irqs_disabled_flags(flags)))
2882 2883 2884 2885 2886
		return;

	mark_held_locks(curr, RECLAIM_FS);
}

2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903
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
2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931
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
2932
					LOCK_ENABLED_HARDIRQ_READ))
P
Peter Zijlstra 已提交
2933 2934 2935
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
2936
						LOCK_ENABLED_SOFTIRQ_READ))
P
Peter Zijlstra 已提交
2937 2938 2939
					return 0;
		} else {
			if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
2940
					LOCK_ENABLED_HARDIRQ))
P
Peter Zijlstra 已提交
2941 2942 2943
				return 0;
			if (curr->softirqs_enabled)
				if (!mark_lock(curr, hlock,
P
Peter Zijlstra 已提交
2944
						LOCK_ENABLED_SOFTIRQ))
P
Peter Zijlstra 已提交
2945 2946 2947 2948
					return 0;
		}
	}

2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964
	/*
	 * 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
2965 2966 2967
	return 1;
}

2968 2969 2970 2971 2972
static inline unsigned int task_irq_context(struct task_struct *task)
{
	return 2 * !!task->hardirq_context + !!task->softirq_context;
}

P
Peter Zijlstra 已提交
2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993
static int separate_irq_context(struct task_struct *curr,
		struct held_lock *hlock)
{
	unsigned int depth = curr->lockdep_depth;

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

		prev_hlock = curr->held_locks + depth-1;
		/*
		 * If we cross into another context, reset the
		 * hash key (this also prevents the checking and the
		 * adding of the dependency to 'prev'):
		 */
		if (prev_hlock->irq_context != hlock->irq_context)
			return 1;
	}
	return 0;
I
Ingo Molnar 已提交
2994 2995
}

P
Peter Zijlstra 已提交
2996
#else /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
I
Ingo Molnar 已提交
2997

P
Peter Zijlstra 已提交
2998 2999 3000
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
3001
{
P
Peter Zijlstra 已提交
3002
	WARN_ON(1); /* Impossible innit? when we don't have TRACE_IRQFLAG */
P
Peter Zijlstra 已提交
3003 3004
	return 1;
}
I
Ingo Molnar 已提交
3005

P
Peter Zijlstra 已提交
3006 3007 3008 3009 3010
static inline int mark_irqflags(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 1;
}
I
Ingo Molnar 已提交
3011

3012 3013 3014 3015 3016
static inline unsigned int task_irq_context(struct task_struct *task)
{
	return 0;
}

P
Peter Zijlstra 已提交
3017 3018 3019 3020
static inline int separate_irq_context(struct task_struct *curr,
		struct held_lock *hlock)
{
	return 0;
I
Ingo Molnar 已提交
3021 3022
}

3023 3024 3025 3026
void lockdep_trace_alloc(gfp_t gfp_mask)
{
}

P
Peter Zijlstra 已提交
3027
#endif /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
I
Ingo Molnar 已提交
3028 3029

/*
P
Peter Zijlstra 已提交
3030
 * Mark a lock with a usage bit, and validate the state transition:
I
Ingo Molnar 已提交
3031
 */
3032
static int mark_lock(struct task_struct *curr, struct held_lock *this,
3033
			     enum lock_usage_bit new_bit)
I
Ingo Molnar 已提交
3034
{
P
Peter Zijlstra 已提交
3035
	unsigned int new_mask = 1 << new_bit, ret = 1;
I
Ingo Molnar 已提交
3036 3037

	/*
P
Peter Zijlstra 已提交
3038 3039
	 * If already set then do not dirty the cacheline,
	 * nor do any checks:
I
Ingo Molnar 已提交
3040
	 */
D
Dave Jones 已提交
3041
	if (likely(hlock_class(this)->usage_mask & new_mask))
P
Peter Zijlstra 已提交
3042 3043 3044 3045
		return 1;

	if (!graph_lock())
		return 0;
I
Ingo Molnar 已提交
3046
	/*
L
Lucas De Marchi 已提交
3047
	 * Make sure we didn't race:
I
Ingo Molnar 已提交
3048
	 */
D
Dave Jones 已提交
3049
	if (unlikely(hlock_class(this)->usage_mask & new_mask)) {
P
Peter Zijlstra 已提交
3050 3051 3052
		graph_unlock();
		return 1;
	}
I
Ingo Molnar 已提交
3053

D
Dave Jones 已提交
3054
	hlock_class(this)->usage_mask |= new_mask;
I
Ingo Molnar 已提交
3055

D
Dave Jones 已提交
3056
	if (!save_trace(hlock_class(this)->usage_traces + new_bit))
P
Peter Zijlstra 已提交
3057
		return 0;
I
Ingo Molnar 已提交
3058

P
Peter Zijlstra 已提交
3059
	switch (new_bit) {
P
Peter Zijlstra 已提交
3060 3061 3062 3063 3064 3065 3066
#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
3067 3068 3069 3070 3071
		ret = mark_lock_irq(curr, this, new_bit);
		if (!ret)
			return 0;
		break;
	case LOCK_USED:
3072
		debug_atomic_dec(nr_unused_locks);
P
Peter Zijlstra 已提交
3073 3074 3075 3076 3077 3078 3079
		break;
	default:
		if (!debug_locks_off_graph_unlock())
			return 0;
		WARN_ON(1);
		return 0;
	}
I
Ingo Molnar 已提交
3080

P
Peter Zijlstra 已提交
3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094
	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
3095 3096 3097 3098 3099

/*
 * Initialize a lock instance's lock-class mapping info:
 */
void lockdep_init_map(struct lockdep_map *lock, const char *name,
3100
		      struct lock_class_key *key, int subclass)
I
Ingo Molnar 已提交
3101
{
3102 3103 3104 3105 3106 3107
	int i;

	kmemcheck_mark_initialized(lock, sizeof(*lock));

	for (i = 0; i < NR_LOCKDEP_CACHING_CLASSES; i++)
		lock->class_cache[i] = NULL;
3108

3109 3110 3111 3112
#ifdef CONFIG_LOCK_STAT
	lock->cpu = raw_smp_processor_id();
#endif

P
Peter Zijlstra 已提交
3113 3114 3115
	/*
	 * Can't be having no nameless bastards around this place!
	 */
3116 3117
	if (DEBUG_LOCKS_WARN_ON(!name)) {
		lock->name = "NULL";
I
Ingo Molnar 已提交
3118
		return;
3119 3120 3121
	}

	lock->name = name;
I
Ingo Molnar 已提交
3122

P
Peter Zijlstra 已提交
3123 3124 3125
	/*
	 * No key, no joy, we need to hash something.
	 */
I
Ingo Molnar 已提交
3126 3127 3128 3129 3130 3131 3132
	if (DEBUG_LOCKS_WARN_ON(!key))
		return;
	/*
	 * Sanity check, the lock-class key must be persistent:
	 */
	if (!static_obj(key)) {
		printk("BUG: key %p not in .data!\n", key);
P
Peter Zijlstra 已提交
3133 3134 3135
		/*
		 * What it says above ^^^^^, I suggest you read it.
		 */
I
Ingo Molnar 已提交
3136 3137 3138 3139
		DEBUG_LOCKS_WARN_ON(1);
		return;
	}
	lock->key = key;
3140 3141 3142 3143

	if (unlikely(!debug_locks))
		return;

3144 3145 3146 3147 3148 3149 3150 3151
	if (subclass) {
		unsigned long flags;

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

		raw_local_irq_save(flags);
		current->lockdep_recursion = 1;
3152
		register_lock_class(lock, subclass, 1);
3153 3154 3155
		current->lockdep_recursion = 0;
		raw_local_irq_restore(flags);
	}
I
Ingo Molnar 已提交
3156 3157 3158
}
EXPORT_SYMBOL_GPL(lockdep_init_map);

3159
struct lock_class_key __lockdep_no_validate__;
K
Kent Overstreet 已提交
3160
EXPORT_SYMBOL_GPL(__lockdep_no_validate__);
3161

3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195
static int
print_lock_nested_lock_not_held(struct task_struct *curr,
				struct held_lock *hlock,
				unsigned long ip)
{
	if (!debug_locks_off())
		return 0;
	if (debug_locks_silent)
		return 0;

	printk("\n");
	printk("==================================\n");
	printk("[ BUG: Nested lock was not taken ]\n");
	print_kernel_ident();
	printk("----------------------------------\n");

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

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

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

	printk("\nother info that might help us debug this:\n");
	lockdep_print_held_locks(curr);

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

	return 0;
}

3196
static int __lock_is_held(struct lockdep_map *lock, int read);
3197

I
Ingo Molnar 已提交
3198 3199 3200 3201 3202 3203
/*
 * 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,
3204
			  struct lockdep_map *nest_lock, unsigned long ip,
3205
			  int references, int pin_count)
I
Ingo Molnar 已提交
3206 3207
{
	struct task_struct *curr = current;
3208
	struct lock_class *class = NULL;
I
Ingo Molnar 已提交
3209
	struct held_lock *hlock;
3210
	unsigned int depth;
I
Ingo Molnar 已提交
3211
	int chain_head = 0;
3212
	int class_idx;
I
Ingo Molnar 已提交
3213 3214 3215 3216 3217
	u64 chain_key;

	if (unlikely(!debug_locks))
		return 0;

P
Peter Zijlstra 已提交
3218 3219 3220 3221 3222
	/*
	 * Lockdep should run with IRQs disabled, otherwise we could
	 * get an interrupt which would want to take locks, which would
	 * end up in lockdep and have you got a head-ache already?
	 */
I
Ingo Molnar 已提交
3223 3224 3225
	if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
		return 0;

3226 3227
	if (!prove_locking || lock->key == &__lockdep_no_validate__)
		check = 0;
3228

3229 3230
	if (subclass < NR_LOCKDEP_CACHING_CLASSES)
		class = lock->class_cache[subclass];
3231
	/*
3232
	 * Not cached?
3233
	 */
I
Ingo Molnar 已提交
3234
	if (unlikely(!class)) {
3235
		class = register_lock_class(lock, subclass, 0);
I
Ingo Molnar 已提交
3236 3237 3238
		if (!class)
			return 0;
	}
3239
	atomic_inc((atomic_t *)&class->ops);
I
Ingo Molnar 已提交
3240 3241 3242
	if (very_verbose(class)) {
		printk("\nacquire class [%p] %s", class->key, class->name);
		if (class->name_version > 1)
D
Dmitry Vyukov 已提交
3243 3244
			printk(KERN_CONT "#%d", class->name_version);
		printk(KERN_CONT "\n");
I
Ingo Molnar 已提交
3245 3246 3247 3248 3249 3250 3251 3252 3253
		dump_stack();
	}

	/*
	 * Add the lock to the list of currently held locks.
	 * (we dont increase the depth just yet, up until the
	 * dependency checks are done)
	 */
	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3254 3255 3256
	/*
	 * Ran out of static storage for our per-task lock stack again have we?
	 */
I
Ingo Molnar 已提交
3257 3258 3259
	if (DEBUG_LOCKS_WARN_ON(depth >= MAX_LOCK_DEPTH))
		return 0;

3260 3261 3262 3263 3264
	class_idx = class - lock_classes + 1;

	if (depth) {
		hlock = curr->held_locks + depth - 1;
		if (hlock->class_idx == class_idx && nest_lock) {
3265 3266 3267 3268 3269 3270 3271
			if (hlock->references) {
				/*
				 * Check: unsigned int references:12, overflow.
				 */
				if (DEBUG_LOCKS_WARN_ON(hlock->references == (1 << 12)-1))
					return 0;

3272
				hlock->references++;
3273
			} else {
3274
				hlock->references = 2;
3275
			}
3276 3277 3278 3279 3280

			return 1;
		}
	}

I
Ingo Molnar 已提交
3281
	hlock = curr->held_locks + depth;
P
Peter Zijlstra 已提交
3282 3283 3284 3285
	/*
	 * Plain impossible, we just registered it and checked it weren't no
	 * NULL like.. I bet this mushroom I ate was good!
	 */
D
Dave Jones 已提交
3286 3287
	if (DEBUG_LOCKS_WARN_ON(!class))
		return 0;
3288
	hlock->class_idx = class_idx;
I
Ingo Molnar 已提交
3289 3290
	hlock->acquire_ip = ip;
	hlock->instance = lock;
P
Peter Zijlstra 已提交
3291
	hlock->nest_lock = nest_lock;
3292
	hlock->irq_context = task_irq_context(curr);
I
Ingo Molnar 已提交
3293 3294 3295
	hlock->trylock = trylock;
	hlock->read = read;
	hlock->check = check;
3296
	hlock->hardirqs_off = !!hardirqs_off;
3297
	hlock->references = references;
P
Peter Zijlstra 已提交
3298 3299
#ifdef CONFIG_LOCK_STAT
	hlock->waittime_stamp = 0;
3300
	hlock->holdtime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
3301
#endif
3302
	hlock->pin_count = pin_count;
I
Ingo Molnar 已提交
3303

3304
	if (check && !mark_irqflags(curr, hlock))
P
Peter Zijlstra 已提交
3305 3306
		return 0;

I
Ingo Molnar 已提交
3307
	/* mark it as used: */
3308
	if (!mark_lock(curr, hlock, LOCK_USED))
I
Ingo Molnar 已提交
3309
		return 0;
P
Peter Zijlstra 已提交
3310

I
Ingo Molnar 已提交
3311
	/*
3312
	 * Calculate the chain hash: it's the combined hash of all the
I
Ingo Molnar 已提交
3313 3314 3315 3316 3317 3318 3319 3320
	 * lock keys along the dependency chain. We save the hash value
	 * at every step so that we can get the current hash easily
	 * after unlock. The chain hash is then used to cache dependency
	 * results.
	 *
	 * The 'key ID' is what is the most compact key value to drive
	 * the hash, not class->key.
	 */
P
Peter Zijlstra 已提交
3321 3322 3323
	/*
	 * Whoops, we did it again.. ran straight out of our static allocation.
	 */
3324
	if (DEBUG_LOCKS_WARN_ON(class_idx > MAX_LOCKDEP_KEYS))
I
Ingo Molnar 已提交
3325 3326 3327 3328
		return 0;

	chain_key = curr->curr_chain_key;
	if (!depth) {
P
Peter Zijlstra 已提交
3329 3330 3331
		/*
		 * How can we have a chain hash when we ain't got no keys?!
		 */
I
Ingo Molnar 已提交
3332 3333 3334 3335 3336 3337
		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
3338 3339 3340
	if (separate_irq_context(curr, hlock)) {
		chain_key = 0;
		chain_head = 1;
I
Ingo Molnar 已提交
3341
	}
3342
	chain_key = iterate_chain_key(chain_key, class_idx);
I
Ingo Molnar 已提交
3343

3344
	if (nest_lock && !__lock_is_held(nest_lock, -1))
3345 3346
		return print_lock_nested_lock_not_held(curr, hlock, ip);

3347
	if (!validate_chain(curr, lock, hlock, chain_head, chain_key))
P
Peter Zijlstra 已提交
3348
		return 0;
3349

3350
	curr->curr_chain_key = chain_key;
I
Ingo Molnar 已提交
3351 3352
	curr->lockdep_depth++;
	check_chain_key(curr);
3353 3354 3355 3356
#ifdef CONFIG_DEBUG_LOCKDEP
	if (unlikely(!debug_locks))
		return 0;
#endif
I
Ingo Molnar 已提交
3357 3358
	if (unlikely(curr->lockdep_depth >= MAX_LOCK_DEPTH)) {
		debug_locks_off();
3359 3360
		print_lockdep_off("BUG: MAX_LOCK_DEPTH too low!");
		printk(KERN_DEBUG "depth: %i  max: %lu!\n",
3361 3362 3363 3364
		       curr->lockdep_depth, MAX_LOCK_DEPTH);

		lockdep_print_held_locks(current);
		debug_show_all_locks();
3365
		dump_stack();
3366

I
Ingo Molnar 已提交
3367 3368
		return 0;
	}
3369

I
Ingo Molnar 已提交
3370 3371 3372 3373 3374 3375 3376
	if (unlikely(curr->lockdep_depth > max_lockdep_depth))
		max_lockdep_depth = curr->lockdep_depth;

	return 1;
}

static int
3377
print_unlock_imbalance_bug(struct task_struct *curr, struct lockdep_map *lock,
I
Ingo Molnar 已提交
3378 3379 3380 3381 3382 3383 3384
			   unsigned long ip)
{
	if (!debug_locks_off())
		return 0;
	if (debug_locks_silent)
		return 0;

3385 3386 3387
	printk("\n");
	printk("=====================================\n");
	printk("[ BUG: bad unlock balance detected! ]\n");
3388
	print_kernel_ident();
3389
	printk("-------------------------------------\n");
I
Ingo Molnar 已提交
3390
	printk("%s/%d is trying to release lock (",
3391
		curr->comm, task_pid_nr(curr));
I
Ingo Molnar 已提交
3392
	print_lockdep_cache(lock);
D
Dmitry Vyukov 已提交
3393
	printk(KERN_CONT ") at:\n");
I
Ingo Molnar 已提交
3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404
	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;
}

3405 3406 3407 3408 3409 3410
static int match_held_lock(struct held_lock *hlock, struct lockdep_map *lock)
{
	if (hlock->instance == lock)
		return 1;

	if (hlock->references) {
3411
		struct lock_class *class = lock->class_cache[0];
3412 3413 3414 3415

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

3416 3417 3418 3419 3420 3421 3422
		/*
		 * If look_up_lock_class() failed to find a class, we're trying
		 * to test if we hold a lock that has never yet been acquired.
		 * Clearly if the lock hasn't been acquired _ever_, we're not
		 * holding it either, so report failure.
		 */
		if (!class)
3423 3424
			return 0;

P
Peter Zijlstra 已提交
3425 3426 3427 3428 3429
		/*
		 * References, but not a lock we're actually ref-counting?
		 * State got messed up, follow the sites that change ->references
		 * and try to make sense of it.
		 */
3430 3431 3432 3433 3434 3435 3436 3437 3438 3439
		if (DEBUG_LOCKS_WARN_ON(!hlock->nest_lock))
			return 0;

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

	return 0;
}

3440
static int
3441 3442 3443
__lock_set_class(struct lockdep_map *lock, const char *name,
		 struct lock_class_key *key, unsigned int subclass,
		 unsigned long ip)
3444 3445 3446 3447 3448 3449 3450 3451
{
	struct task_struct *curr = current;
	struct held_lock *hlock, *prev_hlock;
	struct lock_class *class;
	unsigned int depth;
	int i;

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3452 3453 3454 3455
	/*
	 * This function is about (re)setting the class of a held lock,
	 * yet we're not actually holding any locks. Naughty user!
	 */
3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466
	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;
3467
		if (match_held_lock(hlock, lock))
3468 3469 3470
			goto found_it;
		prev_hlock = hlock;
	}
3471
	return print_unlock_imbalance_bug(curr, lock, ip);
3472 3473

found_it:
3474
	lockdep_init_map(lock, name, key, 0);
3475
	class = register_lock_class(lock, subclass, 0);
D
Dave Jones 已提交
3476
	hlock->class_idx = class - lock_classes + 1;
3477 3478 3479 3480 3481 3482 3483

	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
3484
			hlock_class(hlock)->subclass, hlock->trylock,
3485
				hlock->read, hlock->check, hlock->hardirqs_off,
3486
				hlock->nest_lock, hlock->acquire_ip,
3487
				hlock->references, hlock->pin_count))
3488 3489 3490
			return 0;
	}

P
Peter Zijlstra 已提交
3491 3492 3493 3494
	/*
	 * I took it apart and put it back together again, except now I have
	 * these 'spare' parts.. where shall I put them.
	 */
3495 3496 3497 3498 3499
	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth))
		return 0;
	return 1;
}

I
Ingo Molnar 已提交
3500
/*
3501 3502 3503 3504 3505
 * Remove the lock to the list of currently held locks - this gets
 * called on mutex_unlock()/spin_unlock*() (or on a failed
 * mutex_lock_interruptible()).
 *
 * @nested is an hysterical artifact, needs a tree wide cleanup.
I
Ingo Molnar 已提交
3506 3507
 */
static int
3508
__lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
I
Ingo Molnar 已提交
3509
{
3510
	struct task_struct *curr = current;
I
Ingo Molnar 已提交
3511 3512 3513 3514
	struct held_lock *hlock, *prev_hlock;
	unsigned int depth;
	int i;

3515 3516 3517
	if (unlikely(!debug_locks))
		return 0;

I
Ingo Molnar 已提交
3518
	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3519 3520 3521 3522
	/*
	 * So we're all set to release this lock.. wait what lock? We don't
	 * own any locks, you've been drinking again?
	 */
3523 3524
	if (DEBUG_LOCKS_WARN_ON(depth <= 0))
		 return print_unlock_imbalance_bug(curr, lock, ip);
I
Ingo Molnar 已提交
3525

3526 3527 3528 3529
	/*
	 * Check whether the lock exists in the current stack
	 * of held locks:
	 */
I
Ingo Molnar 已提交
3530 3531 3532 3533 3534 3535 3536 3537
	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;
3538
		if (match_held_lock(hlock, lock))
I
Ingo Molnar 已提交
3539 3540 3541
			goto found_it;
		prev_hlock = hlock;
	}
3542
	return print_unlock_imbalance_bug(curr, lock, ip);
I
Ingo Molnar 已提交
3543 3544

found_it:
3545 3546 3547
	if (hlock->instance == lock)
		lock_release_holdtime(hlock);

P
Peter Zijlstra 已提交
3548 3549
	WARN(hlock->pin_count, "releasing a pinned lock\n");

3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560
	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
3561

I
Ingo Molnar 已提交
3562 3563 3564 3565 3566
	/*
	 * 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:
	 */
3567

I
Ingo Molnar 已提交
3568 3569 3570 3571 3572 3573
	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
3574
			hlock_class(hlock)->subclass, hlock->trylock,
I
Ingo Molnar 已提交
3575
				hlock->read, hlock->check, hlock->hardirqs_off,
3576
				hlock->nest_lock, hlock->acquire_ip,
3577
				hlock->references, hlock->pin_count))
I
Ingo Molnar 已提交
3578 3579 3580
			return 0;
	}

P
Peter Zijlstra 已提交
3581 3582 3583 3584
	/*
	 * We had N bottles of beer on the wall, we drank one, but now
	 * there's not N-1 bottles of beer left on the wall...
	 */
I
Ingo Molnar 已提交
3585 3586
	if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth - 1))
		return 0;
P
Peter Zijlstra 已提交
3587

I
Ingo Molnar 已提交
3588 3589 3590
	return 1;
}

3591
static int __lock_is_held(struct lockdep_map *lock, int read)
I
Ingo Molnar 已提交
3592
{
3593 3594
	struct task_struct *curr = current;
	int i;
I
Ingo Molnar 已提交
3595

3596
	for (i = 0; i < curr->lockdep_depth; i++) {
3597
		struct held_lock *hlock = curr->held_locks + i;
I
Ingo Molnar 已提交
3598

3599 3600 3601 3602 3603 3604
		if (match_held_lock(hlock, lock)) {
			if (read == -1 || hlock->read == read)
				return 1;

			return 0;
		}
3605
	}
P
Peter Zijlstra 已提交
3606

3607
	return 0;
I
Ingo Molnar 已提交
3608 3609
}

3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638
static struct pin_cookie __lock_pin_lock(struct lockdep_map *lock)
{
	struct pin_cookie cookie = NIL_COOKIE;
	struct task_struct *curr = current;
	int i;

	if (unlikely(!debug_locks))
		return cookie;

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

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

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

static void __lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
I
Ingo Molnar 已提交
3639 3640
{
	struct task_struct *curr = current;
P
Peter Zijlstra 已提交
3641
	int i;
I
Ingo Molnar 已提交
3642

P
Peter Zijlstra 已提交
3643
	if (unlikely(!debug_locks))
I
Ingo Molnar 已提交
3644 3645
		return;

P
Peter Zijlstra 已提交
3646 3647 3648 3649
	for (i = 0; i < curr->lockdep_depth; i++) {
		struct held_lock *hlock = curr->held_locks + i;

		if (match_held_lock(hlock, lock)) {
3650
			hlock->pin_count += cookie.val;
I
Ingo Molnar 已提交
3651
			return;
P
Peter Zijlstra 已提交
3652
		}
I
Ingo Molnar 已提交
3653 3654
	}

P
Peter Zijlstra 已提交
3655
	WARN(1, "pinning an unheld lock\n");
I
Ingo Molnar 已提交
3656 3657
}

3658
static void __lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
3659 3660 3661 3662
{
	struct task_struct *curr = current;
	int i;

P
Peter Zijlstra 已提交
3663 3664 3665
	if (unlikely(!debug_locks))
		return;

3666
	for (i = 0; i < curr->lockdep_depth; i++) {
3667 3668
		struct held_lock *hlock = curr->held_locks + i;

P
Peter Zijlstra 已提交
3669 3670 3671 3672
		if (match_held_lock(hlock, lock)) {
			if (WARN(!hlock->pin_count, "unpinning an unpinned lock\n"))
				return;

3673 3674 3675 3676 3677
			hlock->pin_count -= cookie.val;

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

P
Peter Zijlstra 已提交
3678 3679
			return;
		}
3680 3681
	}

P
Peter Zijlstra 已提交
3682
	WARN(1, "unpinning an unheld lock\n");
3683 3684
}

I
Ingo Molnar 已提交
3685 3686 3687
/*
 * Check whether we follow the irq-flags state precisely:
 */
3688
static void check_flags(unsigned long flags)
I
Ingo Molnar 已提交
3689
{
3690 3691
#if defined(CONFIG_PROVE_LOCKING) && defined(CONFIG_DEBUG_LOCKDEP) && \
    defined(CONFIG_TRACE_IRQFLAGS)
I
Ingo Molnar 已提交
3692 3693 3694
	if (!debug_locks)
		return;

3695 3696 3697 3698 3699 3700 3701 3702 3703
	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
3704 3705 3706 3707 3708 3709 3710

	/*
	 * We dont accurately track softirq state in e.g.
	 * hardirq contexts (such as on 4KSTACKS), so only
	 * check if not in hardirq contexts:
	 */
	if (!hardirq_count()) {
P
Peter Zijlstra 已提交
3711 3712
		if (softirq_count()) {
			/* like the above, but with softirqs */
I
Ingo Molnar 已提交
3713
			DEBUG_LOCKS_WARN_ON(current->softirqs_enabled);
P
Peter Zijlstra 已提交
3714 3715
		} else {
			/* lick the above, does it taste good? */
I
Ingo Molnar 已提交
3716
			DEBUG_LOCKS_WARN_ON(!current->softirqs_enabled);
P
Peter Zijlstra 已提交
3717
		}
I
Ingo Molnar 已提交
3718 3719 3720 3721 3722 3723 3724
	}

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

3725 3726 3727
void lock_set_class(struct lockdep_map *lock, const char *name,
		    struct lock_class_key *key, unsigned int subclass,
		    unsigned long ip)
3728 3729 3730 3731 3732 3733 3734 3735 3736
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	current->lockdep_recursion = 1;
	check_flags(flags);
3737
	if (__lock_set_class(lock, name, key, subclass, ip))
3738 3739 3740 3741
		check_chain_key(current);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
3742
EXPORT_SYMBOL_GPL(lock_set_class);
3743

I
Ingo Molnar 已提交
3744 3745 3746 3747
/*
 * We are not always called with irqs disabled - do that here,
 * and also avoid lockdep recursion:
 */
3748
void lock_acquire(struct lockdep_map *lock, unsigned int subclass,
P
Peter Zijlstra 已提交
3749 3750
			  int trylock, int read, int check,
			  struct lockdep_map *nest_lock, unsigned long ip)
I
Ingo Molnar 已提交
3751 3752 3753 3754 3755 3756 3757 3758 3759 3760
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
3761
	trace_lock_acquire(lock, subclass, trylock, read, check, nest_lock, ip);
I
Ingo Molnar 已提交
3762
	__lock_acquire(lock, subclass, trylock, read, check,
3763
		       irqs_disabled_flags(flags), nest_lock, ip, 0, 0);
I
Ingo Molnar 已提交
3764 3765 3766 3767 3768
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquire);

3769
void lock_release(struct lockdep_map *lock, int nested,
3770
			  unsigned long ip)
I
Ingo Molnar 已提交
3771 3772 3773 3774 3775 3776 3777 3778 3779
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);
	current->lockdep_recursion = 1;
3780
	trace_lock_release(lock, ip);
3781 3782
	if (__lock_release(lock, nested, ip))
		check_chain_key(current);
I
Ingo Molnar 已提交
3783 3784 3785 3786 3787
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_release);

3788
int lock_is_held_type(struct lockdep_map *lock, int read)
3789 3790 3791 3792 3793
{
	unsigned long flags;
	int ret = 0;

	if (unlikely(current->lockdep_recursion))
3794
		return 1; /* avoid false negative lockdep_assert_held() */
3795 3796 3797 3798 3799

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
3800
	ret = __lock_is_held(lock, read);
3801 3802 3803 3804 3805
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);

	return ret;
}
3806
EXPORT_SYMBOL_GPL(lock_is_held_type);
3807

3808
struct pin_cookie lock_pin_lock(struct lockdep_map *lock)
P
Peter Zijlstra 已提交
3809
{
3810
	struct pin_cookie cookie = NIL_COOKIE;
P
Peter Zijlstra 已提交
3811 3812 3813
	unsigned long flags;

	if (unlikely(current->lockdep_recursion))
3814
		return cookie;
P
Peter Zijlstra 已提交
3815 3816 3817 3818 3819

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
3820
	cookie = __lock_pin_lock(lock);
P
Peter Zijlstra 已提交
3821 3822
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
3823 3824

	return cookie;
P
Peter Zijlstra 已提交
3825 3826 3827
}
EXPORT_SYMBOL_GPL(lock_pin_lock);

3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845
void lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);

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

void lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
P
Peter Zijlstra 已提交
3846 3847 3848 3849 3850 3851 3852 3853 3854 3855
{
	unsigned long flags;

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

	raw_local_irq_save(flags);
	check_flags(flags);

	current->lockdep_recursion = 1;
3856
	__lock_unpin_lock(lock, cookie);
P
Peter Zijlstra 已提交
3857 3858 3859 3860 3861
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_unpin_lock);

3862 3863 3864 3865 3866 3867 3868 3869 3870 3871
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
3872 3873 3874 3875 3876 3877 3878 3879 3880 3881
#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;

3882 3883 3884
	printk("\n");
	printk("=================================\n");
	printk("[ BUG: bad contention detected! ]\n");
3885
	print_kernel_ident();
3886
	printk("---------------------------------\n");
P
Peter Zijlstra 已提交
3887
	printk("%s/%d is trying to contend lock (",
3888
		curr->comm, task_pid_nr(curr));
P
Peter Zijlstra 已提交
3889
	print_lockdep_cache(lock);
D
Dmitry Vyukov 已提交
3890
	printk(KERN_CONT ") at:\n");
P
Peter Zijlstra 已提交
3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908
	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
3909
	int i, contention_point, contending_point;
P
Peter Zijlstra 已提交
3910 3911

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3912 3913 3914 3915
	/*
	 * Whee, we contended on this lock, except it seems we're not
	 * actually trying to acquire anything much at all..
	 */
P
Peter Zijlstra 已提交
3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926
	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;
3927
		if (match_held_lock(hlock, lock))
P
Peter Zijlstra 已提交
3928 3929 3930 3931 3932 3933 3934
			goto found_it;
		prev_hlock = hlock;
	}
	print_lock_contention_bug(curr, lock, ip);
	return;

found_it:
3935 3936 3937
	if (hlock->instance != lock)
		return;

3938
	hlock->waittime_stamp = lockstat_clock();
P
Peter Zijlstra 已提交
3939

P
Peter Zijlstra 已提交
3940 3941 3942
	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
3943

D
Dave Jones 已提交
3944
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
3945 3946 3947 3948
	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
3949 3950
	if (lock->cpu != smp_processor_id())
		stats->bounces[bounce_contended + !!hlock->read]++;
P
Peter Zijlstra 已提交
3951 3952 3953 3954
	put_lock_stats(stats);
}

static void
P
Peter Zijlstra 已提交
3955
__lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
3956 3957 3958 3959 3960
{
	struct task_struct *curr = current;
	struct held_lock *hlock, *prev_hlock;
	struct lock_class_stats *stats;
	unsigned int depth;
3961
	u64 now, waittime = 0;
P
Peter Zijlstra 已提交
3962
	int i, cpu;
P
Peter Zijlstra 已提交
3963 3964

	depth = curr->lockdep_depth;
P
Peter Zijlstra 已提交
3965 3966 3967 3968
	/*
	 * Yay, we acquired ownership of this lock we didn't try to
	 * acquire, how the heck did that happen?
	 */
P
Peter Zijlstra 已提交
3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979
	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;
3980
		if (match_held_lock(hlock, lock))
P
Peter Zijlstra 已提交
3981 3982 3983 3984 3985 3986 3987
			goto found_it;
		prev_hlock = hlock;
	}
	print_lock_contention_bug(curr, lock, _RET_IP_);
	return;

found_it:
3988 3989 3990
	if (hlock->instance != lock)
		return;

P
Peter Zijlstra 已提交
3991 3992
	cpu = smp_processor_id();
	if (hlock->waittime_stamp) {
3993
		now = lockstat_clock();
P
Peter Zijlstra 已提交
3994 3995 3996
		waittime = now - hlock->waittime_stamp;
		hlock->holdtime_stamp = now;
	}
P
Peter Zijlstra 已提交
3997

3998
	trace_lock_acquired(lock, ip);
3999

D
Dave Jones 已提交
4000
	stats = get_lock_stats(hlock_class(hlock));
P
Peter Zijlstra 已提交
4001 4002 4003 4004 4005 4006 4007 4008
	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
4009
	put_lock_stats(stats);
P
Peter Zijlstra 已提交
4010 4011

	lock->cpu = cpu;
P
Peter Zijlstra 已提交
4012
	lock->ip = ip;
P
Peter Zijlstra 已提交
4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027
}

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;
4028
	trace_lock_contended(lock, ip);
P
Peter Zijlstra 已提交
4029 4030 4031 4032 4033 4034
	__lock_contended(lock, ip);
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_contended);

P
Peter Zijlstra 已提交
4035
void lock_acquired(struct lockdep_map *lock, unsigned long ip)
P
Peter Zijlstra 已提交
4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047
{
	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
4048
	__lock_acquired(lock, ip);
P
Peter Zijlstra 已提交
4049 4050 4051 4052 4053 4054
	current->lockdep_recursion = 0;
	raw_local_irq_restore(flags);
}
EXPORT_SYMBOL_GPL(lock_acquired);
#endif

I
Ingo Molnar 已提交
4055 4056 4057 4058 4059 4060 4061 4062
/*
 * Used by the testsuite, sanitize the validator state
 * after a simulated failure:
 */

void lockdep_reset(void)
{
	unsigned long flags;
4063
	int i;
I
Ingo Molnar 已提交
4064 4065 4066 4067 4068 4069 4070 4071 4072 4073

	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;
4074
	for (i = 0; i < CHAINHASH_SIZE; i++)
4075
		INIT_HLIST_HEAD(chainhash_table + i);
I
Ingo Molnar 已提交
4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093
	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:
	 */
4094
	hlist_del_rcu(&class->hash_entry);
I
Ingo Molnar 已提交
4095 4096
	list_del_rcu(&class->lock_entry);

4097 4098
	RCU_INIT_POINTER(class->key, NULL);
	RCU_INIT_POINTER(class->name, NULL);
I
Ingo Molnar 已提交
4099 4100
}

4101
static inline int within(const void *addr, void *start, unsigned long size)
I
Ingo Molnar 已提交
4102 4103 4104 4105
{
	return addr >= start && addr < start + size;
}

4106 4107 4108 4109 4110 4111 4112 4113
/*
 * Used in module.c to remove lock classes from memory that is going to be
 * freed; and possibly re-used by other modules.
 *
 * We will have had one sync_sched() before getting here, so we're guaranteed
 * nobody will look up these exact classes -- they're properly dead but still
 * allocated.
 */
I
Ingo Molnar 已提交
4114 4115
void lockdep_free_key_range(void *start, unsigned long size)
{
4116
	struct lock_class *class;
4117
	struct hlist_head *head;
I
Ingo Molnar 已提交
4118 4119
	unsigned long flags;
	int i;
4120
	int locked;
I
Ingo Molnar 已提交
4121 4122

	raw_local_irq_save(flags);
4123
	locked = graph_lock();
I
Ingo Molnar 已提交
4124 4125 4126 4127 4128 4129

	/*
	 * Unhash all classes that were created by this module:
	 */
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
4130
		hlist_for_each_entry_rcu(class, head, hash_entry) {
I
Ingo Molnar 已提交
4131 4132
			if (within(class->key, start, size))
				zap_class(class);
4133 4134 4135
			else if (within(class->name, start, size))
				zap_class(class);
		}
I
Ingo Molnar 已提交
4136 4137
	}

4138 4139
	if (locked)
		graph_unlock();
I
Ingo Molnar 已提交
4140
	raw_local_irq_restore(flags);
4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154

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

	/*
	 * XXX at this point we could return the resources to the pool;
	 * instead we leak them. We would need to change to bitmap allocators
	 * instead of the linear allocators we have now.
	 */
I
Ingo Molnar 已提交
4155 4156 4157 4158
}

void lockdep_reset_lock(struct lockdep_map *lock)
{
4159
	struct lock_class *class;
4160
	struct hlist_head *head;
I
Ingo Molnar 已提交
4161 4162
	unsigned long flags;
	int i, j;
4163
	int locked;
I
Ingo Molnar 已提交
4164 4165 4166 4167

	raw_local_irq_save(flags);

	/*
4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180
	 * 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
4181
	 */
4182
	locked = graph_lock();
I
Ingo Molnar 已提交
4183 4184
	for (i = 0; i < CLASSHASH_SIZE; i++) {
		head = classhash_table + i;
4185
		hlist_for_each_entry_rcu(class, head, hash_entry) {
4186 4187 4188 4189 4190 4191
			int match = 0;

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

			if (unlikely(match)) {
P
Peter Zijlstra 已提交
4192 4193 4194 4195
				if (debug_locks_off_graph_unlock()) {
					/*
					 * We all just reset everything, how did it match?
					 */
4196
					WARN_ON(1);
P
Peter Zijlstra 已提交
4197
				}
4198
				goto out_restore;
I
Ingo Molnar 已提交
4199 4200 4201
			}
		}
	}
4202 4203
	if (locked)
		graph_unlock();
4204 4205

out_restore:
I
Ingo Molnar 已提交
4206 4207 4208 4209 4210 4211 4212
	raw_local_irq_restore(flags);
}

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

4213
	printk("... MAX_LOCKDEP_SUBCLASSES:  %lu\n", MAX_LOCKDEP_SUBCLASSES);
I
Ingo Molnar 已提交
4214 4215
	printk("... MAX_LOCK_DEPTH:          %lu\n", MAX_LOCK_DEPTH);
	printk("... MAX_LOCKDEP_KEYS:        %lu\n", MAX_LOCKDEP_KEYS);
4216
	printk("... CLASSHASH_SIZE:          %lu\n", CLASSHASH_SIZE);
I
Ingo Molnar 已提交
4217 4218 4219 4220 4221 4222 4223 4224 4225
	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
4226
		sizeof(struct list_head) * CHAINHASH_SIZE
4227
#ifdef CONFIG_PROVE_LOCKING
4228
		+ sizeof(struct circular_queue)
4229
#endif
M
Ming Lei 已提交
4230
		) / 1024
4231
		);
I
Ingo Molnar 已提交
4232 4233 4234 4235 4236 4237 4238

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

static void
print_freed_lock_bug(struct task_struct *curr, const void *mem_from,
4239
		     const void *mem_to, struct held_lock *hlock)
I
Ingo Molnar 已提交
4240 4241 4242 4243 4244 4245
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

4246 4247 4248
	printk("\n");
	printk("=========================\n");
	printk("[ BUG: held lock freed! ]\n");
4249
	print_kernel_ident();
4250
	printk("-------------------------\n");
I
Ingo Molnar 已提交
4251
	printk("%s/%d is freeing memory %p-%p, with a lock still held there!\n",
4252
		curr->comm, task_pid_nr(curr), mem_from, mem_to-1);
4253
	print_lock(hlock);
I
Ingo Molnar 已提交
4254 4255 4256 4257 4258 4259
	lockdep_print_held_locks(curr);

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

O
Oleg Nesterov 已提交
4260 4261 4262 4263 4264 4265 4266
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
4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285
/*
 * 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
4286 4287
		if (not_in_range(mem_from, mem_len, hlock->instance,
					sizeof(*hlock->instance)))
I
Ingo Molnar 已提交
4288 4289
			continue;

O
Oleg Nesterov 已提交
4290
		print_freed_lock_bug(curr, mem_from, mem_from + mem_len, hlock);
I
Ingo Molnar 已提交
4291 4292 4293 4294
		break;
	}
	local_irq_restore(flags);
}
4295
EXPORT_SYMBOL_GPL(debug_check_no_locks_freed);
I
Ingo Molnar 已提交
4296

4297
static void print_held_locks_bug(void)
I
Ingo Molnar 已提交
4298 4299 4300 4301 4302 4303
{
	if (!debug_locks_off())
		return;
	if (debug_locks_silent)
		return;

4304 4305
	printk("\n");
	printk("=====================================\n");
4306 4307
	printk("[ BUG: %s/%d still has locks held! ]\n",
	       current->comm, task_pid_nr(current));
4308
	print_kernel_ident();
4309
	printk("-------------------------------------\n");
4310
	lockdep_print_held_locks(current);
I
Ingo Molnar 已提交
4311 4312 4313 4314
	printk("\nstack backtrace:\n");
	dump_stack();
}

4315
void debug_check_no_locks_held(void)
I
Ingo Molnar 已提交
4316
{
4317 4318
	if (unlikely(current->lockdep_depth > 0))
		print_held_locks_bug();
I
Ingo Molnar 已提交
4319
}
4320
EXPORT_SYMBOL_GPL(debug_check_no_locks_held);
I
Ingo Molnar 已提交
4321

4322
#ifdef __KERNEL__
I
Ingo Molnar 已提交
4323 4324 4325 4326 4327 4328
void debug_show_all_locks(void)
{
	struct task_struct *g, *p;
	int count = 10;
	int unlock = 1;

4329 4330 4331 4332
	if (unlikely(!debug_locks)) {
		printk("INFO: lockdep is turned off.\n");
		return;
	}
I
Ingo Molnar 已提交
4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352
	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;
4353 4354 4355
	} else {
		if (count != 10)
			printk(KERN_CONT " locked it.\n");
I
Ingo Molnar 已提交
4356 4357 4358
	}

	do_each_thread(g, p) {
I
Ingo Molnar 已提交
4359 4360 4361 4362 4363 4364 4365
		/*
		 * 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
4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379
		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);
4380
#endif
I
Ingo Molnar 已提交
4381

4382 4383 4384 4385
/*
 * Careful: only use this function if you are sure that
 * the task cannot run in parallel!
 */
4386
void debug_show_held_locks(struct task_struct *task)
I
Ingo Molnar 已提交
4387
{
4388 4389 4390 4391
	if (unlikely(!debug_locks)) {
		printk("INFO: lockdep is turned off.\n");
		return;
	}
I
Ingo Molnar 已提交
4392 4393 4394
	lockdep_print_held_locks(task);
}
EXPORT_SYMBOL_GPL(debug_show_held_locks);
P
Peter Zijlstra 已提交
4395

4396
asmlinkage __visible void lockdep_sys_exit(void)
P
Peter Zijlstra 已提交
4397 4398 4399 4400 4401 4402
{
	struct task_struct *curr = current;

	if (unlikely(curr->lockdep_depth)) {
		if (!debug_locks_off())
			return;
4403 4404 4405
		printk("\n");
		printk("================================================\n");
		printk("[ BUG: lock held when returning to user space! ]\n");
4406
		print_kernel_ident();
4407
		printk("------------------------------------------------\n");
P
Peter Zijlstra 已提交
4408 4409 4410 4411 4412
		printk("%s/%d is leaving the kernel with locks still held!\n",
				curr->comm, curr->pid);
		lockdep_print_held_locks(curr);
	}
}
4413

4414
void lockdep_rcu_suspicious(const char *file, const int line, const char *s)
4415 4416 4417
{
	struct task_struct *curr = current;

4418
#ifndef CONFIG_PROVE_RCU_REPEATEDLY
4419 4420
	if (!debug_locks_off())
		return;
4421 4422
#endif /* #ifdef CONFIG_PROVE_RCU_REPEATEDLY */
	/* Note: the following can be executed concurrently, so be careful. */
4423
	printk("\n");
4424 4425
	pr_err("===============================\n");
	pr_err("[ ERR: suspicious RCU usage.  ]\n");
4426
	print_kernel_ident();
4427 4428 4429 4430
	pr_err("-------------------------------\n");
	pr_err("%s:%d %s!\n", file, line, s);
	pr_err("\nother info that might help us debug this:\n\n");
	pr_err("\n%srcu_scheduler_active = %d, debug_locks = %d\n",
4431 4432
	       !rcu_lockdep_current_cpu_online()
			? "RCU used illegally from offline CPU!\n"
4433
			: !rcu_is_watching()
4434 4435 4436
				? "RCU used illegally from idle CPU!\n"
				: "",
	       rcu_scheduler_active, debug_locks);
4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455

	/*
	 * If a CPU is in the RCU-free window in idle (ie: in the section
	 * between rcu_idle_enter() and rcu_idle_exit(), then RCU
	 * considers that CPU to be in an "extended quiescent state",
	 * which means that RCU will be completely ignoring that CPU.
	 * Therefore, rcu_read_lock() and friends have absolutely no
	 * effect on a CPU running in that state. In other words, even if
	 * such an RCU-idle CPU has called rcu_read_lock(), RCU might well
	 * delete data structures out from under it.  RCU really has no
	 * choice here: we need to keep an RCU-free window in idle where
	 * the CPU may possibly enter into low power mode. This way we can
	 * notice an extended quiescent state to other CPUs that started a grace
	 * period. Otherwise we would delay any grace period as long as we run
	 * in the idle task.
	 *
	 * So complain bitterly if someone does call rcu_read_lock(),
	 * rcu_read_lock_bh() and so on from extended quiescent states.
	 */
4456
	if (!rcu_is_watching())
4457 4458
		printk("RCU used illegally from extended quiescent state!\n");

4459 4460 4461 4462
	lockdep_print_held_locks(curr);
	printk("\nstack backtrace:\n");
	dump_stack();
}
4463
EXPORT_SYMBOL_GPL(lockdep_rcu_suspicious);