]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - kernel/locking/lockdep.c
mm, vmstat: suppress pcp stats for unpopulated zones in zoneinfo
[mirror_ubuntu-bionic-kernel.git] / kernel / locking / lockdep.c
CommitLineData
fbb9ce95
IM
1/*
2 * kernel/lockdep.c
3 *
4 * Runtime locking correctness validator
5 *
6 * Started by Ingo Molnar:
7 *
4b32d0a4 8 * Copyright (C) 2006,2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
90eec103 9 * Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra
fbb9ce95
IM
10 *
11 * this code maps all the lock dependencies as they occur in a live kernel
12 * and will warn about the following classes of locking bugs:
13 *
14 * - lock inversion scenarios
15 * - circular lock dependencies
16 * - hardirq/softirq safe/unsafe locking bugs
17 *
18 * Bugs are reported even if the current locking scenario does not cause
19 * any deadlock at this point.
20 *
21 * I.e. if anytime in the past two locks were taken in a different order,
22 * even if it happened for another task, even if those were different
23 * locks (but of the same class as this lock), this code will detect it.
24 *
25 * Thanks to Arjan van de Ven for coming up with the initial idea of
26 * mapping lock dependencies runtime.
27 */
a5e25883 28#define DISABLE_BRANCH_PROFILING
fbb9ce95
IM
29#include <linux/mutex.h>
30#include <linux/sched.h>
e6017571 31#include <linux/sched/clock.h>
29930025 32#include <linux/sched/task.h>
fbb9ce95
IM
33#include <linux/delay.h>
34#include <linux/module.h>
35#include <linux/proc_fs.h>
36#include <linux/seq_file.h>
37#include <linux/spinlock.h>
38#include <linux/kallsyms.h>
39#include <linux/interrupt.h>
40#include <linux/stacktrace.h>
41#include <linux/debug_locks.h>
42#include <linux/irqflags.h>
99de055a 43#include <linux/utsname.h>
4b32d0a4 44#include <linux/hash.h>
81d68a96 45#include <linux/ftrace.h>
b4b136f4 46#include <linux/stringify.h>
d588e461 47#include <linux/bitops.h>
5a0e3ad6 48#include <linux/gfp.h>
d3d03d4f 49#include <linux/kmemcheck.h>
e7904a28 50#include <linux/random.h>
dfaaf3fa 51#include <linux/jhash.h>
af012961 52
fbb9ce95
IM
53#include <asm/sections.h>
54
55#include "lockdep_internals.h"
56
a8d154b0 57#define CREATE_TRACE_POINTS
67178767 58#include <trace/events/lock.h>
a8d154b0 59
f20786ff
PZ
60#ifdef CONFIG_PROVE_LOCKING
61int prove_locking = 1;
62module_param(prove_locking, int, 0644);
63#else
64#define prove_locking 0
65#endif
66
67#ifdef CONFIG_LOCK_STAT
68int lock_stat = 1;
69module_param(lock_stat, int, 0644);
70#else
71#define lock_stat 0
72#endif
73
fbb9ce95 74/*
74c383f1
IM
75 * lockdep_lock: protects the lockdep graph, the hashes and the
76 * class/list/hash allocators.
fbb9ce95
IM
77 *
78 * This is one of the rare exceptions where it's justified
79 * to use a raw spinlock - we really dont want the spinlock
74c383f1 80 * code to recurse back into the lockdep code...
fbb9ce95 81 */
edc35bd7 82static arch_spinlock_t lockdep_lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
74c383f1
IM
83
84static int graph_lock(void)
85{
0199c4e6 86 arch_spin_lock(&lockdep_lock);
74c383f1
IM
87 /*
88 * Make sure that if another CPU detected a bug while
89 * walking the graph we dont change it (while the other
90 * CPU is busy printing out stuff with the graph lock
91 * dropped already)
92 */
93 if (!debug_locks) {
0199c4e6 94 arch_spin_unlock(&lockdep_lock);
74c383f1
IM
95 return 0;
96 }
bb065afb
SR
97 /* prevent any recursions within lockdep from causing deadlocks */
98 current->lockdep_recursion++;
74c383f1
IM
99 return 1;
100}
101
102static inline int graph_unlock(void)
103{
0119fee4
PZ
104 if (debug_locks && !arch_spin_is_locked(&lockdep_lock)) {
105 /*
106 * The lockdep graph lock isn't locked while we expect it to
107 * be, we're confused now, bye!
108 */
381a2292 109 return DEBUG_LOCKS_WARN_ON(1);
0119fee4 110 }
381a2292 111
bb065afb 112 current->lockdep_recursion--;
0199c4e6 113 arch_spin_unlock(&lockdep_lock);
74c383f1
IM
114 return 0;
115}
116
117/*
118 * Turn lock debugging off and return with 0 if it was off already,
119 * and also release the graph lock:
120 */
121static inline int debug_locks_off_graph_unlock(void)
122{
123 int ret = debug_locks_off();
124
0199c4e6 125 arch_spin_unlock(&lockdep_lock);
74c383f1
IM
126
127 return ret;
128}
fbb9ce95 129
fbb9ce95 130unsigned long nr_list_entries;
af012961 131static struct lock_list list_entries[MAX_LOCKDEP_ENTRIES];
fbb9ce95 132
fbb9ce95
IM
133/*
134 * All data structures here are protected by the global debug_lock.
135 *
136 * Mutex key structs only get allocated, once during bootup, and never
137 * get freed - this significantly simplifies the debugging code.
138 */
139unsigned long nr_lock_classes;
140static struct lock_class lock_classes[MAX_LOCKDEP_KEYS];
141
f82b217e
DJ
142static inline struct lock_class *hlock_class(struct held_lock *hlock)
143{
144 if (!hlock->class_idx) {
0119fee4
PZ
145 /*
146 * Someone passed in garbage, we give up.
147 */
f82b217e
DJ
148 DEBUG_LOCKS_WARN_ON(1);
149 return NULL;
150 }
151 return lock_classes + hlock->class_idx - 1;
152}
153
f20786ff 154#ifdef CONFIG_LOCK_STAT
25528213 155static DEFINE_PER_CPU(struct lock_class_stats[MAX_LOCKDEP_KEYS], cpu_lock_stats);
f20786ff 156
3365e779
PZ
157static inline u64 lockstat_clock(void)
158{
c676329a 159 return local_clock();
3365e779
PZ
160}
161
c7e78cff 162static int lock_point(unsigned long points[], unsigned long ip)
f20786ff
PZ
163{
164 int i;
165
c7e78cff
PZ
166 for (i = 0; i < LOCKSTAT_POINTS; i++) {
167 if (points[i] == 0) {
168 points[i] = ip;
f20786ff
PZ
169 break;
170 }
c7e78cff 171 if (points[i] == ip)
f20786ff
PZ
172 break;
173 }
174
175 return i;
176}
177
3365e779 178static void lock_time_inc(struct lock_time *lt, u64 time)
f20786ff
PZ
179{
180 if (time > lt->max)
181 lt->max = time;
182
109d71c6 183 if (time < lt->min || !lt->nr)
f20786ff
PZ
184 lt->min = time;
185
186 lt->total += time;
187 lt->nr++;
188}
189
c46261de
PZ
190static inline void lock_time_add(struct lock_time *src, struct lock_time *dst)
191{
109d71c6
FR
192 if (!src->nr)
193 return;
194
195 if (src->max > dst->max)
196 dst->max = src->max;
197
198 if (src->min < dst->min || !dst->nr)
199 dst->min = src->min;
200
c46261de
PZ
201 dst->total += src->total;
202 dst->nr += src->nr;
203}
204
205struct lock_class_stats lock_stats(struct lock_class *class)
206{
207 struct lock_class_stats stats;
208 int cpu, i;
209
210 memset(&stats, 0, sizeof(struct lock_class_stats));
211 for_each_possible_cpu(cpu) {
212 struct lock_class_stats *pcs =
1871e52c 213 &per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
c46261de
PZ
214
215 for (i = 0; i < ARRAY_SIZE(stats.contention_point); i++)
216 stats.contention_point[i] += pcs->contention_point[i];
217
c7e78cff
PZ
218 for (i = 0; i < ARRAY_SIZE(stats.contending_point); i++)
219 stats.contending_point[i] += pcs->contending_point[i];
220
c46261de
PZ
221 lock_time_add(&pcs->read_waittime, &stats.read_waittime);
222 lock_time_add(&pcs->write_waittime, &stats.write_waittime);
223
224 lock_time_add(&pcs->read_holdtime, &stats.read_holdtime);
225 lock_time_add(&pcs->write_holdtime, &stats.write_holdtime);
96645678
PZ
226
227 for (i = 0; i < ARRAY_SIZE(stats.bounces); i++)
228 stats.bounces[i] += pcs->bounces[i];
c46261de
PZ
229 }
230
231 return stats;
232}
233
234void clear_lock_stats(struct lock_class *class)
235{
236 int cpu;
237
238 for_each_possible_cpu(cpu) {
239 struct lock_class_stats *cpu_stats =
1871e52c 240 &per_cpu(cpu_lock_stats, cpu)[class - lock_classes];
c46261de
PZ
241
242 memset(cpu_stats, 0, sizeof(struct lock_class_stats));
243 }
244 memset(class->contention_point, 0, sizeof(class->contention_point));
c7e78cff 245 memset(class->contending_point, 0, sizeof(class->contending_point));
c46261de
PZ
246}
247
f20786ff
PZ
248static struct lock_class_stats *get_lock_stats(struct lock_class *class)
249{
1871e52c 250 return &get_cpu_var(cpu_lock_stats)[class - lock_classes];
f20786ff
PZ
251}
252
253static void put_lock_stats(struct lock_class_stats *stats)
254{
1871e52c 255 put_cpu_var(cpu_lock_stats);
f20786ff
PZ
256}
257
258static void lock_release_holdtime(struct held_lock *hlock)
259{
260 struct lock_class_stats *stats;
3365e779 261 u64 holdtime;
f20786ff
PZ
262
263 if (!lock_stat)
264 return;
265
3365e779 266 holdtime = lockstat_clock() - hlock->holdtime_stamp;
f20786ff 267
f82b217e 268 stats = get_lock_stats(hlock_class(hlock));
f20786ff
PZ
269 if (hlock->read)
270 lock_time_inc(&stats->read_holdtime, holdtime);
271 else
272 lock_time_inc(&stats->write_holdtime, holdtime);
273 put_lock_stats(stats);
274}
275#else
276static inline void lock_release_holdtime(struct held_lock *hlock)
277{
278}
279#endif
280
fbb9ce95
IM
281/*
282 * We keep a global list of all lock classes. The list only grows,
283 * never shrinks. The list is only accessed with the lockdep
284 * spinlock lock held.
285 */
286LIST_HEAD(all_lock_classes);
287
288/*
289 * The lockdep classes are in a hash-table as well, for fast lookup:
290 */
291#define CLASSHASH_BITS (MAX_LOCKDEP_KEYS_BITS - 1)
292#define CLASSHASH_SIZE (1UL << CLASSHASH_BITS)
4b32d0a4 293#define __classhashfn(key) hash_long((unsigned long)key, CLASSHASH_BITS)
fbb9ce95
IM
294#define classhashentry(key) (classhash_table + __classhashfn((key)))
295
a63f38cc 296static struct hlist_head classhash_table[CLASSHASH_SIZE];
fbb9ce95 297
fbb9ce95
IM
298/*
299 * We put the lock dependency chains into a hash-table as well, to cache
300 * their existence:
301 */
302#define CHAINHASH_BITS (MAX_LOCKDEP_CHAINS_BITS-1)
303#define CHAINHASH_SIZE (1UL << CHAINHASH_BITS)
4b32d0a4 304#define __chainhashfn(chain) hash_long(chain, CHAINHASH_BITS)
fbb9ce95
IM
305#define chainhashentry(chain) (chainhash_table + __chainhashfn((chain)))
306
a63f38cc 307static struct hlist_head chainhash_table[CHAINHASH_SIZE];
fbb9ce95
IM
308
309/*
310 * The hash key of the lock dependency chains is a hash itself too:
311 * it's a hash of all locks taken up to that lock, including that lock.
312 * It's a 64-bit hash, because it's important for the keys to be
313 * unique.
314 */
dfaaf3fa
PZ
315static inline u64 iterate_chain_key(u64 key, u32 idx)
316{
317 u32 k0 = key, k1 = key >> 32;
318
319 __jhash_mix(idx, k0, k1); /* Macro that modifies arguments! */
320
321 return k0 | (u64)k1 << 32;
322}
fbb9ce95 323
1d09daa5 324void lockdep_off(void)
fbb9ce95
IM
325{
326 current->lockdep_recursion++;
327}
fbb9ce95
IM
328EXPORT_SYMBOL(lockdep_off);
329
1d09daa5 330void lockdep_on(void)
fbb9ce95
IM
331{
332 current->lockdep_recursion--;
333}
fbb9ce95
IM
334EXPORT_SYMBOL(lockdep_on);
335
fbb9ce95
IM
336/*
337 * Debugging switches:
338 */
339
340#define VERBOSE 0
33e94e96 341#define VERY_VERBOSE 0
fbb9ce95
IM
342
343#if VERBOSE
344# define HARDIRQ_VERBOSE 1
345# define SOFTIRQ_VERBOSE 1
cf40bd16 346# define RECLAIM_VERBOSE 1
fbb9ce95
IM
347#else
348# define HARDIRQ_VERBOSE 0
349# define SOFTIRQ_VERBOSE 0
cf40bd16 350# define RECLAIM_VERBOSE 0
fbb9ce95
IM
351#endif
352
cf40bd16 353#if VERBOSE || HARDIRQ_VERBOSE || SOFTIRQ_VERBOSE || RECLAIM_VERBOSE
fbb9ce95
IM
354/*
355 * Quick filtering for interesting events:
356 */
357static int class_filter(struct lock_class *class)
358{
f9829cce
AK
359#if 0
360 /* Example */
fbb9ce95 361 if (class->name_version == 1 &&
f9829cce 362 !strcmp(class->name, "lockname"))
fbb9ce95
IM
363 return 1;
364 if (class->name_version == 1 &&
f9829cce 365 !strcmp(class->name, "&struct->lockfield"))
fbb9ce95 366 return 1;
f9829cce 367#endif
a6640897
IM
368 /* Filter everything else. 1 would be to allow everything else */
369 return 0;
fbb9ce95
IM
370}
371#endif
372
373static int verbose(struct lock_class *class)
374{
375#if VERBOSE
376 return class_filter(class);
377#endif
378 return 0;
379}
380
fbb9ce95
IM
381/*
382 * Stack-trace: tightly packed array of stack backtrace
74c383f1 383 * addresses. Protected by the graph_lock.
fbb9ce95
IM
384 */
385unsigned long nr_stack_trace_entries;
386static unsigned long stack_trace[MAX_STACK_TRACE_ENTRIES];
387
2c522836
DJ
388static void print_lockdep_off(const char *bug_msg)
389{
390 printk(KERN_DEBUG "%s\n", bug_msg);
391 printk(KERN_DEBUG "turning off the locking correctness validator.\n");
acf59377 392#ifdef CONFIG_LOCK_STAT
2c522836 393 printk(KERN_DEBUG "Please attach the output of /proc/lock_stat to the bug report\n");
acf59377 394#endif
2c522836
DJ
395}
396
fbb9ce95
IM
397static int save_trace(struct stack_trace *trace)
398{
399 trace->nr_entries = 0;
400 trace->max_entries = MAX_STACK_TRACE_ENTRIES - nr_stack_trace_entries;
401 trace->entries = stack_trace + nr_stack_trace_entries;
402
5a1b3999 403 trace->skip = 3;
5a1b3999 404
ab1b6f03 405 save_stack_trace(trace);
fbb9ce95 406
4f84f433
PZ
407 /*
408 * Some daft arches put -1 at the end to indicate its a full trace.
409 *
410 * <rant> this is buggy anyway, since it takes a whole extra entry so a
411 * complete trace that maxes out the entries provided will be reported
412 * as incomplete, friggin useless </rant>
413 */
ea5b41f9
TL
414 if (trace->nr_entries != 0 &&
415 trace->entries[trace->nr_entries-1] == ULONG_MAX)
4f84f433
PZ
416 trace->nr_entries--;
417
fbb9ce95
IM
418 trace->max_entries = trace->nr_entries;
419
420 nr_stack_trace_entries += trace->nr_entries;
fbb9ce95 421
4f84f433 422 if (nr_stack_trace_entries >= MAX_STACK_TRACE_ENTRIES-1) {
74c383f1
IM
423 if (!debug_locks_off_graph_unlock())
424 return 0;
425
2c522836 426 print_lockdep_off("BUG: MAX_STACK_TRACE_ENTRIES too low!");
74c383f1
IM
427 dump_stack();
428
fbb9ce95
IM
429 return 0;
430 }
431
432 return 1;
433}
434
435unsigned int nr_hardirq_chains;
436unsigned int nr_softirq_chains;
437unsigned int nr_process_chains;
438unsigned int max_lockdep_depth;
fbb9ce95
IM
439
440#ifdef CONFIG_DEBUG_LOCKDEP
fbb9ce95
IM
441/*
442 * Various lockdep statistics:
443 */
bd6d29c2 444DEFINE_PER_CPU(struct lockdep_stats, lockdep_stats);
fbb9ce95
IM
445#endif
446
447/*
448 * Locking printouts:
449 */
450
fabe9c42 451#define __USAGE(__STATE) \
b4b136f4
PZ
452 [LOCK_USED_IN_##__STATE] = "IN-"__stringify(__STATE)"-W", \
453 [LOCK_ENABLED_##__STATE] = __stringify(__STATE)"-ON-W", \
454 [LOCK_USED_IN_##__STATE##_READ] = "IN-"__stringify(__STATE)"-R",\
455 [LOCK_ENABLED_##__STATE##_READ] = __stringify(__STATE)"-ON-R",
fabe9c42 456
fbb9ce95
IM
457static const char *usage_str[] =
458{
fabe9c42
PZ
459#define LOCKDEP_STATE(__STATE) __USAGE(__STATE)
460#include "lockdep_states.h"
461#undef LOCKDEP_STATE
462 [LOCK_USED] = "INITIAL USE",
fbb9ce95
IM
463};
464
465const char * __get_key_name(struct lockdep_subclass_key *key, char *str)
466{
ffb45122 467 return kallsyms_lookup((unsigned long)key, NULL, NULL, NULL, str);
fbb9ce95
IM
468}
469
3ff176ca 470static inline unsigned long lock_flag(enum lock_usage_bit bit)
fbb9ce95 471{
3ff176ca
PZ
472 return 1UL << bit;
473}
fbb9ce95 474
3ff176ca
PZ
475static char get_usage_char(struct lock_class *class, enum lock_usage_bit bit)
476{
477 char c = '.';
478
479 if (class->usage_mask & lock_flag(bit + 2))
480 c = '+';
481 if (class->usage_mask & lock_flag(bit)) {
482 c = '-';
483 if (class->usage_mask & lock_flag(bit + 2))
484 c = '?';
fbb9ce95
IM
485 }
486
3ff176ca
PZ
487 return c;
488}
cf40bd16 489
f510b233 490void get_usage_chars(struct lock_class *class, char usage[LOCK_USAGE_CHARS])
3ff176ca 491{
f510b233 492 int i = 0;
cf40bd16 493
f510b233
PZ
494#define LOCKDEP_STATE(__STATE) \
495 usage[i++] = get_usage_char(class, LOCK_USED_IN_##__STATE); \
496 usage[i++] = get_usage_char(class, LOCK_USED_IN_##__STATE##_READ);
497#include "lockdep_states.h"
498#undef LOCKDEP_STATE
499
500 usage[i] = '\0';
fbb9ce95
IM
501}
502
e5e78d08 503static void __print_lock_name(struct lock_class *class)
3003eba3
SR
504{
505 char str[KSYM_NAME_LEN];
506 const char *name;
507
fbb9ce95
IM
508 name = class->name;
509 if (!name) {
510 name = __get_key_name(class->key, str);
f943fe0f 511 printk(KERN_CONT "%s", name);
fbb9ce95 512 } else {
f943fe0f 513 printk(KERN_CONT "%s", name);
fbb9ce95 514 if (class->name_version > 1)
f943fe0f 515 printk(KERN_CONT "#%d", class->name_version);
fbb9ce95 516 if (class->subclass)
f943fe0f 517 printk(KERN_CONT "/%d", class->subclass);
fbb9ce95 518 }
e5e78d08
SR
519}
520
521static void print_lock_name(struct lock_class *class)
522{
523 char usage[LOCK_USAGE_CHARS];
524
525 get_usage_chars(class, usage);
526
f943fe0f 527 printk(KERN_CONT " (");
e5e78d08 528 __print_lock_name(class);
f943fe0f 529 printk(KERN_CONT "){%s}", usage);
fbb9ce95
IM
530}
531
532static void print_lockdep_cache(struct lockdep_map *lock)
533{
534 const char *name;
9281acea 535 char str[KSYM_NAME_LEN];
fbb9ce95
IM
536
537 name = lock->name;
538 if (!name)
539 name = __get_key_name(lock->key->subkeys, str);
540
f943fe0f 541 printk(KERN_CONT "%s", name);
fbb9ce95
IM
542}
543
544static void print_lock(struct held_lock *hlock)
545{
d7bc3197
PZ
546 /*
547 * We can be called locklessly through debug_show_all_locks() so be
548 * extra careful, the hlock might have been released and cleared.
549 */
550 unsigned int class_idx = hlock->class_idx;
551
552 /* Don't re-read hlock->class_idx, can't use READ_ONCE() on bitfields: */
553 barrier();
554
555 if (!class_idx || (class_idx - 1) >= MAX_LOCKDEP_KEYS) {
f943fe0f 556 printk(KERN_CONT "<RELEASED>\n");
d7bc3197
PZ
557 return;
558 }
559
560 print_lock_name(lock_classes + class_idx - 1);
f943fe0f
DV
561 printk(KERN_CONT ", at: [<%p>] %pS\n",
562 (void *)hlock->acquire_ip, (void *)hlock->acquire_ip);
fbb9ce95
IM
563}
564
565static void lockdep_print_held_locks(struct task_struct *curr)
566{
567 int i, depth = curr->lockdep_depth;
568
569 if (!depth) {
ba25f9dc 570 printk("no locks held by %s/%d.\n", curr->comm, task_pid_nr(curr));
fbb9ce95
IM
571 return;
572 }
573 printk("%d lock%s held by %s/%d:\n",
ba25f9dc 574 depth, depth > 1 ? "s" : "", curr->comm, task_pid_nr(curr));
fbb9ce95
IM
575
576 for (i = 0; i < depth; i++) {
577 printk(" #%d: ", i);
578 print_lock(curr->held_locks + i);
579 }
580}
fbb9ce95 581
fbdc4b9a 582static void print_kernel_ident(void)
8e18257d 583{
fbdc4b9a 584 printk("%s %.*s %s\n", init_utsname()->release,
8e18257d 585 (int)strcspn(init_utsname()->version, " "),
fbdc4b9a
BH
586 init_utsname()->version,
587 print_tainted());
8e18257d
PZ
588}
589
590static int very_verbose(struct lock_class *class)
591{
592#if VERY_VERBOSE
593 return class_filter(class);
594#endif
595 return 0;
596}
597
fbb9ce95 598/*
8e18257d 599 * Is this the address of a static object:
fbb9ce95 600 */
8dce7a9a 601#ifdef __KERNEL__
8e18257d 602static int static_obj(void *obj)
fbb9ce95 603{
8e18257d
PZ
604 unsigned long start = (unsigned long) &_stext,
605 end = (unsigned long) &_end,
606 addr = (unsigned long) obj;
8e18257d 607
fbb9ce95 608 /*
8e18257d 609 * static variable?
fbb9ce95 610 */
8e18257d
PZ
611 if ((addr >= start) && (addr < end))
612 return 1;
fbb9ce95 613
2a9ad18d
MF
614 if (arch_is_kernel_data(addr))
615 return 1;
616
fbb9ce95 617 /*
10fad5e4 618 * in-kernel percpu var?
fbb9ce95 619 */
10fad5e4
TH
620 if (is_kernel_percpu_address(addr))
621 return 1;
fbb9ce95 622
8e18257d 623 /*
10fad5e4 624 * module static or percpu var?
8e18257d 625 */
10fad5e4 626 return is_module_address(addr) || is_module_percpu_address(addr);
99de055a 627}
8dce7a9a 628#endif
99de055a 629
fbb9ce95 630/*
8e18257d
PZ
631 * To make lock name printouts unique, we calculate a unique
632 * class->name_version generation counter:
fbb9ce95 633 */
8e18257d 634static int count_matching_names(struct lock_class *new_class)
fbb9ce95 635{
8e18257d
PZ
636 struct lock_class *class;
637 int count = 0;
fbb9ce95 638
8e18257d 639 if (!new_class->name)
fbb9ce95
IM
640 return 0;
641
35a9393c 642 list_for_each_entry_rcu(class, &all_lock_classes, lock_entry) {
8e18257d
PZ
643 if (new_class->key - new_class->subclass == class->key)
644 return class->name_version;
645 if (class->name && !strcmp(class->name, new_class->name))
646 count = max(count, class->name_version);
647 }
fbb9ce95 648
8e18257d 649 return count + 1;
fbb9ce95
IM
650}
651
8e18257d
PZ
652/*
653 * Register a lock's class in the hash-table, if the class is not present
654 * yet. Otherwise we look it up. We cache the result in the lock object
655 * itself, so actual lookup of the hash should be once per lock object.
656 */
657static inline struct lock_class *
658look_up_lock_class(struct lockdep_map *lock, unsigned int subclass)
fbb9ce95 659{
8e18257d 660 struct lockdep_subclass_key *key;
a63f38cc 661 struct hlist_head *hash_head;
8e18257d 662 struct lock_class *class;
383776fa 663 bool is_static = false;
fbb9ce95 664
4ba053c0
HM
665 if (unlikely(subclass >= MAX_LOCKDEP_SUBCLASSES)) {
666 debug_locks_off();
667 printk(KERN_ERR
668 "BUG: looking up invalid subclass: %u\n", subclass);
669 printk(KERN_ERR
670 "turning off the locking correctness validator.\n");
671 dump_stack();
672 return NULL;
673 }
674
8e18257d
PZ
675 /*
676 * Static locks do not have their class-keys yet - for them the key
383776fa
TG
677 * is the lock object itself. If the lock is in the per cpu area,
678 * the canonical address of the lock (per cpu offset removed) is
679 * used.
8e18257d 680 */
383776fa
TG
681 if (unlikely(!lock->key)) {
682 unsigned long can_addr, addr = (unsigned long)lock;
683
684 if (__is_kernel_percpu_address(addr, &can_addr))
685 lock->key = (void *)can_addr;
686 else if (__is_module_percpu_address(addr, &can_addr))
687 lock->key = (void *)can_addr;
688 else if (static_obj(lock))
689 lock->key = (void *)lock;
690 else
691 return ERR_PTR(-EINVAL);
692 is_static = true;
693 }
fbb9ce95 694
8e18257d
PZ
695 /*
696 * NOTE: the class-key must be unique. For dynamic locks, a static
697 * lock_class_key variable is passed in through the mutex_init()
698 * (or spin_lock_init()) call - which acts as the key. For static
699 * locks we use the lock object itself as the key.
700 */
4b32d0a4
PZ
701 BUILD_BUG_ON(sizeof(struct lock_class_key) >
702 sizeof(struct lockdep_map));
fbb9ce95 703
8e18257d 704 key = lock->key->subkeys + subclass;
ca268c69 705
8e18257d 706 hash_head = classhashentry(key);
74c383f1 707
8e18257d 708 /*
35a9393c 709 * We do an RCU walk of the hash, see lockdep_free_key_range().
8e18257d 710 */
35a9393c
PZ
711 if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
712 return NULL;
713
a63f38cc 714 hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
4b32d0a4 715 if (class->key == key) {
0119fee4
PZ
716 /*
717 * Huh! same key, different name? Did someone trample
718 * on some memory? We're most confused.
719 */
4b32d0a4 720 WARN_ON_ONCE(class->name != lock->name);
8e18257d 721 return class;
4b32d0a4
PZ
722 }
723 }
fbb9ce95 724
383776fa 725 return is_static || static_obj(lock->key) ? NULL : ERR_PTR(-EINVAL);
fbb9ce95
IM
726}
727
728/*
8e18257d
PZ
729 * Register a lock's class in the hash-table, if the class is not present
730 * yet. Otherwise we look it up. We cache the result in the lock object
731 * itself, so actual lookup of the hash should be once per lock object.
fbb9ce95 732 */
c003ed92 733static struct lock_class *
8e18257d 734register_lock_class(struct lockdep_map *lock, unsigned int subclass, int force)
fbb9ce95 735{
8e18257d 736 struct lockdep_subclass_key *key;
a63f38cc 737 struct hlist_head *hash_head;
8e18257d 738 struct lock_class *class;
35a9393c
PZ
739
740 DEBUG_LOCKS_WARN_ON(!irqs_disabled());
8e18257d
PZ
741
742 class = look_up_lock_class(lock, subclass);
383776fa 743 if (likely(!IS_ERR_OR_NULL(class)))
87cdee71 744 goto out_set_class_cache;
8e18257d
PZ
745
746 /*
747 * Debug-check: all keys must be persistent!
383776fa
TG
748 */
749 if (IS_ERR(class)) {
8e18257d
PZ
750 debug_locks_off();
751 printk("INFO: trying to register non-static key.\n");
752 printk("the code is fine but needs lockdep annotation.\n");
753 printk("turning off the locking correctness validator.\n");
754 dump_stack();
8e18257d
PZ
755 return NULL;
756 }
757
758 key = lock->key->subkeys + subclass;
759 hash_head = classhashentry(key);
760
8e18257d 761 if (!graph_lock()) {
8e18257d
PZ
762 return NULL;
763 }
764 /*
765 * We have to do the hash-walk again, to avoid races
766 * with another CPU:
767 */
a63f38cc 768 hlist_for_each_entry_rcu(class, hash_head, hash_entry) {
8e18257d
PZ
769 if (class->key == key)
770 goto out_unlock_set;
35a9393c
PZ
771 }
772
8e18257d
PZ
773 /*
774 * Allocate a new key from the static array, and add it to
775 * the hash:
776 */
777 if (nr_lock_classes >= MAX_LOCKDEP_KEYS) {
778 if (!debug_locks_off_graph_unlock()) {
8e18257d
PZ
779 return NULL;
780 }
8e18257d 781
2c522836 782 print_lockdep_off("BUG: MAX_LOCKDEP_KEYS too low!");
eedeeabd 783 dump_stack();
8e18257d
PZ
784 return NULL;
785 }
786 class = lock_classes + nr_lock_classes++;
bd6d29c2 787 debug_atomic_inc(nr_unused_locks);
8e18257d
PZ
788 class->key = key;
789 class->name = lock->name;
790 class->subclass = subclass;
791 INIT_LIST_HEAD(&class->lock_entry);
792 INIT_LIST_HEAD(&class->locks_before);
793 INIT_LIST_HEAD(&class->locks_after);
794 class->name_version = count_matching_names(class);
795 /*
796 * We use RCU's safe list-add method to make
797 * parallel walking of the hash-list safe:
798 */
a63f38cc 799 hlist_add_head_rcu(&class->hash_entry, hash_head);
1481197b
DF
800 /*
801 * Add it to the global list of classes:
802 */
803 list_add_tail_rcu(&class->lock_entry, &all_lock_classes);
8e18257d
PZ
804
805 if (verbose(class)) {
806 graph_unlock();
8e18257d
PZ
807
808 printk("\nnew class %p: %s", class->key, class->name);
809 if (class->name_version > 1)
f943fe0f
DV
810 printk(KERN_CONT "#%d", class->name_version);
811 printk(KERN_CONT "\n");
8e18257d
PZ
812 dump_stack();
813
8e18257d 814 if (!graph_lock()) {
8e18257d
PZ
815 return NULL;
816 }
817 }
818out_unlock_set:
819 graph_unlock();
8e18257d 820
87cdee71 821out_set_class_cache:
8e18257d 822 if (!subclass || force)
62016250
HM
823 lock->class_cache[0] = class;
824 else if (subclass < NR_LOCKDEP_CACHING_CLASSES)
825 lock->class_cache[subclass] = class;
8e18257d 826
0119fee4
PZ
827 /*
828 * Hash collision, did we smoke some? We found a class with a matching
829 * hash but the subclass -- which is hashed in -- didn't match.
830 */
8e18257d
PZ
831 if (DEBUG_LOCKS_WARN_ON(class->subclass != subclass))
832 return NULL;
833
834 return class;
835}
836
837#ifdef CONFIG_PROVE_LOCKING
838/*
839 * Allocate a lockdep entry. (assumes the graph_lock held, returns
840 * with NULL on failure)
841 */
842static struct lock_list *alloc_list_entry(void)
843{
844 if (nr_list_entries >= MAX_LOCKDEP_ENTRIES) {
845 if (!debug_locks_off_graph_unlock())
846 return NULL;
847
2c522836 848 print_lockdep_off("BUG: MAX_LOCKDEP_ENTRIES too low!");
eedeeabd 849 dump_stack();
8e18257d
PZ
850 return NULL;
851 }
852 return list_entries + nr_list_entries++;
853}
854
855/*
856 * Add a new dependency to the head of the list:
857 */
83f06168
TE
858static int add_lock_to_list(struct lock_class *this, struct list_head *head,
859 unsigned long ip, int distance,
860 struct stack_trace *trace)
8e18257d
PZ
861{
862 struct lock_list *entry;
863 /*
864 * Lock not present yet - get a new dependency struct and
865 * add it to the list:
866 */
867 entry = alloc_list_entry();
868 if (!entry)
869 return 0;
870
74870172
ZY
871 entry->class = this;
872 entry->distance = distance;
4726f2a6 873 entry->trace = *trace;
8e18257d 874 /*
35a9393c
PZ
875 * Both allocation and removal are done under the graph lock; but
876 * iteration is under RCU-sched; see look_up_lock_class() and
877 * lockdep_free_key_range().
8e18257d
PZ
878 */
879 list_add_tail_rcu(&entry->entry, head);
880
881 return 1;
882}
883
98c33edd
PZ
884/*
885 * For good efficiency of modular, we use power of 2
886 */
af012961
PZ
887#define MAX_CIRCULAR_QUEUE_SIZE 4096UL
888#define CQ_MASK (MAX_CIRCULAR_QUEUE_SIZE-1)
889
98c33edd
PZ
890/*
891 * The circular_queue and helpers is used to implement the
af012961
PZ
892 * breadth-first search(BFS)algorithem, by which we can build
893 * the shortest path from the next lock to be acquired to the
894 * previous held lock if there is a circular between them.
98c33edd 895 */
af012961
PZ
896struct circular_queue {
897 unsigned long element[MAX_CIRCULAR_QUEUE_SIZE];
898 unsigned int front, rear;
899};
900
901static struct circular_queue lock_cq;
af012961 902
12f3dfd0 903unsigned int max_bfs_queue_depth;
af012961 904
e351b660
ML
905static unsigned int lockdep_dependency_gen_id;
906
af012961
PZ
907static inline void __cq_init(struct circular_queue *cq)
908{
909 cq->front = cq->rear = 0;
e351b660 910 lockdep_dependency_gen_id++;
af012961
PZ
911}
912
913static inline int __cq_empty(struct circular_queue *cq)
914{
915 return (cq->front == cq->rear);
916}
917
918static inline int __cq_full(struct circular_queue *cq)
919{
920 return ((cq->rear + 1) & CQ_MASK) == cq->front;
921}
922
923static inline int __cq_enqueue(struct circular_queue *cq, unsigned long elem)
924{
925 if (__cq_full(cq))
926 return -1;
927
928 cq->element[cq->rear] = elem;
929 cq->rear = (cq->rear + 1) & CQ_MASK;
930 return 0;
931}
932
933static inline int __cq_dequeue(struct circular_queue *cq, unsigned long *elem)
934{
935 if (__cq_empty(cq))
936 return -1;
937
938 *elem = cq->element[cq->front];
939 cq->front = (cq->front + 1) & CQ_MASK;
940 return 0;
941}
942
943static inline unsigned int __cq_get_elem_count(struct circular_queue *cq)
944{
945 return (cq->rear - cq->front) & CQ_MASK;
946}
947
948static inline void mark_lock_accessed(struct lock_list *lock,
949 struct lock_list *parent)
950{
951 unsigned long nr;
98c33edd 952
af012961 953 nr = lock - list_entries;
0119fee4 954 WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
af012961 955 lock->parent = parent;
e351b660 956 lock->class->dep_gen_id = lockdep_dependency_gen_id;
af012961
PZ
957}
958
959static inline unsigned long lock_accessed(struct lock_list *lock)
960{
961 unsigned long nr;
98c33edd 962
af012961 963 nr = lock - list_entries;
0119fee4 964 WARN_ON(nr >= nr_list_entries); /* Out-of-bounds, input fail */
e351b660 965 return lock->class->dep_gen_id == lockdep_dependency_gen_id;
af012961
PZ
966}
967
968static inline struct lock_list *get_lock_parent(struct lock_list *child)
969{
970 return child->parent;
971}
972
973static inline int get_lock_depth(struct lock_list *child)
974{
975 int depth = 0;
976 struct lock_list *parent;
977
978 while ((parent = get_lock_parent(child))) {
979 child = parent;
980 depth++;
981 }
982 return depth;
983}
984
9e2d551e 985static int __bfs(struct lock_list *source_entry,
af012961
PZ
986 void *data,
987 int (*match)(struct lock_list *entry, void *data),
988 struct lock_list **target_entry,
989 int forward)
c94aa5ca
ML
990{
991 struct lock_list *entry;
d588e461 992 struct list_head *head;
c94aa5ca
ML
993 struct circular_queue *cq = &lock_cq;
994 int ret = 1;
995
9e2d551e 996 if (match(source_entry, data)) {
c94aa5ca
ML
997 *target_entry = source_entry;
998 ret = 0;
999 goto exit;
1000 }
1001
d588e461
ML
1002 if (forward)
1003 head = &source_entry->class->locks_after;
1004 else
1005 head = &source_entry->class->locks_before;
1006
1007 if (list_empty(head))
1008 goto exit;
1009
1010 __cq_init(cq);
c94aa5ca
ML
1011 __cq_enqueue(cq, (unsigned long)source_entry);
1012
1013 while (!__cq_empty(cq)) {
1014 struct lock_list *lock;
c94aa5ca
ML
1015
1016 __cq_dequeue(cq, (unsigned long *)&lock);
1017
1018 if (!lock->class) {
1019 ret = -2;
1020 goto exit;
1021 }
1022
1023 if (forward)
1024 head = &lock->class->locks_after;
1025 else
1026 head = &lock->class->locks_before;
1027
35a9393c
PZ
1028 DEBUG_LOCKS_WARN_ON(!irqs_disabled());
1029
1030 list_for_each_entry_rcu(entry, head, entry) {
c94aa5ca 1031 if (!lock_accessed(entry)) {
12f3dfd0 1032 unsigned int cq_depth;
c94aa5ca 1033 mark_lock_accessed(entry, lock);
9e2d551e 1034 if (match(entry, data)) {
c94aa5ca
ML
1035 *target_entry = entry;
1036 ret = 0;
1037 goto exit;
1038 }
1039
1040 if (__cq_enqueue(cq, (unsigned long)entry)) {
1041 ret = -1;
1042 goto exit;
1043 }
12f3dfd0
ML
1044 cq_depth = __cq_get_elem_count(cq);
1045 if (max_bfs_queue_depth < cq_depth)
1046 max_bfs_queue_depth = cq_depth;
c94aa5ca
ML
1047 }
1048 }
1049 }
1050exit:
1051 return ret;
1052}
1053
d7aaba14 1054static inline int __bfs_forwards(struct lock_list *src_entry,
9e2d551e
ML
1055 void *data,
1056 int (*match)(struct lock_list *entry, void *data),
1057 struct lock_list **target_entry)
c94aa5ca 1058{
9e2d551e 1059 return __bfs(src_entry, data, match, target_entry, 1);
c94aa5ca
ML
1060
1061}
1062
d7aaba14 1063static inline int __bfs_backwards(struct lock_list *src_entry,
9e2d551e
ML
1064 void *data,
1065 int (*match)(struct lock_list *entry, void *data),
1066 struct lock_list **target_entry)
c94aa5ca 1067{
9e2d551e 1068 return __bfs(src_entry, data, match, target_entry, 0);
c94aa5ca
ML
1069
1070}
1071
8e18257d
PZ
1072/*
1073 * Recursive, forwards-direction lock-dependency checking, used for
1074 * both noncyclic checking and for hardirq-unsafe/softirq-unsafe
1075 * checking.
8e18257d 1076 */
8e18257d
PZ
1077
1078/*
1079 * Print a dependency chain entry (this is only done when a deadlock
1080 * has been detected):
1081 */
1082static noinline int
24208ca7 1083print_circular_bug_entry(struct lock_list *target, int depth)
8e18257d
PZ
1084{
1085 if (debug_locks_silent)
1086 return 0;
1087 printk("\n-> #%u", depth);
1088 print_lock_name(target->class);
f943fe0f 1089 printk(KERN_CONT ":\n");
8e18257d
PZ
1090 print_stack_trace(&target->trace, 6);
1091
1092 return 0;
1093}
1094
f4185812
SR
1095static void
1096print_circular_lock_scenario(struct held_lock *src,
1097 struct held_lock *tgt,
1098 struct lock_list *prt)
1099{
1100 struct lock_class *source = hlock_class(src);
1101 struct lock_class *target = hlock_class(tgt);
1102 struct lock_class *parent = prt->class;
1103
1104 /*
1105 * A direct locking problem where unsafe_class lock is taken
1106 * directly by safe_class lock, then all we need to show
1107 * is the deadlock scenario, as it is obvious that the
1108 * unsafe lock is taken under the safe lock.
1109 *
1110 * But if there is a chain instead, where the safe lock takes
1111 * an intermediate lock (middle_class) where this lock is
1112 * not the same as the safe lock, then the lock chain is
1113 * used to describe the problem. Otherwise we would need
1114 * to show a different CPU case for each link in the chain
1115 * from the safe_class lock to the unsafe_class lock.
1116 */
1117 if (parent != source) {
1118 printk("Chain exists of:\n ");
1119 __print_lock_name(source);
f943fe0f 1120 printk(KERN_CONT " --> ");
f4185812 1121 __print_lock_name(parent);
f943fe0f 1122 printk(KERN_CONT " --> ");
f4185812 1123 __print_lock_name(target);
f943fe0f 1124 printk(KERN_CONT "\n\n");
f4185812
SR
1125 }
1126
1127 printk(" Possible unsafe locking scenario:\n\n");
1128 printk(" CPU0 CPU1\n");
1129 printk(" ---- ----\n");
1130 printk(" lock(");
1131 __print_lock_name(target);
f943fe0f 1132 printk(KERN_CONT ");\n");
f4185812
SR
1133 printk(" lock(");
1134 __print_lock_name(parent);
f943fe0f 1135 printk(KERN_CONT ");\n");
f4185812
SR
1136 printk(" lock(");
1137 __print_lock_name(target);
f943fe0f 1138 printk(KERN_CONT ");\n");
f4185812
SR
1139 printk(" lock(");
1140 __print_lock_name(source);
f943fe0f 1141 printk(KERN_CONT ");\n");
f4185812
SR
1142 printk("\n *** DEADLOCK ***\n\n");
1143}
1144
8e18257d
PZ
1145/*
1146 * When a circular dependency is detected, print the
1147 * header first:
1148 */
1149static noinline int
db0002a3
ML
1150print_circular_bug_header(struct lock_list *entry, unsigned int depth,
1151 struct held_lock *check_src,
1152 struct held_lock *check_tgt)
8e18257d
PZ
1153{
1154 struct task_struct *curr = current;
1155
c94aa5ca 1156 if (debug_locks_silent)
8e18257d
PZ
1157 return 0;
1158
b3fbab05
PM
1159 printk("\n");
1160 printk("======================================================\n");
1161 printk("[ INFO: possible circular locking dependency detected ]\n");
fbdc4b9a 1162 print_kernel_ident();
b3fbab05 1163 printk("-------------------------------------------------------\n");
8e18257d 1164 printk("%s/%d is trying to acquire lock:\n",
ba25f9dc 1165 curr->comm, task_pid_nr(curr));
db0002a3 1166 print_lock(check_src);
8e18257d 1167 printk("\nbut task is already holding lock:\n");
db0002a3 1168 print_lock(check_tgt);
8e18257d
PZ
1169 printk("\nwhich lock already depends on the new lock.\n\n");
1170 printk("\nthe existing dependency chain (in reverse order) is:\n");
1171
1172 print_circular_bug_entry(entry, depth);
1173
1174 return 0;
1175}
1176
9e2d551e
ML
1177static inline int class_equal(struct lock_list *entry, void *data)
1178{
1179 return entry->class == data;
1180}
1181
db0002a3
ML
1182static noinline int print_circular_bug(struct lock_list *this,
1183 struct lock_list *target,
1184 struct held_lock *check_src,
1185 struct held_lock *check_tgt)
8e18257d
PZ
1186{
1187 struct task_struct *curr = current;
c94aa5ca 1188 struct lock_list *parent;
f4185812 1189 struct lock_list *first_parent;
24208ca7 1190 int depth;
8e18257d 1191
c94aa5ca 1192 if (!debug_locks_off_graph_unlock() || debug_locks_silent)
8e18257d
PZ
1193 return 0;
1194
db0002a3 1195 if (!save_trace(&this->trace))
8e18257d
PZ
1196 return 0;
1197
c94aa5ca
ML
1198 depth = get_lock_depth(target);
1199
db0002a3 1200 print_circular_bug_header(target, depth, check_src, check_tgt);
c94aa5ca
ML
1201
1202 parent = get_lock_parent(target);
f4185812 1203 first_parent = parent;
c94aa5ca
ML
1204
1205 while (parent) {
1206 print_circular_bug_entry(parent, --depth);
1207 parent = get_lock_parent(parent);
1208 }
8e18257d
PZ
1209
1210 printk("\nother info that might help us debug this:\n\n");
f4185812
SR
1211 print_circular_lock_scenario(check_src, check_tgt,
1212 first_parent);
1213
8e18257d
PZ
1214 lockdep_print_held_locks(curr);
1215
1216 printk("\nstack backtrace:\n");
1217 dump_stack();
1218
1219 return 0;
1220}
1221
db0002a3
ML
1222static noinline int print_bfs_bug(int ret)
1223{
1224 if (!debug_locks_off_graph_unlock())
1225 return 0;
1226
0119fee4
PZ
1227 /*
1228 * Breadth-first-search failed, graph got corrupted?
1229 */
db0002a3
ML
1230 WARN(1, "lockdep bfs error:%d\n", ret);
1231
1232 return 0;
1233}
1234
ef681026 1235static int noop_count(struct lock_list *entry, void *data)
419ca3f1 1236{
ef681026
ML
1237 (*(unsigned long *)data)++;
1238 return 0;
1239}
419ca3f1 1240
5216d530 1241static unsigned long __lockdep_count_forward_deps(struct lock_list *this)
ef681026
ML
1242{
1243 unsigned long count = 0;
1244 struct lock_list *uninitialized_var(target_entry);
419ca3f1 1245
ef681026 1246 __bfs_forwards(this, (void *)&count, noop_count, &target_entry);
419ca3f1 1247
ef681026 1248 return count;
419ca3f1 1249}
419ca3f1
DM
1250unsigned long lockdep_count_forward_deps(struct lock_class *class)
1251{
1252 unsigned long ret, flags;
ef681026
ML
1253 struct lock_list this;
1254
1255 this.parent = NULL;
1256 this.class = class;
419ca3f1
DM
1257
1258 local_irq_save(flags);
0199c4e6 1259 arch_spin_lock(&lockdep_lock);
ef681026 1260 ret = __lockdep_count_forward_deps(&this);
0199c4e6 1261 arch_spin_unlock(&lockdep_lock);
419ca3f1
DM
1262 local_irq_restore(flags);
1263
1264 return ret;
1265}
1266
5216d530 1267static unsigned long __lockdep_count_backward_deps(struct lock_list *this)
419ca3f1 1268{
ef681026
ML
1269 unsigned long count = 0;
1270 struct lock_list *uninitialized_var(target_entry);
419ca3f1 1271
ef681026 1272 __bfs_backwards(this, (void *)&count, noop_count, &target_entry);
419ca3f1 1273
ef681026 1274 return count;
419ca3f1
DM
1275}
1276
1277unsigned long lockdep_count_backward_deps(struct lock_class *class)
1278{
1279 unsigned long ret, flags;
ef681026
ML
1280 struct lock_list this;
1281
1282 this.parent = NULL;
1283 this.class = class;
419ca3f1
DM
1284
1285 local_irq_save(flags);
0199c4e6 1286 arch_spin_lock(&lockdep_lock);
ef681026 1287 ret = __lockdep_count_backward_deps(&this);
0199c4e6 1288 arch_spin_unlock(&lockdep_lock);
419ca3f1
DM
1289 local_irq_restore(flags);
1290
1291 return ret;
1292}
1293
8e18257d
PZ
1294/*
1295 * Prove that the dependency graph starting at <entry> can not
1296 * lead to <target>. Print an error and return 0 if it does.
1297 */
1298static noinline int
db0002a3
ML
1299check_noncircular(struct lock_list *root, struct lock_class *target,
1300 struct lock_list **target_entry)
8e18257d 1301{
db0002a3 1302 int result;
8e18257d 1303
bd6d29c2 1304 debug_atomic_inc(nr_cyclic_checks);
419ca3f1 1305
d7aaba14 1306 result = __bfs_forwards(root, target, class_equal, target_entry);
fbb9ce95 1307
db0002a3
ML
1308 return result;
1309}
c94aa5ca 1310
81d68a96 1311#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
fbb9ce95
IM
1312/*
1313 * Forwards and backwards subgraph searching, for the purposes of
1314 * proving that two subgraphs can be connected by a new dependency
1315 * without creating any illegal irq-safe -> irq-unsafe lock dependency.
1316 */
fbb9ce95 1317
d7aaba14
ML
1318static inline int usage_match(struct lock_list *entry, void *bit)
1319{
1320 return entry->class->usage_mask & (1 << (enum lock_usage_bit)bit);
1321}
1322
1323
1324
fbb9ce95
IM
1325/*
1326 * Find a node in the forwards-direction dependency sub-graph starting
d7aaba14 1327 * at @root->class that matches @bit.
fbb9ce95 1328 *
d7aaba14
ML
1329 * Return 0 if such a node exists in the subgraph, and put that node
1330 * into *@target_entry.
fbb9ce95 1331 *
d7aaba14
ML
1332 * Return 1 otherwise and keep *@target_entry unchanged.
1333 * Return <0 on error.
fbb9ce95 1334 */
d7aaba14
ML
1335static int
1336find_usage_forwards(struct lock_list *root, enum lock_usage_bit bit,
1337 struct lock_list **target_entry)
fbb9ce95 1338{
d7aaba14 1339 int result;
fbb9ce95 1340
bd6d29c2 1341 debug_atomic_inc(nr_find_usage_forwards_checks);
fbb9ce95 1342
d7aaba14
ML
1343 result = __bfs_forwards(root, (void *)bit, usage_match, target_entry);
1344
1345 return result;
fbb9ce95
IM
1346}
1347
1348/*
1349 * Find a node in the backwards-direction dependency sub-graph starting
d7aaba14 1350 * at @root->class that matches @bit.
fbb9ce95 1351 *
d7aaba14
ML
1352 * Return 0 if such a node exists in the subgraph, and put that node
1353 * into *@target_entry.
fbb9ce95 1354 *
d7aaba14
ML
1355 * Return 1 otherwise and keep *@target_entry unchanged.
1356 * Return <0 on error.
fbb9ce95 1357 */
d7aaba14
ML
1358static int
1359find_usage_backwards(struct lock_list *root, enum lock_usage_bit bit,
1360 struct lock_list **target_entry)
fbb9ce95 1361{
d7aaba14 1362 int result;
fbb9ce95 1363
bd6d29c2 1364 debug_atomic_inc(nr_find_usage_backwards_checks);
fbb9ce95 1365
d7aaba14 1366 result = __bfs_backwards(root, (void *)bit, usage_match, target_entry);
f82b217e 1367
d7aaba14 1368 return result;
fbb9ce95
IM
1369}
1370
af012961
PZ
1371static void print_lock_class_header(struct lock_class *class, int depth)
1372{
1373 int bit;
1374
1375 printk("%*s->", depth, "");
1376 print_lock_name(class);
f943fe0f
DV
1377 printk(KERN_CONT " ops: %lu", class->ops);
1378 printk(KERN_CONT " {\n");
af012961
PZ
1379
1380 for (bit = 0; bit < LOCK_USAGE_STATES; bit++) {
1381 if (class->usage_mask & (1 << bit)) {
1382 int len = depth;
1383
1384 len += printk("%*s %s", depth, "", usage_str[bit]);
f943fe0f 1385 len += printk(KERN_CONT " at:\n");
af012961
PZ
1386 print_stack_trace(class->usage_traces + bit, len);
1387 }
1388 }
1389 printk("%*s }\n", depth, "");
1390
f943fe0f
DV
1391 printk("%*s ... key at: [<%p>] %pS\n",
1392 depth, "", class->key, class->key);
af012961
PZ
1393}
1394
1395/*
1396 * printk the shortest lock dependencies from @start to @end in reverse order:
1397 */
1398static void __used
1399print_shortest_lock_dependencies(struct lock_list *leaf,
1400 struct lock_list *root)
1401{
1402 struct lock_list *entry = leaf;
1403 int depth;
1404
1405 /*compute depth from generated tree by BFS*/
1406 depth = get_lock_depth(leaf);
1407
1408 do {
1409 print_lock_class_header(entry->class, depth);
1410 printk("%*s ... acquired at:\n", depth, "");
1411 print_stack_trace(&entry->trace, 2);
1412 printk("\n");
1413
1414 if (depth == 0 && (entry != root)) {
6be8c393 1415 printk("lockdep:%s bad path found in chain graph\n", __func__);
af012961
PZ
1416 break;
1417 }
1418
1419 entry = get_lock_parent(entry);
1420 depth--;
1421 } while (entry && (depth >= 0));
1422
1423 return;
1424}
d7aaba14 1425
3003eba3
SR
1426static void
1427print_irq_lock_scenario(struct lock_list *safe_entry,
1428 struct lock_list *unsafe_entry,
dad3d743
SR
1429 struct lock_class *prev_class,
1430 struct lock_class *next_class)
3003eba3
SR
1431{
1432 struct lock_class *safe_class = safe_entry->class;
1433 struct lock_class *unsafe_class = unsafe_entry->class;
dad3d743 1434 struct lock_class *middle_class = prev_class;
3003eba3
SR
1435
1436 if (middle_class == safe_class)
dad3d743 1437 middle_class = next_class;
3003eba3
SR
1438
1439 /*
1440 * A direct locking problem where unsafe_class lock is taken
1441 * directly by safe_class lock, then all we need to show
1442 * is the deadlock scenario, as it is obvious that the
1443 * unsafe lock is taken under the safe lock.
1444 *
1445 * But if there is a chain instead, where the safe lock takes
1446 * an intermediate lock (middle_class) where this lock is
1447 * not the same as the safe lock, then the lock chain is
1448 * used to describe the problem. Otherwise we would need
1449 * to show a different CPU case for each link in the chain
1450 * from the safe_class lock to the unsafe_class lock.
1451 */
1452 if (middle_class != unsafe_class) {
1453 printk("Chain exists of:\n ");
1454 __print_lock_name(safe_class);
f943fe0f 1455 printk(KERN_CONT " --> ");
3003eba3 1456 __print_lock_name(middle_class);
f943fe0f 1457 printk(KERN_CONT " --> ");
3003eba3 1458 __print_lock_name(unsafe_class);
f943fe0f 1459 printk(KERN_CONT "\n\n");
3003eba3
SR
1460 }
1461
1462 printk(" Possible interrupt unsafe locking scenario:\n\n");
1463 printk(" CPU0 CPU1\n");
1464 printk(" ---- ----\n");
1465 printk(" lock(");
1466 __print_lock_name(unsafe_class);
f943fe0f 1467 printk(KERN_CONT ");\n");
3003eba3
SR
1468 printk(" local_irq_disable();\n");
1469 printk(" lock(");
1470 __print_lock_name(safe_class);
f943fe0f 1471 printk(KERN_CONT ");\n");
3003eba3
SR
1472 printk(" lock(");
1473 __print_lock_name(middle_class);
f943fe0f 1474 printk(KERN_CONT ");\n");
3003eba3
SR
1475 printk(" <Interrupt>\n");
1476 printk(" lock(");
1477 __print_lock_name(safe_class);
f943fe0f 1478 printk(KERN_CONT ");\n");
3003eba3
SR
1479 printk("\n *** DEADLOCK ***\n\n");
1480}
1481
fbb9ce95
IM
1482static int
1483print_bad_irq_dependency(struct task_struct *curr,
24208ca7
ML
1484 struct lock_list *prev_root,
1485 struct lock_list *next_root,
1486 struct lock_list *backwards_entry,
1487 struct lock_list *forwards_entry,
fbb9ce95
IM
1488 struct held_lock *prev,
1489 struct held_lock *next,
1490 enum lock_usage_bit bit1,
1491 enum lock_usage_bit bit2,
1492 const char *irqclass)
1493{
74c383f1 1494 if (!debug_locks_off_graph_unlock() || debug_locks_silent)
fbb9ce95
IM
1495 return 0;
1496
b3fbab05
PM
1497 printk("\n");
1498 printk("======================================================\n");
1499 printk("[ INFO: %s-safe -> %s-unsafe lock order detected ]\n",
fbb9ce95 1500 irqclass, irqclass);
fbdc4b9a 1501 print_kernel_ident();
b3fbab05 1502 printk("------------------------------------------------------\n");
fbb9ce95 1503 printk("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] is trying to acquire:\n",
ba25f9dc 1504 curr->comm, task_pid_nr(curr),
fbb9ce95
IM
1505 curr->hardirq_context, hardirq_count() >> HARDIRQ_SHIFT,
1506 curr->softirq_context, softirq_count() >> SOFTIRQ_SHIFT,
1507 curr->hardirqs_enabled,
1508 curr->softirqs_enabled);
1509 print_lock(next);
1510
1511 printk("\nand this task is already holding:\n");
1512 print_lock(prev);
1513 printk("which would create a new lock dependency:\n");
f82b217e 1514 print_lock_name(hlock_class(prev));
f943fe0f 1515 printk(KERN_CONT " ->");
f82b217e 1516 print_lock_name(hlock_class(next));
f943fe0f 1517 printk(KERN_CONT "\n");
fbb9ce95
IM
1518
1519 printk("\nbut this new dependency connects a %s-irq-safe lock:\n",
1520 irqclass);
24208ca7 1521 print_lock_name(backwards_entry->class);
fbb9ce95
IM
1522 printk("\n... which became %s-irq-safe at:\n", irqclass);
1523
24208ca7 1524 print_stack_trace(backwards_entry->class->usage_traces + bit1, 1);
fbb9ce95
IM
1525
1526 printk("\nto a %s-irq-unsafe lock:\n", irqclass);
24208ca7 1527 print_lock_name(forwards_entry->class);
fbb9ce95
IM
1528 printk("\n... which became %s-irq-unsafe at:\n", irqclass);
1529 printk("...");
1530
24208ca7 1531 print_stack_trace(forwards_entry->class->usage_traces + bit2, 1);
fbb9ce95
IM
1532
1533 printk("\nother info that might help us debug this:\n\n");
dad3d743
SR
1534 print_irq_lock_scenario(backwards_entry, forwards_entry,
1535 hlock_class(prev), hlock_class(next));
3003eba3 1536
fbb9ce95
IM
1537 lockdep_print_held_locks(curr);
1538
f943fe0f 1539 printk("\nthe dependencies between %s-irq-safe lock and the holding lock:\n", irqclass);
24208ca7
ML
1540 if (!save_trace(&prev_root->trace))
1541 return 0;
1542 print_shortest_lock_dependencies(backwards_entry, prev_root);
fbb9ce95 1543
24208ca7
ML
1544 printk("\nthe dependencies between the lock to be acquired");
1545 printk(" and %s-irq-unsafe lock:\n", irqclass);
1546 if (!save_trace(&next_root->trace))
1547 return 0;
1548 print_shortest_lock_dependencies(forwards_entry, next_root);
fbb9ce95
IM
1549
1550 printk("\nstack backtrace:\n");
1551 dump_stack();
1552
1553 return 0;
1554}
1555
1556static int
1557check_usage(struct task_struct *curr, struct held_lock *prev,
1558 struct held_lock *next, enum lock_usage_bit bit_backwards,
1559 enum lock_usage_bit bit_forwards, const char *irqclass)
1560{
1561 int ret;
24208ca7 1562 struct lock_list this, that;
d7aaba14 1563 struct lock_list *uninitialized_var(target_entry);
24208ca7 1564 struct lock_list *uninitialized_var(target_entry1);
d7aaba14
ML
1565
1566 this.parent = NULL;
1567
1568 this.class = hlock_class(prev);
1569 ret = find_usage_backwards(&this, bit_backwards, &target_entry);
af012961
PZ
1570 if (ret < 0)
1571 return print_bfs_bug(ret);
1572 if (ret == 1)
1573 return ret;
d7aaba14 1574
24208ca7
ML
1575 that.parent = NULL;
1576 that.class = hlock_class(next);
1577 ret = find_usage_forwards(&that, bit_forwards, &target_entry1);
af012961
PZ
1578 if (ret < 0)
1579 return print_bfs_bug(ret);
1580 if (ret == 1)
1581 return ret;
fbb9ce95 1582
24208ca7
ML
1583 return print_bad_irq_dependency(curr, &this, &that,
1584 target_entry, target_entry1,
1585 prev, next,
fbb9ce95
IM
1586 bit_backwards, bit_forwards, irqclass);
1587}
1588
4f367d8a
PZ
1589static const char *state_names[] = {
1590#define LOCKDEP_STATE(__STATE) \
b4b136f4 1591 __stringify(__STATE),
4f367d8a
PZ
1592#include "lockdep_states.h"
1593#undef LOCKDEP_STATE
1594};
1595
1596static const char *state_rnames[] = {
1597#define LOCKDEP_STATE(__STATE) \
b4b136f4 1598 __stringify(__STATE)"-READ",
4f367d8a
PZ
1599#include "lockdep_states.h"
1600#undef LOCKDEP_STATE
1601};
1602
1603static inline const char *state_name(enum lock_usage_bit bit)
8e18257d 1604{
4f367d8a
PZ
1605 return (bit & 1) ? state_rnames[bit >> 2] : state_names[bit >> 2];
1606}
8e18257d 1607
4f367d8a
PZ
1608static int exclusive_bit(int new_bit)
1609{
8e18257d 1610 /*
4f367d8a
PZ
1611 * USED_IN
1612 * USED_IN_READ
1613 * ENABLED
1614 * ENABLED_READ
1615 *
1616 * bit 0 - write/read
1617 * bit 1 - used_in/enabled
1618 * bit 2+ state
8e18257d 1619 */
4f367d8a
PZ
1620
1621 int state = new_bit & ~3;
1622 int dir = new_bit & 2;
8e18257d
PZ
1623
1624 /*
4f367d8a 1625 * keep state, bit flip the direction and strip read.
8e18257d 1626 */
4f367d8a
PZ
1627 return state | (dir ^ 2);
1628}
1629
1630static int check_irq_usage(struct task_struct *curr, struct held_lock *prev,
1631 struct held_lock *next, enum lock_usage_bit bit)
1632{
8e18257d 1633 /*
4f367d8a
PZ
1634 * Prove that the new dependency does not connect a hardirq-safe
1635 * lock with a hardirq-unsafe lock - to achieve this we search
8e18257d
PZ
1636 * the backwards-subgraph starting at <prev>, and the
1637 * forwards-subgraph starting at <next>:
1638 */
4f367d8a
PZ
1639 if (!check_usage(curr, prev, next, bit,
1640 exclusive_bit(bit), state_name(bit)))
8e18257d
PZ
1641 return 0;
1642
4f367d8a
PZ
1643 bit++; /* _READ */
1644
cf40bd16 1645 /*
4f367d8a
PZ
1646 * Prove that the new dependency does not connect a hardirq-safe-read
1647 * lock with a hardirq-unsafe lock - to achieve this we search
cf40bd16
NP
1648 * the backwards-subgraph starting at <prev>, and the
1649 * forwards-subgraph starting at <next>:
1650 */
4f367d8a
PZ
1651 if (!check_usage(curr, prev, next, bit,
1652 exclusive_bit(bit), state_name(bit)))
cf40bd16
NP
1653 return 0;
1654
4f367d8a
PZ
1655 return 1;
1656}
1657
1658static int
1659check_prev_add_irq(struct task_struct *curr, struct held_lock *prev,
1660 struct held_lock *next)
1661{
1662#define LOCKDEP_STATE(__STATE) \
1663 if (!check_irq_usage(curr, prev, next, LOCK_USED_IN_##__STATE)) \
cf40bd16 1664 return 0;
4f367d8a
PZ
1665#include "lockdep_states.h"
1666#undef LOCKDEP_STATE
cf40bd16 1667
8e18257d
PZ
1668 return 1;
1669}
1670
1671static void inc_chains(void)
1672{
1673 if (current->hardirq_context)
1674 nr_hardirq_chains++;
1675 else {
1676 if (current->softirq_context)
1677 nr_softirq_chains++;
1678 else
1679 nr_process_chains++;
1680 }
1681}
1682
1683#else
1684
1685static inline int
1686check_prev_add_irq(struct task_struct *curr, struct held_lock *prev,
1687 struct held_lock *next)
1688{
1689 return 1;
1690}
1691
1692static inline void inc_chains(void)
1693{
1694 nr_process_chains++;
1695}
1696
fbb9ce95
IM
1697#endif
1698
48702ecf
SR
1699static void
1700print_deadlock_scenario(struct held_lock *nxt,
1701 struct held_lock *prv)
1702{
1703 struct lock_class *next = hlock_class(nxt);
1704 struct lock_class *prev = hlock_class(prv);
1705
1706 printk(" Possible unsafe locking scenario:\n\n");
1707 printk(" CPU0\n");
1708 printk(" ----\n");
1709 printk(" lock(");
1710 __print_lock_name(prev);
f943fe0f 1711 printk(KERN_CONT ");\n");
48702ecf
SR
1712 printk(" lock(");
1713 __print_lock_name(next);
f943fe0f 1714 printk(KERN_CONT ");\n");
48702ecf
SR
1715 printk("\n *** DEADLOCK ***\n\n");
1716 printk(" May be due to missing lock nesting notation\n\n");
1717}
1718
fbb9ce95
IM
1719static int
1720print_deadlock_bug(struct task_struct *curr, struct held_lock *prev,
1721 struct held_lock *next)
1722{
74c383f1 1723 if (!debug_locks_off_graph_unlock() || debug_locks_silent)
fbb9ce95
IM
1724 return 0;
1725
b3fbab05
PM
1726 printk("\n");
1727 printk("=============================================\n");
1728 printk("[ INFO: possible recursive locking detected ]\n");
fbdc4b9a 1729 print_kernel_ident();
b3fbab05 1730 printk("---------------------------------------------\n");
fbb9ce95 1731 printk("%s/%d is trying to acquire lock:\n",
ba25f9dc 1732 curr->comm, task_pid_nr(curr));
fbb9ce95
IM
1733 print_lock(next);
1734 printk("\nbut task is already holding lock:\n");
1735 print_lock(prev);
1736
1737 printk("\nother info that might help us debug this:\n");
48702ecf 1738 print_deadlock_scenario(next, prev);
fbb9ce95
IM
1739 lockdep_print_held_locks(curr);
1740
1741 printk("\nstack backtrace:\n");
1742 dump_stack();
1743
1744 return 0;
1745}
1746
1747/*
1748 * Check whether we are holding such a class already.
1749 *
1750 * (Note that this has to be done separately, because the graph cannot
1751 * detect such classes of deadlocks.)
1752 *
1753 * Returns: 0 on deadlock detected, 1 on OK, 2 on recursive read
1754 */
1755static int
1756check_deadlock(struct task_struct *curr, struct held_lock *next,
1757 struct lockdep_map *next_instance, int read)
1758{
1759 struct held_lock *prev;
7531e2f3 1760 struct held_lock *nest = NULL;
fbb9ce95
IM
1761 int i;
1762
1763 for (i = 0; i < curr->lockdep_depth; i++) {
1764 prev = curr->held_locks + i;
7531e2f3
PZ
1765
1766 if (prev->instance == next->nest_lock)
1767 nest = prev;
1768
f82b217e 1769 if (hlock_class(prev) != hlock_class(next))
fbb9ce95 1770 continue;
7531e2f3 1771
fbb9ce95
IM
1772 /*
1773 * Allow read-after-read recursion of the same
6c9076ec 1774 * lock class (i.e. read_lock(lock)+read_lock(lock)):
fbb9ce95 1775 */
6c9076ec 1776 if ((read == 2) && prev->read)
fbb9ce95 1777 return 2;
7531e2f3
PZ
1778
1779 /*
1780 * We're holding the nest_lock, which serializes this lock's
1781 * nesting behaviour.
1782 */
1783 if (nest)
1784 return 2;
1785
fbb9ce95
IM
1786 return print_deadlock_bug(curr, prev, next);
1787 }
1788 return 1;
1789}
1790
1791/*
1792 * There was a chain-cache miss, and we are about to add a new dependency
1793 * to a previous lock. We recursively validate the following rules:
1794 *
1795 * - would the adding of the <prev> -> <next> dependency create a
1796 * circular dependency in the graph? [== circular deadlock]
1797 *
1798 * - does the new prev->next dependency connect any hardirq-safe lock
1799 * (in the full backwards-subgraph starting at <prev>) with any
1800 * hardirq-unsafe lock (in the full forwards-subgraph starting at
1801 * <next>)? [== illegal lock inversion with hardirq contexts]
1802 *
1803 * - does the new prev->next dependency connect any softirq-safe lock
1804 * (in the full backwards-subgraph starting at <prev>) with any
1805 * softirq-unsafe lock (in the full forwards-subgraph starting at
1806 * <next>)? [== illegal lock inversion with softirq contexts]
1807 *
1808 * any of these scenarios could lead to a deadlock.
1809 *
1810 * Then if all the validations pass, we add the forwards and backwards
1811 * dependency.
1812 */
1813static int
1814check_prev_add(struct task_struct *curr, struct held_lock *prev,
8a5fd564 1815 struct held_lock *next, int distance, int *stack_saved)
fbb9ce95
IM
1816{
1817 struct lock_list *entry;
1818 int ret;
db0002a3
ML
1819 struct lock_list this;
1820 struct lock_list *uninitialized_var(target_entry);
4726f2a6
YZ
1821 /*
1822 * Static variable, serialized by the graph_lock().
1823 *
1824 * We use this static variable to save the stack trace in case
1825 * we call into this function multiple times due to encountering
1826 * trylocks in the held lock stack.
1827 */
1828 static struct stack_trace trace;
fbb9ce95
IM
1829
1830 /*
1831 * Prove that the new <prev> -> <next> dependency would not
1832 * create a circular dependency in the graph. (We do this by
1833 * forward-recursing into the graph starting at <next>, and
1834 * checking whether we can reach <prev>.)
1835 *
1836 * We are using global variables to control the recursion, to
1837 * keep the stackframe size of the recursive functions low:
1838 */
db0002a3
ML
1839 this.class = hlock_class(next);
1840 this.parent = NULL;
1841 ret = check_noncircular(&this, hlock_class(prev), &target_entry);
1842 if (unlikely(!ret))
1843 return print_circular_bug(&this, target_entry, next, prev);
1844 else if (unlikely(ret < 0))
1845 return print_bfs_bug(ret);
c94aa5ca 1846
8e18257d 1847 if (!check_prev_add_irq(curr, prev, next))
fbb9ce95
IM
1848 return 0;
1849
fbb9ce95
IM
1850 /*
1851 * For recursive read-locks we do all the dependency checks,
1852 * but we dont store read-triggered dependencies (only
1853 * write-triggered dependencies). This ensures that only the
1854 * write-side dependencies matter, and that if for example a
1855 * write-lock never takes any other locks, then the reads are
1856 * equivalent to a NOP.
1857 */
1858 if (next->read == 2 || prev->read == 2)
1859 return 1;
1860 /*
1861 * Is the <prev> -> <next> dependency already present?
1862 *
1863 * (this may occur even though this is a new chain: consider
1864 * e.g. the L1 -> L2 -> L3 -> L4 and the L5 -> L1 -> L2 -> L3
1865 * chains - the second one will be new, but L1 already has
1866 * L2 added to its dependency list, due to the first chain.)
1867 */
f82b217e
DJ
1868 list_for_each_entry(entry, &hlock_class(prev)->locks_after, entry) {
1869 if (entry->class == hlock_class(next)) {
068135e6
JB
1870 if (distance == 1)
1871 entry->distance = 1;
fbb9ce95 1872 return 2;
068135e6 1873 }
fbb9ce95
IM
1874 }
1875
8a5fd564
DV
1876 if (!*stack_saved) {
1877 if (!save_trace(&trace))
1878 return 0;
1879 *stack_saved = 1;
1880 }
4726f2a6 1881
fbb9ce95
IM
1882 /*
1883 * Ok, all validations passed, add the new lock
1884 * to the previous lock's dependency list:
1885 */
83f06168 1886 ret = add_lock_to_list(hlock_class(next),
f82b217e 1887 &hlock_class(prev)->locks_after,
4726f2a6 1888 next->acquire_ip, distance, &trace);
068135e6 1889
fbb9ce95
IM
1890 if (!ret)
1891 return 0;
910b1b2e 1892
83f06168 1893 ret = add_lock_to_list(hlock_class(prev),
f82b217e 1894 &hlock_class(next)->locks_before,
4726f2a6 1895 next->acquire_ip, distance, &trace);
910b1b2e
JP
1896 if (!ret)
1897 return 0;
fbb9ce95
IM
1898
1899 /*
8e18257d
PZ
1900 * Debugging printouts:
1901 */
f82b217e 1902 if (verbose(hlock_class(prev)) || verbose(hlock_class(next))) {
8a5fd564
DV
1903 /* We drop graph lock, so another thread can overwrite trace. */
1904 *stack_saved = 0;
8e18257d
PZ
1905 graph_unlock();
1906 printk("\n new dependency: ");
f82b217e 1907 print_lock_name(hlock_class(prev));
f943fe0f 1908 printk(KERN_CONT " => ");
f82b217e 1909 print_lock_name(hlock_class(next));
f943fe0f 1910 printk(KERN_CONT "\n");
fbb9ce95 1911 dump_stack();
8e18257d 1912 return graph_lock();
fbb9ce95 1913 }
8e18257d
PZ
1914 return 1;
1915}
fbb9ce95 1916
8e18257d
PZ
1917/*
1918 * Add the dependency to all directly-previous locks that are 'relevant'.
1919 * The ones that are relevant are (in increasing distance from curr):
1920 * all consecutive trylock entries and the final non-trylock entry - or
1921 * the end of this context's lock-chain - whichever comes first.
1922 */
1923static int
1924check_prevs_add(struct task_struct *curr, struct held_lock *next)
1925{
1926 int depth = curr->lockdep_depth;
8a5fd564 1927 int stack_saved = 0;
8e18257d 1928 struct held_lock *hlock;
d6d897ce 1929
fbb9ce95 1930 /*
8e18257d
PZ
1931 * Debugging checks.
1932 *
1933 * Depth must not be zero for a non-head lock:
fbb9ce95 1934 */
8e18257d
PZ
1935 if (!depth)
1936 goto out_bug;
fbb9ce95 1937 /*
8e18257d
PZ
1938 * At least two relevant locks must exist for this
1939 * to be a head:
fbb9ce95 1940 */
8e18257d
PZ
1941 if (curr->held_locks[depth].irq_context !=
1942 curr->held_locks[depth-1].irq_context)
1943 goto out_bug;
74c383f1 1944
8e18257d
PZ
1945 for (;;) {
1946 int distance = curr->lockdep_depth - depth + 1;
1b5ff816 1947 hlock = curr->held_locks + depth - 1;
8e18257d
PZ
1948 /*
1949 * Only non-recursive-read entries get new dependencies
1950 * added:
1951 */
1b5ff816 1952 if (hlock->read != 2 && hlock->check) {
4726f2a6 1953 if (!check_prev_add(curr, hlock, next,
8a5fd564 1954 distance, &stack_saved))
8e18257d
PZ
1955 return 0;
1956 /*
1957 * Stop after the first non-trylock entry,
1958 * as non-trylock entries have added their
1959 * own direct dependencies already, so this
1960 * lock is connected to them indirectly:
1961 */
1962 if (!hlock->trylock)
1963 break;
74c383f1 1964 }
8e18257d
PZ
1965 depth--;
1966 /*
1967 * End of lock-stack?
1968 */
1969 if (!depth)
1970 break;
1971 /*
1972 * Stop the search if we cross into another context:
1973 */
1974 if (curr->held_locks[depth].irq_context !=
1975 curr->held_locks[depth-1].irq_context)
1976 break;
fbb9ce95 1977 }
8e18257d
PZ
1978 return 1;
1979out_bug:
1980 if (!debug_locks_off_graph_unlock())
1981 return 0;
fbb9ce95 1982
0119fee4
PZ
1983 /*
1984 * Clearly we all shouldn't be here, but since we made it we
1985 * can reliable say we messed up our state. See the above two
1986 * gotos for reasons why we could possibly end up here.
1987 */
8e18257d 1988 WARN_ON(1);
fbb9ce95 1989
8e18257d 1990 return 0;
fbb9ce95
IM
1991}
1992
8e18257d 1993unsigned long nr_lock_chains;
443cd507 1994struct lock_chain lock_chains[MAX_LOCKDEP_CHAINS];
cd1a28e8 1995int nr_chain_hlocks;
443cd507
HY
1996static u16 chain_hlocks[MAX_LOCKDEP_CHAIN_HLOCKS];
1997
1998struct lock_class *lock_chain_get_class(struct lock_chain *chain, int i)
1999{
2000 return lock_classes + chain_hlocks[chain->base + i];
2001}
8e18257d 2002
9e4e7554
IM
2003/*
2004 * Returns the index of the first held_lock of the current chain
2005 */
2006static inline int get_first_held_lock(struct task_struct *curr,
2007 struct held_lock *hlock)
2008{
2009 int i;
2010 struct held_lock *hlock_curr;
2011
2012 for (i = curr->lockdep_depth - 1; i >= 0; i--) {
2013 hlock_curr = curr->held_locks + i;
2014 if (hlock_curr->irq_context != hlock->irq_context)
2015 break;
2016
2017 }
2018
2019 return ++i;
2020}
2021
5c8a010c 2022#ifdef CONFIG_DEBUG_LOCKDEP
39e2e173
AAF
2023/*
2024 * Returns the next chain_key iteration
2025 */
2026static u64 print_chain_key_iteration(int class_idx, u64 chain_key)
2027{
2028 u64 new_chain_key = iterate_chain_key(chain_key, class_idx);
2029
2030 printk(" class_idx:%d -> chain_key:%016Lx",
2031 class_idx,
2032 (unsigned long long)new_chain_key);
2033 return new_chain_key;
2034}
2035
2036static void
2037print_chain_keys_held_locks(struct task_struct *curr, struct held_lock *hlock_next)
2038{
2039 struct held_lock *hlock;
2040 u64 chain_key = 0;
2041 int depth = curr->lockdep_depth;
2042 int i;
2043
2044 printk("depth: %u\n", depth + 1);
2045 for (i = get_first_held_lock(curr, hlock_next); i < depth; i++) {
2046 hlock = curr->held_locks + i;
2047 chain_key = print_chain_key_iteration(hlock->class_idx, chain_key);
2048
2049 print_lock(hlock);
2050 }
2051
2052 print_chain_key_iteration(hlock_next->class_idx, chain_key);
2053 print_lock(hlock_next);
2054}
2055
2056static void print_chain_keys_chain(struct lock_chain *chain)
2057{
2058 int i;
2059 u64 chain_key = 0;
2060 int class_id;
2061
2062 printk("depth: %u\n", chain->depth);
2063 for (i = 0; i < chain->depth; i++) {
2064 class_id = chain_hlocks[chain->base + i];
2065 chain_key = print_chain_key_iteration(class_id + 1, chain_key);
2066
2067 print_lock_name(lock_classes + class_id);
2068 printk("\n");
2069 }
2070}
2071
2072static void print_collision(struct task_struct *curr,
2073 struct held_lock *hlock_next,
2074 struct lock_chain *chain)
2075{
2076 printk("\n");
2077 printk("======================\n");
2078 printk("[chain_key collision ]\n");
2079 print_kernel_ident();
2080 printk("----------------------\n");
2081 printk("%s/%d: ", current->comm, task_pid_nr(current));
2082 printk("Hash chain already cached but the contents don't match!\n");
2083
2084 printk("Held locks:");
2085 print_chain_keys_held_locks(curr, hlock_next);
2086
2087 printk("Locks in cached chain:");
2088 print_chain_keys_chain(chain);
2089
2090 printk("\nstack backtrace:\n");
2091 dump_stack();
2092}
5c8a010c 2093#endif
39e2e173 2094
9e4e7554
IM
2095/*
2096 * Checks whether the chain and the current held locks are consistent
2097 * in depth and also in content. If they are not it most likely means
2098 * that there was a collision during the calculation of the chain_key.
2099 * Returns: 0 not passed, 1 passed
2100 */
2101static int check_no_collision(struct task_struct *curr,
2102 struct held_lock *hlock,
2103 struct lock_chain *chain)
2104{
2105#ifdef CONFIG_DEBUG_LOCKDEP
2106 int i, j, id;
2107
2108 i = get_first_held_lock(curr, hlock);
2109
39e2e173
AAF
2110 if (DEBUG_LOCKS_WARN_ON(chain->depth != curr->lockdep_depth - (i - 1))) {
2111 print_collision(curr, hlock, chain);
9e4e7554 2112 return 0;
39e2e173 2113 }
9e4e7554
IM
2114
2115 for (j = 0; j < chain->depth - 1; j++, i++) {
2116 id = curr->held_locks[i].class_idx - 1;
2117
39e2e173
AAF
2118 if (DEBUG_LOCKS_WARN_ON(chain_hlocks[chain->base + j] != id)) {
2119 print_collision(curr, hlock, chain);
9e4e7554 2120 return 0;
39e2e173 2121 }
9e4e7554
IM
2122 }
2123#endif
2124 return 1;
2125}
2126
fbb9ce95
IM
2127/*
2128 * Look up a dependency chain. If the key is not present yet then
9e860d00
JP
2129 * add it and return 1 - in this case the new dependency chain is
2130 * validated. If the key is already hashed, return 0.
2131 * (On return with 1 graph_lock is held.)
fbb9ce95 2132 */
443cd507
HY
2133static inline int lookup_chain_cache(struct task_struct *curr,
2134 struct held_lock *hlock,
2135 u64 chain_key)
fbb9ce95 2136{
f82b217e 2137 struct lock_class *class = hlock_class(hlock);
a63f38cc 2138 struct hlist_head *hash_head = chainhashentry(chain_key);
fbb9ce95 2139 struct lock_chain *chain;
e0944ee6 2140 int i, j;
fbb9ce95 2141
0119fee4
PZ
2142 /*
2143 * We might need to take the graph lock, ensure we've got IRQs
2144 * disabled to make this an IRQ-safe lock.. for recursion reasons
2145 * lockdep won't complain about its own locking errors.
2146 */
381a2292
JP
2147 if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
2148 return 0;
fbb9ce95
IM
2149 /*
2150 * We can walk it lock-free, because entries only get added
2151 * to the hash:
2152 */
a63f38cc 2153 hlist_for_each_entry_rcu(chain, hash_head, entry) {
fbb9ce95
IM
2154 if (chain->chain_key == chain_key) {
2155cache_hit:
bd6d29c2 2156 debug_atomic_inc(chain_lookup_hits);
9e4e7554
IM
2157 if (!check_no_collision(curr, hlock, chain))
2158 return 0;
2159
81fc685a 2160 if (very_verbose(class))
755cd900
AM
2161 printk("\nhash chain already cached, key: "
2162 "%016Lx tail class: [%p] %s\n",
2163 (unsigned long long)chain_key,
2164 class->key, class->name);
fbb9ce95
IM
2165 return 0;
2166 }
2167 }
81fc685a 2168 if (very_verbose(class))
755cd900
AM
2169 printk("\nnew hash chain, key: %016Lx tail class: [%p] %s\n",
2170 (unsigned long long)chain_key, class->key, class->name);
fbb9ce95
IM
2171 /*
2172 * Allocate a new chain entry from the static array, and add
2173 * it to the hash:
2174 */
74c383f1
IM
2175 if (!graph_lock())
2176 return 0;
fbb9ce95
IM
2177 /*
2178 * We have to walk the chain again locked - to avoid duplicates:
2179 */
a63f38cc 2180 hlist_for_each_entry(chain, hash_head, entry) {
fbb9ce95 2181 if (chain->chain_key == chain_key) {
74c383f1 2182 graph_unlock();
fbb9ce95
IM
2183 goto cache_hit;
2184 }
2185 }
2186 if (unlikely(nr_lock_chains >= MAX_LOCKDEP_CHAINS)) {
74c383f1
IM
2187 if (!debug_locks_off_graph_unlock())
2188 return 0;
2189
2c522836 2190 print_lockdep_off("BUG: MAX_LOCKDEP_CHAINS too low!");
eedeeabd 2191 dump_stack();
fbb9ce95
IM
2192 return 0;
2193 }
2194 chain = lock_chains + nr_lock_chains++;
2195 chain->chain_key = chain_key;
443cd507 2196 chain->irq_context = hlock->irq_context;
9e4e7554 2197 i = get_first_held_lock(curr, hlock);
443cd507 2198 chain->depth = curr->lockdep_depth + 1 - i;
75dd602a
PZ
2199
2200 BUILD_BUG_ON((1UL << 24) <= ARRAY_SIZE(chain_hlocks));
2201 BUILD_BUG_ON((1UL << 6) <= ARRAY_SIZE(curr->held_locks));
2202 BUILD_BUG_ON((1UL << 8*sizeof(chain_hlocks[0])) <= ARRAY_SIZE(lock_classes));
2203
e0944ee6
SR
2204 if (likely(nr_chain_hlocks + chain->depth <= MAX_LOCKDEP_CHAIN_HLOCKS)) {
2205 chain->base = nr_chain_hlocks;
443cd507 2206 for (j = 0; j < chain->depth - 1; j++, i++) {
f82b217e 2207 int lock_id = curr->held_locks[i].class_idx - 1;
443cd507
HY
2208 chain_hlocks[chain->base + j] = lock_id;
2209 }
2210 chain_hlocks[chain->base + j] = class - lock_classes;
2211 }
75dd602a
PZ
2212
2213 if (nr_chain_hlocks < MAX_LOCKDEP_CHAIN_HLOCKS)
2214 nr_chain_hlocks += chain->depth;
2215
2216#ifdef CONFIG_DEBUG_LOCKDEP
2217 /*
2218 * Important for check_no_collision().
2219 */
2220 if (unlikely(nr_chain_hlocks > MAX_LOCKDEP_CHAIN_HLOCKS)) {
f9af456a 2221 if (!debug_locks_off_graph_unlock())
75dd602a
PZ
2222 return 0;
2223
2224 print_lockdep_off("BUG: MAX_LOCKDEP_CHAIN_HLOCKS too low!");
2225 dump_stack();
2226 return 0;
2227 }
2228#endif
2229
a63f38cc 2230 hlist_add_head_rcu(&chain->entry, hash_head);
bd6d29c2 2231 debug_atomic_inc(chain_lookup_misses);
8e18257d
PZ
2232 inc_chains();
2233
2234 return 1;
2235}
2236
2237static int validate_chain(struct task_struct *curr, struct lockdep_map *lock,
4e6045f1 2238 struct held_lock *hlock, int chain_head, u64 chain_key)
8e18257d
PZ
2239{
2240 /*
2241 * Trylock needs to maintain the stack of held locks, but it
2242 * does not add new dependencies, because trylock can be done
2243 * in any order.
2244 *
2245 * We look up the chain_key and do the O(N^2) check and update of
2246 * the dependencies only if this is a new dependency chain.
2247 * (If lookup_chain_cache() returns with 1 it acquires
2248 * graph_lock for us)
2249 */
fb9edbe9 2250 if (!hlock->trylock && hlock->check &&
443cd507 2251 lookup_chain_cache(curr, hlock, chain_key)) {
8e18257d
PZ
2252 /*
2253 * Check whether last held lock:
2254 *
2255 * - is irq-safe, if this lock is irq-unsafe
2256 * - is softirq-safe, if this lock is hardirq-unsafe
2257 *
2258 * And check whether the new lock's dependency graph
2259 * could lead back to the previous lock.
2260 *
2261 * any of these scenarios could lead to a deadlock. If
2262 * All validations
2263 */
2264 int ret = check_deadlock(curr, hlock, lock, hlock->read);
2265
2266 if (!ret)
2267 return 0;
2268 /*
2269 * Mark recursive read, as we jump over it when
2270 * building dependencies (just like we jump over
2271 * trylock entries):
2272 */
2273 if (ret == 2)
2274 hlock->read = 2;
2275 /*
2276 * Add dependency only if this lock is not the head
2277 * of the chain, and if it's not a secondary read-lock:
2278 */
2279 if (!chain_head && ret != 2)
2280 if (!check_prevs_add(curr, hlock))
2281 return 0;
2282 graph_unlock();
2283 } else
2284 /* after lookup_chain_cache(): */
2285 if (unlikely(!debug_locks))
2286 return 0;
fbb9ce95
IM
2287
2288 return 1;
2289}
8e18257d
PZ
2290#else
2291static inline int validate_chain(struct task_struct *curr,
2292 struct lockdep_map *lock, struct held_lock *hlock,
3aa416b0 2293 int chain_head, u64 chain_key)
8e18257d
PZ
2294{
2295 return 1;
2296}
ca58abcb 2297#endif
fbb9ce95
IM
2298
2299/*
2300 * We are building curr_chain_key incrementally, so double-check
2301 * it from scratch, to make sure that it's done correctly:
2302 */
1d09daa5 2303static void check_chain_key(struct task_struct *curr)
fbb9ce95
IM
2304{
2305#ifdef CONFIG_DEBUG_LOCKDEP
2306 struct held_lock *hlock, *prev_hlock = NULL;
5f18ab5c 2307 unsigned int i;
fbb9ce95
IM
2308 u64 chain_key = 0;
2309
2310 for (i = 0; i < curr->lockdep_depth; i++) {
2311 hlock = curr->held_locks + i;
2312 if (chain_key != hlock->prev_chain_key) {
2313 debug_locks_off();
0119fee4
PZ
2314 /*
2315 * We got mighty confused, our chain keys don't match
2316 * with what we expect, someone trample on our task state?
2317 */
2df8b1d6 2318 WARN(1, "hm#1, depth: %u [%u], %016Lx != %016Lx\n",
fbb9ce95
IM
2319 curr->lockdep_depth, i,
2320 (unsigned long long)chain_key,
2321 (unsigned long long)hlock->prev_chain_key);
fbb9ce95
IM
2322 return;
2323 }
0119fee4
PZ
2324 /*
2325 * Whoops ran out of static storage again?
2326 */
5f18ab5c 2327 if (DEBUG_LOCKS_WARN_ON(hlock->class_idx > MAX_LOCKDEP_KEYS))
381a2292
JP
2328 return;
2329
fbb9ce95
IM
2330 if (prev_hlock && (prev_hlock->irq_context !=
2331 hlock->irq_context))
2332 chain_key = 0;
5f18ab5c 2333 chain_key = iterate_chain_key(chain_key, hlock->class_idx);
fbb9ce95
IM
2334 prev_hlock = hlock;
2335 }
2336 if (chain_key != curr->curr_chain_key) {
2337 debug_locks_off();
0119fee4
PZ
2338 /*
2339 * More smoking hash instead of calculating it, damn see these
2340 * numbers float.. I bet that a pink elephant stepped on my memory.
2341 */
2df8b1d6 2342 WARN(1, "hm#2, depth: %u [%u], %016Lx != %016Lx\n",
fbb9ce95
IM
2343 curr->lockdep_depth, i,
2344 (unsigned long long)chain_key,
2345 (unsigned long long)curr->curr_chain_key);
fbb9ce95
IM
2346 }
2347#endif
2348}
2349
282b5c2f
SR
2350static void
2351print_usage_bug_scenario(struct held_lock *lock)
2352{
2353 struct lock_class *class = hlock_class(lock);
2354
2355 printk(" Possible unsafe locking scenario:\n\n");
2356 printk(" CPU0\n");
2357 printk(" ----\n");
2358 printk(" lock(");
2359 __print_lock_name(class);
f943fe0f 2360 printk(KERN_CONT ");\n");
282b5c2f
SR
2361 printk(" <Interrupt>\n");
2362 printk(" lock(");
2363 __print_lock_name(class);
f943fe0f 2364 printk(KERN_CONT ");\n");
282b5c2f
SR
2365 printk("\n *** DEADLOCK ***\n\n");
2366}
2367
8e18257d
PZ
2368static int
2369print_usage_bug(struct task_struct *curr, struct held_lock *this,
2370 enum lock_usage_bit prev_bit, enum lock_usage_bit new_bit)
2371{
2372 if (!debug_locks_off_graph_unlock() || debug_locks_silent)
2373 return 0;
2374
b3fbab05
PM
2375 printk("\n");
2376 printk("=================================\n");
2377 printk("[ INFO: inconsistent lock state ]\n");
fbdc4b9a 2378 print_kernel_ident();
b3fbab05 2379 printk("---------------------------------\n");
8e18257d
PZ
2380
2381 printk("inconsistent {%s} -> {%s} usage.\n",
2382 usage_str[prev_bit], usage_str[new_bit]);
2383
2384 printk("%s/%d [HC%u[%lu]:SC%u[%lu]:HE%u:SE%u] takes:\n",
ba25f9dc 2385 curr->comm, task_pid_nr(curr),
8e18257d
PZ
2386 trace_hardirq_context(curr), hardirq_count() >> HARDIRQ_SHIFT,
2387 trace_softirq_context(curr), softirq_count() >> SOFTIRQ_SHIFT,
2388 trace_hardirqs_enabled(curr),
2389 trace_softirqs_enabled(curr));
2390 print_lock(this);
2391
2392 printk("{%s} state was registered at:\n", usage_str[prev_bit]);
f82b217e 2393 print_stack_trace(hlock_class(this)->usage_traces + prev_bit, 1);
8e18257d
PZ
2394
2395 print_irqtrace_events(curr);
2396 printk("\nother info that might help us debug this:\n");
282b5c2f
SR
2397 print_usage_bug_scenario(this);
2398
8e18257d
PZ
2399 lockdep_print_held_locks(curr);
2400
2401 printk("\nstack backtrace:\n");
2402 dump_stack();
2403
2404 return 0;
2405}
2406
2407/*
2408 * Print out an error if an invalid bit is set:
2409 */
2410static inline int
2411valid_state(struct task_struct *curr, struct held_lock *this,
2412 enum lock_usage_bit new_bit, enum lock_usage_bit bad_bit)
2413{
f82b217e 2414 if (unlikely(hlock_class(this)->usage_mask & (1 << bad_bit)))
8e18257d
PZ
2415 return print_usage_bug(curr, this, bad_bit, new_bit);
2416 return 1;
2417}
2418
2419static int mark_lock(struct task_struct *curr, struct held_lock *this,
2420 enum lock_usage_bit new_bit);
2421
81d68a96 2422#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
fbb9ce95
IM
2423
2424/*
2425 * print irq inversion bug:
2426 */
2427static int
24208ca7
ML
2428print_irq_inversion_bug(struct task_struct *curr,
2429 struct lock_list *root, struct lock_list *other,
fbb9ce95
IM
2430 struct held_lock *this, int forwards,
2431 const char *irqclass)
2432{
dad3d743
SR
2433 struct lock_list *entry = other;
2434 struct lock_list *middle = NULL;
2435 int depth;
2436
74c383f1 2437 if (!debug_locks_off_graph_unlock() || debug_locks_silent)
fbb9ce95
IM
2438 return 0;
2439
b3fbab05
PM
2440 printk("\n");
2441 printk("=========================================================\n");
2442 printk("[ INFO: possible irq lock inversion dependency detected ]\n");
fbdc4b9a 2443 print_kernel_ident();
b3fbab05 2444 printk("---------------------------------------------------------\n");
fbb9ce95 2445 printk("%s/%d just changed the state of lock:\n",
ba25f9dc 2446 curr->comm, task_pid_nr(curr));
fbb9ce95
IM
2447 print_lock(this);
2448 if (forwards)
26575e28 2449 printk("but this lock took another, %s-unsafe lock in the past:\n", irqclass);
fbb9ce95 2450 else
26575e28 2451 printk("but this lock was taken by another, %s-safe lock in the past:\n", irqclass);
24208ca7 2452 print_lock_name(other->class);
fbb9ce95
IM
2453 printk("\n\nand interrupts could create inverse lock ordering between them.\n\n");
2454
2455 printk("\nother info that might help us debug this:\n");
dad3d743
SR
2456
2457 /* Find a middle lock (if one exists) */
2458 depth = get_lock_depth(other);
2459 do {
2460 if (depth == 0 && (entry != root)) {
2461 printk("lockdep:%s bad path found in chain graph\n", __func__);
2462 break;
2463 }
2464 middle = entry;
2465 entry = get_lock_parent(entry);
2466 depth--;
2467 } while (entry && entry != root && (depth >= 0));
2468 if (forwards)
2469 print_irq_lock_scenario(root, other,
2470 middle ? middle->class : root->class, other->class);
2471 else
2472 print_irq_lock_scenario(other, root,
2473 middle ? middle->class : other->class, root->class);
2474
fbb9ce95
IM
2475 lockdep_print_held_locks(curr);
2476
24208ca7
ML
2477 printk("\nthe shortest dependencies between 2nd lock and 1st lock:\n");
2478 if (!save_trace(&root->trace))
2479 return 0;
2480 print_shortest_lock_dependencies(other, root);
fbb9ce95
IM
2481
2482 printk("\nstack backtrace:\n");
2483 dump_stack();
2484
2485 return 0;
2486}
2487
2488/*
2489 * Prove that in the forwards-direction subgraph starting at <this>
2490 * there is no lock matching <mask>:
2491 */
2492static int
2493check_usage_forwards(struct task_struct *curr, struct held_lock *this,
2494 enum lock_usage_bit bit, const char *irqclass)
2495{
2496 int ret;
d7aaba14
ML
2497 struct lock_list root;
2498 struct lock_list *uninitialized_var(target_entry);
fbb9ce95 2499
d7aaba14
ML
2500 root.parent = NULL;
2501 root.class = hlock_class(this);
2502 ret = find_usage_forwards(&root, bit, &target_entry);
af012961
PZ
2503 if (ret < 0)
2504 return print_bfs_bug(ret);
2505 if (ret == 1)
2506 return ret;
fbb9ce95 2507
24208ca7 2508 return print_irq_inversion_bug(curr, &root, target_entry,
d7aaba14 2509 this, 1, irqclass);
fbb9ce95
IM
2510}
2511
2512/*
2513 * Prove that in the backwards-direction subgraph starting at <this>
2514 * there is no lock matching <mask>:
2515 */
2516static int
2517check_usage_backwards(struct task_struct *curr, struct held_lock *this,
2518 enum lock_usage_bit bit, const char *irqclass)
2519{
2520 int ret;
d7aaba14
ML
2521 struct lock_list root;
2522 struct lock_list *uninitialized_var(target_entry);
fbb9ce95 2523
d7aaba14
ML
2524 root.parent = NULL;
2525 root.class = hlock_class(this);
2526 ret = find_usage_backwards(&root, bit, &target_entry);
af012961
PZ
2527 if (ret < 0)
2528 return print_bfs_bug(ret);
2529 if (ret == 1)
2530 return ret;
fbb9ce95 2531
24208ca7 2532 return print_irq_inversion_bug(curr, &root, target_entry,
48d50674 2533 this, 0, irqclass);
fbb9ce95
IM
2534}
2535
3117df04 2536void print_irqtrace_events(struct task_struct *curr)
fbb9ce95
IM
2537{
2538 printk("irq event stamp: %u\n", curr->irq_events);
f943fe0f
DV
2539 printk("hardirqs last enabled at (%u): [<%p>] %pS\n",
2540 curr->hardirq_enable_event, (void *)curr->hardirq_enable_ip,
2541 (void *)curr->hardirq_enable_ip);
2542 printk("hardirqs last disabled at (%u): [<%p>] %pS\n",
2543 curr->hardirq_disable_event, (void *)curr->hardirq_disable_ip,
2544 (void *)curr->hardirq_disable_ip);
2545 printk("softirqs last enabled at (%u): [<%p>] %pS\n",
2546 curr->softirq_enable_event, (void *)curr->softirq_enable_ip,
2547 (void *)curr->softirq_enable_ip);
2548 printk("softirqs last disabled at (%u): [<%p>] %pS\n",
2549 curr->softirq_disable_event, (void *)curr->softirq_disable_ip,
2550 (void *)curr->softirq_disable_ip);
fbb9ce95
IM
2551}
2552
cd95302d 2553static int HARDIRQ_verbose(struct lock_class *class)
fbb9ce95 2554{
8e18257d
PZ
2555#if HARDIRQ_VERBOSE
2556 return class_filter(class);
2557#endif
fbb9ce95
IM
2558 return 0;
2559}
2560
cd95302d 2561static int SOFTIRQ_verbose(struct lock_class *class)
fbb9ce95 2562{
8e18257d
PZ
2563#if SOFTIRQ_VERBOSE
2564 return class_filter(class);
2565#endif
2566 return 0;
fbb9ce95
IM
2567}
2568
cd95302d 2569static int RECLAIM_FS_verbose(struct lock_class *class)
cf40bd16
NP
2570{
2571#if RECLAIM_VERBOSE
2572 return class_filter(class);
2573#endif
2574 return 0;
2575}
2576
fbb9ce95
IM
2577#define STRICT_READ_CHECKS 1
2578
cd95302d
PZ
2579static int (*state_verbose_f[])(struct lock_class *class) = {
2580#define LOCKDEP_STATE(__STATE) \
2581 __STATE##_verbose,
2582#include "lockdep_states.h"
2583#undef LOCKDEP_STATE
2584};
2585
2586static inline int state_verbose(enum lock_usage_bit bit,
2587 struct lock_class *class)
2588{
2589 return state_verbose_f[bit >> 2](class);
2590}
2591
42c50d54
PZ
2592typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
2593 enum lock_usage_bit bit, const char *name);
2594
6a6904d3 2595static int
1c21f14e
PZ
2596mark_lock_irq(struct task_struct *curr, struct held_lock *this,
2597 enum lock_usage_bit new_bit)
6a6904d3 2598{
f989209e 2599 int excl_bit = exclusive_bit(new_bit);
9d3651a2 2600 int read = new_bit & 1;
42c50d54
PZ
2601 int dir = new_bit & 2;
2602
38aa2714
PZ
2603 /*
2604 * mark USED_IN has to look forwards -- to ensure no dependency
2605 * has ENABLED state, which would allow recursion deadlocks.
2606 *
2607 * mark ENABLED has to look backwards -- to ensure no dependee
2608 * has USED_IN state, which, again, would allow recursion deadlocks.
2609 */
42c50d54
PZ
2610 check_usage_f usage = dir ?
2611 check_usage_backwards : check_usage_forwards;
f989209e 2612
38aa2714
PZ
2613 /*
2614 * Validate that this particular lock does not have conflicting
2615 * usage states.
2616 */
6a6904d3
PZ
2617 if (!valid_state(curr, this, new_bit, excl_bit))
2618 return 0;
42c50d54 2619
38aa2714
PZ
2620 /*
2621 * Validate that the lock dependencies don't have conflicting usage
2622 * states.
2623 */
2624 if ((!read || !dir || STRICT_READ_CHECKS) &&
1c21f14e 2625 !usage(curr, this, excl_bit, state_name(new_bit & ~1)))
6a6904d3 2626 return 0;
780e820b 2627
38aa2714
PZ
2628 /*
2629 * Check for read in write conflicts
2630 */
2631 if (!read) {
2632 if (!valid_state(curr, this, new_bit, excl_bit + 1))
2633 return 0;
2634
2635 if (STRICT_READ_CHECKS &&
4f367d8a
PZ
2636 !usage(curr, this, excl_bit + 1,
2637 state_name(new_bit + 1)))
38aa2714
PZ
2638 return 0;
2639 }
780e820b 2640
cd95302d 2641 if (state_verbose(new_bit, hlock_class(this)))
6a6904d3
PZ
2642 return 2;
2643
2644 return 1;
2645}
2646
cf40bd16 2647enum mark_type {
36bfb9bb
PZ
2648#define LOCKDEP_STATE(__STATE) __STATE,
2649#include "lockdep_states.h"
2650#undef LOCKDEP_STATE
cf40bd16
NP
2651};
2652
fbb9ce95
IM
2653/*
2654 * Mark all held locks with a usage bit:
2655 */
1d09daa5 2656static int
cf40bd16 2657mark_held_locks(struct task_struct *curr, enum mark_type mark)
fbb9ce95
IM
2658{
2659 enum lock_usage_bit usage_bit;
2660 struct held_lock *hlock;
2661 int i;
2662
2663 for (i = 0; i < curr->lockdep_depth; i++) {
2664 hlock = curr->held_locks + i;
2665
cf2ad4d1
PZ
2666 usage_bit = 2 + (mark << 2); /* ENABLED */
2667 if (hlock->read)
2668 usage_bit += 1; /* READ */
2669
2670 BUG_ON(usage_bit >= LOCK_USAGE_STATES);
cf40bd16 2671
34d0ed5e 2672 if (!hlock->check)
efbe2eee
PZ
2673 continue;
2674
4ff773bb 2675 if (!mark_lock(curr, hlock, usage_bit))
fbb9ce95
IM
2676 return 0;
2677 }
2678
2679 return 1;
2680}
2681
fbb9ce95
IM
2682/*
2683 * Hardirqs will be enabled:
2684 */
dd4e5d3a 2685static void __trace_hardirqs_on_caller(unsigned long ip)
fbb9ce95
IM
2686{
2687 struct task_struct *curr = current;
fbb9ce95 2688
fbb9ce95
IM
2689 /* we'll do an OFF -> ON transition: */
2690 curr->hardirqs_enabled = 1;
fbb9ce95 2691
fbb9ce95
IM
2692 /*
2693 * We are going to turn hardirqs on, so set the
2694 * usage bit for all held locks:
2695 */
cf40bd16 2696 if (!mark_held_locks(curr, HARDIRQ))
fbb9ce95
IM
2697 return;
2698 /*
2699 * If we have softirqs enabled, then set the usage
2700 * bit for all held locks. (disabled hardirqs prevented
2701 * this bit from being set before)
2702 */
2703 if (curr->softirqs_enabled)
cf40bd16 2704 if (!mark_held_locks(curr, SOFTIRQ))
fbb9ce95
IM
2705 return;
2706
8e18257d
PZ
2707 curr->hardirq_enable_ip = ip;
2708 curr->hardirq_enable_event = ++curr->irq_events;
bd6d29c2 2709 debug_atomic_inc(hardirqs_on_events);
8e18257d 2710}
dd4e5d3a 2711
b35f8305 2712__visible void trace_hardirqs_on_caller(unsigned long ip)
dd4e5d3a
PZ
2713{
2714 time_hardirqs_on(CALLER_ADDR0, ip);
2715
2716 if (unlikely(!debug_locks || current->lockdep_recursion))
2717 return;
2718
7d36b26b
PZ
2719 if (unlikely(current->hardirqs_enabled)) {
2720 /*
2721 * Neither irq nor preemption are disabled here
2722 * so this is racy by nature but losing one hit
2723 * in a stat is not a big deal.
2724 */
2725 __debug_atomic_inc(redundant_hardirqs_on);
2726 return;
2727 }
2728
0119fee4
PZ
2729 /*
2730 * We're enabling irqs and according to our state above irqs weren't
2731 * already enabled, yet we find the hardware thinks they are in fact
2732 * enabled.. someone messed up their IRQ state tracing.
2733 */
dd4e5d3a
PZ
2734 if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
2735 return;
2736
0119fee4
PZ
2737 /*
2738 * See the fine text that goes along with this variable definition.
2739 */
7d36b26b
PZ
2740 if (DEBUG_LOCKS_WARN_ON(unlikely(early_boot_irqs_disabled)))
2741 return;
2742
0119fee4
PZ
2743 /*
2744 * Can't allow enabling interrupts while in an interrupt handler,
2745 * that's general bad form and such. Recursion, limited stack etc..
2746 */
7d36b26b
PZ
2747 if (DEBUG_LOCKS_WARN_ON(current->hardirq_context))
2748 return;
2749
dd4e5d3a
PZ
2750 current->lockdep_recursion = 1;
2751 __trace_hardirqs_on_caller(ip);
2752 current->lockdep_recursion = 0;
2753}
81d68a96 2754EXPORT_SYMBOL(trace_hardirqs_on_caller);
8e18257d 2755
1d09daa5 2756void trace_hardirqs_on(void)
81d68a96
SR
2757{
2758 trace_hardirqs_on_caller(CALLER_ADDR0);
2759}
8e18257d
PZ
2760EXPORT_SYMBOL(trace_hardirqs_on);
2761
2762/*
2763 * Hardirqs were disabled:
2764 */
b35f8305 2765__visible void trace_hardirqs_off_caller(unsigned long ip)
8e18257d
PZ
2766{
2767 struct task_struct *curr = current;
2768
6afe40b4 2769 time_hardirqs_off(CALLER_ADDR0, ip);
81d68a96 2770
8e18257d
PZ
2771 if (unlikely(!debug_locks || current->lockdep_recursion))
2772 return;
2773
0119fee4
PZ
2774 /*
2775 * So we're supposed to get called after you mask local IRQs, but for
2776 * some reason the hardware doesn't quite think you did a proper job.
2777 */
8e18257d
PZ
2778 if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
2779 return;
2780
2781 if (curr->hardirqs_enabled) {
2782 /*
2783 * We have done an ON -> OFF transition:
2784 */
2785 curr->hardirqs_enabled = 0;
6afe40b4 2786 curr->hardirq_disable_ip = ip;
8e18257d 2787 curr->hardirq_disable_event = ++curr->irq_events;
bd6d29c2 2788 debug_atomic_inc(hardirqs_off_events);
8e18257d 2789 } else
bd6d29c2 2790 debug_atomic_inc(redundant_hardirqs_off);
8e18257d 2791}
81d68a96 2792EXPORT_SYMBOL(trace_hardirqs_off_caller);
8e18257d 2793
1d09daa5 2794void trace_hardirqs_off(void)
81d68a96
SR
2795{
2796 trace_hardirqs_off_caller(CALLER_ADDR0);
2797}
8e18257d
PZ
2798EXPORT_SYMBOL(trace_hardirqs_off);
2799
2800/*
2801 * Softirqs will be enabled:
2802 */
2803void trace_softirqs_on(unsigned long ip)
2804{
2805 struct task_struct *curr = current;
2806
dd4e5d3a 2807 if (unlikely(!debug_locks || current->lockdep_recursion))
8e18257d
PZ
2808 return;
2809
0119fee4
PZ
2810 /*
2811 * We fancy IRQs being disabled here, see softirq.c, avoids
2812 * funny state and nesting things.
2813 */
8e18257d
PZ
2814 if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
2815 return;
2816
2817 if (curr->softirqs_enabled) {
bd6d29c2 2818 debug_atomic_inc(redundant_softirqs_on);
8e18257d
PZ
2819 return;
2820 }
2821
dd4e5d3a 2822 current->lockdep_recursion = 1;
8e18257d
PZ
2823 /*
2824 * We'll do an OFF -> ON transition:
2825 */
2826 curr->softirqs_enabled = 1;
2827 curr->softirq_enable_ip = ip;
2828 curr->softirq_enable_event = ++curr->irq_events;
bd6d29c2 2829 debug_atomic_inc(softirqs_on_events);
8e18257d
PZ
2830 /*
2831 * We are going to turn softirqs on, so set the
2832 * usage bit for all held locks, if hardirqs are
2833 * enabled too:
2834 */
2835 if (curr->hardirqs_enabled)
cf40bd16 2836 mark_held_locks(curr, SOFTIRQ);
dd4e5d3a 2837 current->lockdep_recursion = 0;
8e18257d
PZ
2838}
2839
2840/*
2841 * Softirqs were disabled:
2842 */
2843void trace_softirqs_off(unsigned long ip)
2844{
2845 struct task_struct *curr = current;
2846
dd4e5d3a 2847 if (unlikely(!debug_locks || current->lockdep_recursion))
8e18257d
PZ
2848 return;
2849
0119fee4
PZ
2850 /*
2851 * We fancy IRQs being disabled here, see softirq.c
2852 */
8e18257d
PZ
2853 if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
2854 return;
2855
2856 if (curr->softirqs_enabled) {
2857 /*
2858 * We have done an ON -> OFF transition:
2859 */
2860 curr->softirqs_enabled = 0;
2861 curr->softirq_disable_ip = ip;
2862 curr->softirq_disable_event = ++curr->irq_events;
bd6d29c2 2863 debug_atomic_inc(softirqs_off_events);
0119fee4
PZ
2864 /*
2865 * Whoops, we wanted softirqs off, so why aren't they?
2866 */
8e18257d
PZ
2867 DEBUG_LOCKS_WARN_ON(!softirq_count());
2868 } else
bd6d29c2 2869 debug_atomic_inc(redundant_softirqs_off);
8e18257d
PZ
2870}
2871
2f850181 2872static void __lockdep_trace_alloc(gfp_t gfp_mask, unsigned long flags)
cf40bd16
NP
2873{
2874 struct task_struct *curr = current;
2875
2876 if (unlikely(!debug_locks))
2877 return;
2878
2879 /* no reclaim without waiting on it */
d0164adc 2880 if (!(gfp_mask & __GFP_DIRECT_RECLAIM))
cf40bd16
NP
2881 return;
2882
2883 /* this guy won't enter reclaim */
2884 if ((curr->flags & PF_MEMALLOC) && !(gfp_mask & __GFP_NOMEMALLOC))
2885 return;
2886
2887 /* We're only interested __GFP_FS allocations for now */
2888 if (!(gfp_mask & __GFP_FS))
2889 return;
2890
0119fee4
PZ
2891 /*
2892 * Oi! Can't be having __GFP_FS allocations with IRQs disabled.
2893 */
2f850181 2894 if (DEBUG_LOCKS_WARN_ON(irqs_disabled_flags(flags)))
cf40bd16
NP
2895 return;
2896
2897 mark_held_locks(curr, RECLAIM_FS);
2898}
2899
2f850181
PZ
2900static void check_flags(unsigned long flags);
2901
2902void lockdep_trace_alloc(gfp_t gfp_mask)
2903{
2904 unsigned long flags;
2905
2906 if (unlikely(current->lockdep_recursion))
2907 return;
2908
2909 raw_local_irq_save(flags);
2910 check_flags(flags);
2911 current->lockdep_recursion = 1;
2912 __lockdep_trace_alloc(gfp_mask, flags);
2913 current->lockdep_recursion = 0;
2914 raw_local_irq_restore(flags);
2915}
2916
8e18257d
PZ
2917static int mark_irqflags(struct task_struct *curr, struct held_lock *hlock)
2918{
2919 /*
2920 * If non-trylock use in a hardirq or softirq context, then
2921 * mark the lock as used in these contexts:
2922 */
2923 if (!hlock->trylock) {
2924 if (hlock->read) {
2925 if (curr->hardirq_context)
2926 if (!mark_lock(curr, hlock,
2927 LOCK_USED_IN_HARDIRQ_READ))
2928 return 0;
2929 if (curr->softirq_context)
2930 if (!mark_lock(curr, hlock,
2931 LOCK_USED_IN_SOFTIRQ_READ))
2932 return 0;
2933 } else {
2934 if (curr->hardirq_context)
2935 if (!mark_lock(curr, hlock, LOCK_USED_IN_HARDIRQ))
2936 return 0;
2937 if (curr->softirq_context)
2938 if (!mark_lock(curr, hlock, LOCK_USED_IN_SOFTIRQ))
2939 return 0;
2940 }
2941 }
2942 if (!hlock->hardirqs_off) {
2943 if (hlock->read) {
2944 if (!mark_lock(curr, hlock,
4fc95e86 2945 LOCK_ENABLED_HARDIRQ_READ))
8e18257d
PZ
2946 return 0;
2947 if (curr->softirqs_enabled)
2948 if (!mark_lock(curr, hlock,
4fc95e86 2949 LOCK_ENABLED_SOFTIRQ_READ))
8e18257d
PZ
2950 return 0;
2951 } else {
2952 if (!mark_lock(curr, hlock,
4fc95e86 2953 LOCK_ENABLED_HARDIRQ))
8e18257d
PZ
2954 return 0;
2955 if (curr->softirqs_enabled)
2956 if (!mark_lock(curr, hlock,
4fc95e86 2957 LOCK_ENABLED_SOFTIRQ))
8e18257d
PZ
2958 return 0;
2959 }
2960 }
2961
cf40bd16
NP
2962 /*
2963 * We reuse the irq context infrastructure more broadly as a general
2964 * context checking code. This tests GFP_FS recursion (a lock taken
2965 * during reclaim for a GFP_FS allocation is held over a GFP_FS
2966 * allocation).
2967 */
2968 if (!hlock->trylock && (curr->lockdep_reclaim_gfp & __GFP_FS)) {
2969 if (hlock->read) {
2970 if (!mark_lock(curr, hlock, LOCK_USED_IN_RECLAIM_FS_READ))
2971 return 0;
2972 } else {
2973 if (!mark_lock(curr, hlock, LOCK_USED_IN_RECLAIM_FS))
2974 return 0;
2975 }
2976 }
2977
8e18257d
PZ
2978 return 1;
2979}
2980
c2469756
BF
2981static inline unsigned int task_irq_context(struct task_struct *task)
2982{
2983 return 2 * !!task->hardirq_context + !!task->softirq_context;
2984}
2985
8e18257d
PZ
2986static int separate_irq_context(struct task_struct *curr,
2987 struct held_lock *hlock)
2988{
2989 unsigned int depth = curr->lockdep_depth;
2990
2991 /*
2992 * Keep track of points where we cross into an interrupt context:
2993 */
8e18257d
PZ
2994 if (depth) {
2995 struct held_lock *prev_hlock;
2996
2997 prev_hlock = curr->held_locks + depth-1;
2998 /*
2999 * If we cross into another context, reset the
3000 * hash key (this also prevents the checking and the
3001 * adding of the dependency to 'prev'):
3002 */
3003 if (prev_hlock->irq_context != hlock->irq_context)
3004 return 1;
3005 }
3006 return 0;
fbb9ce95
IM
3007}
3008
0119fee4 3009#else /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
fbb9ce95 3010
8e18257d
PZ
3011static inline
3012int mark_lock_irq(struct task_struct *curr, struct held_lock *this,
3013 enum lock_usage_bit new_bit)
fbb9ce95 3014{
0119fee4 3015 WARN_ON(1); /* Impossible innit? when we don't have TRACE_IRQFLAG */
8e18257d
PZ
3016 return 1;
3017}
fbb9ce95 3018
8e18257d
PZ
3019static inline int mark_irqflags(struct task_struct *curr,
3020 struct held_lock *hlock)
3021{
3022 return 1;
3023}
fbb9ce95 3024
c2469756
BF
3025static inline unsigned int task_irq_context(struct task_struct *task)
3026{
3027 return 0;
3028}
3029
8e18257d
PZ
3030static inline int separate_irq_context(struct task_struct *curr,
3031 struct held_lock *hlock)
3032{
3033 return 0;
fbb9ce95
IM
3034}
3035
868a23a8
PZ
3036void lockdep_trace_alloc(gfp_t gfp_mask)
3037{
3038}
3039
0119fee4 3040#endif /* defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING) */
fbb9ce95
IM
3041
3042/*
8e18257d 3043 * Mark a lock with a usage bit, and validate the state transition:
fbb9ce95 3044 */
1d09daa5 3045static int mark_lock(struct task_struct *curr, struct held_lock *this,
0764d23c 3046 enum lock_usage_bit new_bit)
fbb9ce95 3047{
8e18257d 3048 unsigned int new_mask = 1 << new_bit, ret = 1;
fbb9ce95
IM
3049
3050 /*
8e18257d
PZ
3051 * If already set then do not dirty the cacheline,
3052 * nor do any checks:
fbb9ce95 3053 */
f82b217e 3054 if (likely(hlock_class(this)->usage_mask & new_mask))
8e18257d
PZ
3055 return 1;
3056
3057 if (!graph_lock())
3058 return 0;
fbb9ce95 3059 /*
25985edc 3060 * Make sure we didn't race:
fbb9ce95 3061 */
f82b217e 3062 if (unlikely(hlock_class(this)->usage_mask & new_mask)) {
8e18257d
PZ
3063 graph_unlock();
3064 return 1;
3065 }
fbb9ce95 3066
f82b217e 3067 hlock_class(this)->usage_mask |= new_mask;
fbb9ce95 3068
f82b217e 3069 if (!save_trace(hlock_class(this)->usage_traces + new_bit))
8e18257d 3070 return 0;
fbb9ce95 3071
8e18257d 3072 switch (new_bit) {
5346417e
PZ
3073#define LOCKDEP_STATE(__STATE) \
3074 case LOCK_USED_IN_##__STATE: \
3075 case LOCK_USED_IN_##__STATE##_READ: \
3076 case LOCK_ENABLED_##__STATE: \
3077 case LOCK_ENABLED_##__STATE##_READ:
3078#include "lockdep_states.h"
3079#undef LOCKDEP_STATE
8e18257d
PZ
3080 ret = mark_lock_irq(curr, this, new_bit);
3081 if (!ret)
3082 return 0;
3083 break;
3084 case LOCK_USED:
bd6d29c2 3085 debug_atomic_dec(nr_unused_locks);
8e18257d
PZ
3086 break;
3087 default:
3088 if (!debug_locks_off_graph_unlock())
3089 return 0;
3090 WARN_ON(1);
3091 return 0;
3092 }
fbb9ce95 3093
8e18257d
PZ
3094 graph_unlock();
3095
3096 /*
3097 * We must printk outside of the graph_lock:
3098 */
3099 if (ret == 2) {
3100 printk("\nmarked lock as {%s}:\n", usage_str[new_bit]);
3101 print_lock(this);
3102 print_irqtrace_events(curr);
3103 dump_stack();
3104 }
3105
3106 return ret;
3107}
fbb9ce95
IM
3108
3109/*
3110 * Initialize a lock instance's lock-class mapping info:
3111 */
3112void lockdep_init_map(struct lockdep_map *lock, const char *name,
4dfbb9d8 3113 struct lock_class_key *key, int subclass)
fbb9ce95 3114{
d3d03d4f
YZ
3115 int i;
3116
3117 kmemcheck_mark_initialized(lock, sizeof(*lock));
3118
3119 for (i = 0; i < NR_LOCKDEP_CACHING_CLASSES; i++)
3120 lock->class_cache[i] = NULL;
62016250 3121
c8a25005
PZ
3122#ifdef CONFIG_LOCK_STAT
3123 lock->cpu = raw_smp_processor_id();
3124#endif
3125
0119fee4
PZ
3126 /*
3127 * Can't be having no nameless bastards around this place!
3128 */
c8a25005
PZ
3129 if (DEBUG_LOCKS_WARN_ON(!name)) {
3130 lock->name = "NULL";
fbb9ce95 3131 return;
c8a25005
PZ
3132 }
3133
3134 lock->name = name;
fbb9ce95 3135
0119fee4
PZ
3136 /*
3137 * No key, no joy, we need to hash something.
3138 */
fbb9ce95
IM
3139 if (DEBUG_LOCKS_WARN_ON(!key))
3140 return;
fbb9ce95
IM
3141 /*
3142 * Sanity check, the lock-class key must be persistent:
3143 */
3144 if (!static_obj(key)) {
3145 printk("BUG: key %p not in .data!\n", key);
0119fee4
PZ
3146 /*
3147 * What it says above ^^^^^, I suggest you read it.
3148 */
fbb9ce95
IM
3149 DEBUG_LOCKS_WARN_ON(1);
3150 return;
3151 }
fbb9ce95 3152 lock->key = key;
c8a25005
PZ
3153
3154 if (unlikely(!debug_locks))
3155 return;
3156
35a9393c
PZ
3157 if (subclass) {
3158 unsigned long flags;
3159
3160 if (DEBUG_LOCKS_WARN_ON(current->lockdep_recursion))
3161 return;
3162
3163 raw_local_irq_save(flags);
3164 current->lockdep_recursion = 1;
4dfbb9d8 3165 register_lock_class(lock, subclass, 1);
35a9393c
PZ
3166 current->lockdep_recursion = 0;
3167 raw_local_irq_restore(flags);
3168 }
fbb9ce95 3169}
fbb9ce95
IM
3170EXPORT_SYMBOL_GPL(lockdep_init_map);
3171
1704f47b 3172struct lock_class_key __lockdep_no_validate__;
ea6749c7 3173EXPORT_SYMBOL_GPL(__lockdep_no_validate__);
1704f47b 3174
d0945950
ML
3175static int
3176print_lock_nested_lock_not_held(struct task_struct *curr,
3177 struct held_lock *hlock,
3178 unsigned long ip)
3179{
3180 if (!debug_locks_off())
3181 return 0;
3182 if (debug_locks_silent)
3183 return 0;
3184
3185 printk("\n");
3186 printk("==================================\n");
3187 printk("[ BUG: Nested lock was not taken ]\n");
3188 print_kernel_ident();
3189 printk("----------------------------------\n");
3190
3191 printk("%s/%d is trying to lock:\n", curr->comm, task_pid_nr(curr));
3192 print_lock(hlock);
3193
3194 printk("\nbut this task is not holding:\n");
3195 printk("%s\n", hlock->nest_lock->name);
3196
3197 printk("\nstack backtrace:\n");
3198 dump_stack();
3199
3200 printk("\nother info that might help us debug this:\n");
3201 lockdep_print_held_locks(curr);
3202
3203 printk("\nstack backtrace:\n");
3204 dump_stack();
3205
3206 return 0;
3207}
3208
f8319483 3209static int __lock_is_held(struct lockdep_map *lock, int read);
d0945950 3210
fbb9ce95
IM
3211/*
3212 * This gets called for every mutex_lock*()/spin_lock*() operation.
3213 * We maintain the dependency maps and validate the locking attempt:
3214 */
3215static int __lock_acquire(struct lockdep_map *lock, unsigned int subclass,
3216 int trylock, int read, int check, int hardirqs_off,
bb97a91e 3217 struct lockdep_map *nest_lock, unsigned long ip,
21199f27 3218 int references, int pin_count)
fbb9ce95
IM
3219{
3220 struct task_struct *curr = current;
d6d897ce 3221 struct lock_class *class = NULL;
fbb9ce95 3222 struct held_lock *hlock;
5f18ab5c 3223 unsigned int depth;
fbb9ce95 3224 int chain_head = 0;
bb97a91e 3225 int class_idx;
fbb9ce95
IM
3226 u64 chain_key;
3227
3228 if (unlikely(!debug_locks))
3229 return 0;
3230
0119fee4
PZ
3231 /*
3232 * Lockdep should run with IRQs disabled, otherwise we could
3233 * get an interrupt which would want to take locks, which would
3234 * end up in lockdep and have you got a head-ache already?
3235 */
fbb9ce95
IM
3236 if (DEBUG_LOCKS_WARN_ON(!irqs_disabled()))
3237 return 0;
3238
fb9edbe9
ON
3239 if (!prove_locking || lock->key == &__lockdep_no_validate__)
3240 check = 0;
1704f47b 3241
62016250
HM
3242 if (subclass < NR_LOCKDEP_CACHING_CLASSES)
3243 class = lock->class_cache[subclass];
d6d897ce 3244 /*
62016250 3245 * Not cached?
d6d897ce 3246 */
fbb9ce95 3247 if (unlikely(!class)) {
4dfbb9d8 3248 class = register_lock_class(lock, subclass, 0);
fbb9ce95
IM
3249 if (!class)
3250 return 0;
3251 }
bd6d29c2 3252 atomic_inc((atomic_t *)&class->ops);
fbb9ce95
IM
3253 if (very_verbose(class)) {
3254 printk("\nacquire class [%p] %s", class->key, class->name);
3255 if (class->name_version > 1)
f943fe0f
DV
3256 printk(KERN_CONT "#%d", class->name_version);
3257 printk(KERN_CONT "\n");
fbb9ce95
IM
3258 dump_stack();
3259 }
3260
3261 /*
3262 * Add the lock to the list of currently held locks.
3263 * (we dont increase the depth just yet, up until the
3264 * dependency checks are done)
3265 */
3266 depth = curr->lockdep_depth;
0119fee4
PZ
3267 /*
3268 * Ran out of static storage for our per-task lock stack again have we?
3269 */
fbb9ce95
IM
3270 if (DEBUG_LOCKS_WARN_ON(depth >= MAX_LOCK_DEPTH))
3271 return 0;
3272
bb97a91e
PZ
3273 class_idx = class - lock_classes + 1;
3274
3275 if (depth) {
3276 hlock = curr->held_locks + depth - 1;
3277 if (hlock->class_idx == class_idx && nest_lock) {
7fb4a2ce
PZ
3278 if (hlock->references) {
3279 /*
3280 * Check: unsigned int references:12, overflow.
3281 */
3282 if (DEBUG_LOCKS_WARN_ON(hlock->references == (1 << 12)-1))
3283 return 0;
3284
bb97a91e 3285 hlock->references++;
7fb4a2ce 3286 } else {
bb97a91e 3287 hlock->references = 2;
7fb4a2ce 3288 }
bb97a91e
PZ
3289
3290 return 1;
3291 }
3292 }
3293
fbb9ce95 3294 hlock = curr->held_locks + depth;
0119fee4
PZ
3295 /*
3296 * Plain impossible, we just registered it and checked it weren't no
3297 * NULL like.. I bet this mushroom I ate was good!
3298 */
f82b217e
DJ
3299 if (DEBUG_LOCKS_WARN_ON(!class))
3300 return 0;
bb97a91e 3301 hlock->class_idx = class_idx;
fbb9ce95
IM
3302 hlock->acquire_ip = ip;
3303 hlock->instance = lock;
7531e2f3 3304 hlock->nest_lock = nest_lock;
c2469756 3305 hlock->irq_context = task_irq_context(curr);
fbb9ce95
IM
3306 hlock->trylock = trylock;
3307 hlock->read = read;
3308 hlock->check = check;
6951b12a 3309 hlock->hardirqs_off = !!hardirqs_off;
bb97a91e 3310 hlock->references = references;
f20786ff
PZ
3311#ifdef CONFIG_LOCK_STAT
3312 hlock->waittime_stamp = 0;
3365e779 3313 hlock->holdtime_stamp = lockstat_clock();
f20786ff 3314#endif
21199f27 3315 hlock->pin_count = pin_count;
fbb9ce95 3316
fb9edbe9 3317 if (check && !mark_irqflags(curr, hlock))
8e18257d
PZ
3318 return 0;
3319
fbb9ce95 3320 /* mark it as used: */
4ff773bb 3321 if (!mark_lock(curr, hlock, LOCK_USED))
fbb9ce95 3322 return 0;
8e18257d 3323
fbb9ce95 3324 /*
17aacfb9 3325 * Calculate the chain hash: it's the combined hash of all the
fbb9ce95
IM
3326 * lock keys along the dependency chain. We save the hash value
3327 * at every step so that we can get the current hash easily
3328 * after unlock. The chain hash is then used to cache dependency
3329 * results.
3330 *
3331 * The 'key ID' is what is the most compact key value to drive
3332 * the hash, not class->key.
3333 */
0119fee4
PZ
3334 /*
3335 * Whoops, we did it again.. ran straight out of our static allocation.
3336 */
5f18ab5c 3337 if (DEBUG_LOCKS_WARN_ON(class_idx > MAX_LOCKDEP_KEYS))
fbb9ce95
IM
3338 return 0;
3339
3340 chain_key = curr->curr_chain_key;
3341 if (!depth) {
0119fee4
PZ
3342 /*
3343 * How can we have a chain hash when we ain't got no keys?!
3344 */
fbb9ce95
IM
3345 if (DEBUG_LOCKS_WARN_ON(chain_key != 0))
3346 return 0;
3347 chain_head = 1;
3348 }
3349
3350 hlock->prev_chain_key = chain_key;
8e18257d
PZ
3351 if (separate_irq_context(curr, hlock)) {
3352 chain_key = 0;
3353 chain_head = 1;
fbb9ce95 3354 }
5f18ab5c 3355 chain_key = iterate_chain_key(chain_key, class_idx);
fbb9ce95 3356
f8319483 3357 if (nest_lock && !__lock_is_held(nest_lock, -1))
d0945950
ML
3358 return print_lock_nested_lock_not_held(curr, hlock, ip);
3359
3aa416b0 3360 if (!validate_chain(curr, lock, hlock, chain_head, chain_key))
8e18257d 3361 return 0;
381a2292 3362
3aa416b0 3363 curr->curr_chain_key = chain_key;
fbb9ce95
IM
3364 curr->lockdep_depth++;
3365 check_chain_key(curr);
60e114d1
JP
3366#ifdef CONFIG_DEBUG_LOCKDEP
3367 if (unlikely(!debug_locks))
3368 return 0;
3369#endif
fbb9ce95
IM
3370 if (unlikely(curr->lockdep_depth >= MAX_LOCK_DEPTH)) {
3371 debug_locks_off();
2c522836
DJ
3372 print_lockdep_off("BUG: MAX_LOCK_DEPTH too low!");
3373 printk(KERN_DEBUG "depth: %i max: %lu!\n",
c0540606 3374 curr->lockdep_depth, MAX_LOCK_DEPTH);
c0540606
BG
3375
3376 lockdep_print_held_locks(current);
3377 debug_show_all_locks();
eedeeabd 3378 dump_stack();
c0540606 3379
fbb9ce95
IM
3380 return 0;
3381 }
381a2292 3382
fbb9ce95
IM
3383 if (unlikely(curr->lockdep_depth > max_lockdep_depth))
3384 max_lockdep_depth = curr->lockdep_depth;
3385
3386 return 1;
3387}
3388
3389static int
f86f7554 3390print_unlock_imbalance_bug(struct task_struct *curr, struct lockdep_map *lock,
fbb9ce95
IM
3391 unsigned long ip)
3392{
3393 if (!debug_locks_off())
3394 return 0;
3395 if (debug_locks_silent)
3396 return 0;
3397
b3fbab05
PM
3398 printk("\n");
3399 printk("=====================================\n");
3400 printk("[ BUG: bad unlock balance detected! ]\n");
fbdc4b9a 3401 print_kernel_ident();
b3fbab05 3402 printk("-------------------------------------\n");
fbb9ce95 3403 printk("%s/%d is trying to release lock (",
ba25f9dc 3404 curr->comm, task_pid_nr(curr));
fbb9ce95 3405 print_lockdep_cache(lock);
f943fe0f 3406 printk(KERN_CONT ") at:\n");
fbb9ce95
IM
3407 print_ip_sym(ip);
3408 printk("but there are no more locks to release!\n");
3409 printk("\nother info that might help us debug this:\n");
3410 lockdep_print_held_locks(curr);
3411
3412 printk("\nstack backtrace:\n");
3413 dump_stack();
3414
3415 return 0;
3416}
3417
bb97a91e
PZ
3418static int match_held_lock(struct held_lock *hlock, struct lockdep_map *lock)
3419{
3420 if (hlock->instance == lock)
3421 return 1;
3422
3423 if (hlock->references) {
62016250 3424 struct lock_class *class = lock->class_cache[0];
bb97a91e
PZ
3425
3426 if (!class)
3427 class = look_up_lock_class(lock, 0);
3428
80e0401e
PZ
3429 /*
3430 * If look_up_lock_class() failed to find a class, we're trying
3431 * to test if we hold a lock that has never yet been acquired.
3432 * Clearly if the lock hasn't been acquired _ever_, we're not
3433 * holding it either, so report failure.
3434 */
383776fa 3435 if (IS_ERR_OR_NULL(class))
bb97a91e
PZ
3436 return 0;
3437
0119fee4
PZ
3438 /*
3439 * References, but not a lock we're actually ref-counting?
3440 * State got messed up, follow the sites that change ->references
3441 * and try to make sense of it.
3442 */
bb97a91e
PZ
3443 if (DEBUG_LOCKS_WARN_ON(!hlock->nest_lock))
3444 return 0;
3445
3446 if (hlock->class_idx == class - lock_classes + 1)
3447 return 1;
3448 }
3449
3450 return 0;
3451}
3452
41c2c5b8
O
3453/* @depth must not be zero */
3454static struct held_lock *find_held_lock(struct task_struct *curr,
3455 struct lockdep_map *lock,
3456 unsigned int depth, int *idx)
3457{
3458 struct held_lock *ret, *hlock, *prev_hlock;
3459 int i;
3460
3461 i = depth - 1;
3462 hlock = curr->held_locks + i;
3463 ret = hlock;
3464 if (match_held_lock(hlock, lock))
3465 goto out;
3466
3467 ret = NULL;
3468 for (i--, prev_hlock = hlock--;
3469 i >= 0;
3470 i--, prev_hlock = hlock--) {
3471 /*
3472 * We must not cross into another context:
3473 */
3474 if (prev_hlock->irq_context != hlock->irq_context) {
3475 ret = NULL;
3476 break;
3477 }
3478 if (match_held_lock(hlock, lock)) {
3479 ret = hlock;
3480 break;
3481 }
3482 }
3483
3484out:
3485 *idx = i;
3486 return ret;
3487}
3488
e969970b
O
3489static int reacquire_held_locks(struct task_struct *curr, unsigned int depth,
3490 int idx)
3491{
3492 struct held_lock *hlock;
3493
3494 for (hlock = curr->held_locks + idx; idx < depth; idx++, hlock++) {
3495 if (!__lock_acquire(hlock->instance,
3496 hlock_class(hlock)->subclass,
3497 hlock->trylock,
3498 hlock->read, hlock->check,
3499 hlock->hardirqs_off,
3500 hlock->nest_lock, hlock->acquire_ip,
3501 hlock->references, hlock->pin_count))
3502 return 1;
3503 }
3504 return 0;
3505}
3506
64aa348e 3507static int
00ef9f73
PZ
3508__lock_set_class(struct lockdep_map *lock, const char *name,
3509 struct lock_class_key *key, unsigned int subclass,
3510 unsigned long ip)
64aa348e
PZ
3511{
3512 struct task_struct *curr = current;
41c2c5b8 3513 struct held_lock *hlock;
64aa348e
PZ
3514 struct lock_class *class;
3515 unsigned int depth;
3516 int i;
3517
3518 depth = curr->lockdep_depth;
0119fee4
PZ
3519 /*
3520 * This function is about (re)setting the class of a held lock,
3521 * yet we're not actually holding any locks. Naughty user!
3522 */
64aa348e
PZ
3523 if (DEBUG_LOCKS_WARN_ON(!depth))
3524 return 0;
3525
41c2c5b8
O
3526 hlock = find_held_lock(curr, lock, depth, &i);
3527 if (!hlock)
3528 return print_unlock_imbalance_bug(curr, lock, ip);
64aa348e 3529
00ef9f73 3530 lockdep_init_map(lock, name, key, 0);
64aa348e 3531 class = register_lock_class(lock, subclass, 0);
f82b217e 3532 hlock->class_idx = class - lock_classes + 1;
64aa348e
PZ
3533
3534 curr->lockdep_depth = i;
3535 curr->curr_chain_key = hlock->prev_chain_key;
3536
e969970b
O
3537 if (reacquire_held_locks(curr, depth, i))
3538 return 0;
64aa348e 3539
0119fee4
PZ
3540 /*
3541 * I took it apart and put it back together again, except now I have
3542 * these 'spare' parts.. where shall I put them.
3543 */
64aa348e
PZ
3544 if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth))
3545 return 0;
3546 return 1;
3547}
3548
6419c4af
O
3549static int __lock_downgrade(struct lockdep_map *lock, unsigned long ip)
3550{
3551 struct task_struct *curr = current;
3552 struct held_lock *hlock;
3553 unsigned int depth;
3554 int i;
3555
3556 depth = curr->lockdep_depth;
3557 /*
3558 * This function is about (re)setting the class of a held lock,
3559 * yet we're not actually holding any locks. Naughty user!
3560 */
3561 if (DEBUG_LOCKS_WARN_ON(!depth))
3562 return 0;
3563
3564 hlock = find_held_lock(curr, lock, depth, &i);
3565 if (!hlock)
3566 return print_unlock_imbalance_bug(curr, lock, ip);
3567
3568 curr->lockdep_depth = i;
3569 curr->curr_chain_key = hlock->prev_chain_key;
3570
3571 WARN(hlock->read, "downgrading a read lock");
3572 hlock->read = 1;
3573 hlock->acquire_ip = ip;
3574
3575 if (reacquire_held_locks(curr, depth, i))
3576 return 0;
3577
3578 /*
3579 * I took it apart and put it back together again, except now I have
3580 * these 'spare' parts.. where shall I put them.
3581 */
3582 if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth))
3583 return 0;
3584 return 1;
3585}
3586
fbb9ce95 3587/*
e0f56fd7
PZ
3588 * Remove the lock to the list of currently held locks - this gets
3589 * called on mutex_unlock()/spin_unlock*() (or on a failed
3590 * mutex_lock_interruptible()).
3591 *
3592 * @nested is an hysterical artifact, needs a tree wide cleanup.
fbb9ce95
IM
3593 */
3594static int
e0f56fd7 3595__lock_release(struct lockdep_map *lock, int nested, unsigned long ip)
fbb9ce95 3596{
e0f56fd7 3597 struct task_struct *curr = current;
41c2c5b8 3598 struct held_lock *hlock;
fbb9ce95
IM
3599 unsigned int depth;
3600 int i;
3601
e0f56fd7
PZ
3602 if (unlikely(!debug_locks))
3603 return 0;
3604
fbb9ce95 3605 depth = curr->lockdep_depth;
0119fee4
PZ
3606 /*
3607 * So we're all set to release this lock.. wait what lock? We don't
3608 * own any locks, you've been drinking again?
3609 */
e0f56fd7
PZ
3610 if (DEBUG_LOCKS_WARN_ON(depth <= 0))
3611 return print_unlock_imbalance_bug(curr, lock, ip);
fbb9ce95 3612
e0f56fd7
PZ
3613 /*
3614 * Check whether the lock exists in the current stack
3615 * of held locks:
3616 */
41c2c5b8
O
3617 hlock = find_held_lock(curr, lock, depth, &i);
3618 if (!hlock)
3619 return print_unlock_imbalance_bug(curr, lock, ip);
fbb9ce95 3620
bb97a91e
PZ
3621 if (hlock->instance == lock)
3622 lock_release_holdtime(hlock);
3623
a24fc60d
PZ
3624 WARN(hlock->pin_count, "releasing a pinned lock\n");
3625
bb97a91e
PZ
3626 if (hlock->references) {
3627 hlock->references--;
3628 if (hlock->references) {
3629 /*
3630 * We had, and after removing one, still have
3631 * references, the current lock stack is still
3632 * valid. We're done!
3633 */
3634 return 1;
3635 }
3636 }
f20786ff 3637
fbb9ce95
IM
3638 /*
3639 * We have the right lock to unlock, 'hlock' points to it.
3640 * Now we remove it from the stack, and add back the other
3641 * entries (if any), recalculating the hash along the way:
3642 */
bb97a91e 3643
fbb9ce95
IM
3644 curr->lockdep_depth = i;
3645 curr->curr_chain_key = hlock->prev_chain_key;
3646
e969970b
O
3647 if (reacquire_held_locks(curr, depth, i + 1))
3648 return 0;
fbb9ce95 3649
0119fee4
PZ
3650 /*
3651 * We had N bottles of beer on the wall, we drank one, but now
3652 * there's not N-1 bottles of beer left on the wall...
3653 */
fbb9ce95
IM
3654 if (DEBUG_LOCKS_WARN_ON(curr->lockdep_depth != depth - 1))
3655 return 0;
f20786ff 3656
fbb9ce95
IM
3657 return 1;
3658}
3659
f8319483 3660static int __lock_is_held(struct lockdep_map *lock, int read)
fbb9ce95 3661{
f607c668
PZ
3662 struct task_struct *curr = current;
3663 int i;
fbb9ce95 3664
f607c668 3665 for (i = 0; i < curr->lockdep_depth; i++) {
bb97a91e 3666 struct held_lock *hlock = curr->held_locks + i;
fbb9ce95 3667
f8319483
PZ
3668 if (match_held_lock(hlock, lock)) {
3669 if (read == -1 || hlock->read == read)
3670 return 1;
3671
3672 return 0;
3673 }
f607c668 3674 }
f20786ff 3675
f607c668 3676 return 0;
fbb9ce95
IM
3677}
3678
e7904a28
PZ
3679static struct pin_cookie __lock_pin_lock(struct lockdep_map *lock)
3680{
3681 struct pin_cookie cookie = NIL_COOKIE;
3682 struct task_struct *curr = current;
3683 int i;
3684
3685 if (unlikely(!debug_locks))
3686 return cookie;
3687
3688 for (i = 0; i < curr->lockdep_depth; i++) {
3689 struct held_lock *hlock = curr->held_locks + i;
3690
3691 if (match_held_lock(hlock, lock)) {
3692 /*
3693 * Grab 16bits of randomness; this is sufficient to not
3694 * be guessable and still allows some pin nesting in
3695 * our u32 pin_count.
3696 */
3697 cookie.val = 1 + (prandom_u32() >> 16);
3698 hlock->pin_count += cookie.val;
3699 return cookie;
3700 }
3701 }
3702
3703 WARN(1, "pinning an unheld lock\n");
3704 return cookie;
3705}
3706
3707static void __lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
fbb9ce95
IM
3708{
3709 struct task_struct *curr = current;
a24fc60d 3710 int i;
fbb9ce95 3711
a24fc60d 3712 if (unlikely(!debug_locks))
fbb9ce95
IM
3713 return;
3714
a24fc60d
PZ
3715 for (i = 0; i < curr->lockdep_depth; i++) {
3716 struct held_lock *hlock = curr->held_locks + i;
3717
3718 if (match_held_lock(hlock, lock)) {
e7904a28 3719 hlock->pin_count += cookie.val;
fbb9ce95 3720 return;
a24fc60d 3721 }
fbb9ce95
IM
3722 }
3723
a24fc60d 3724 WARN(1, "pinning an unheld lock\n");
fbb9ce95
IM
3725}
3726
e7904a28 3727static void __lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
f607c668
PZ
3728{
3729 struct task_struct *curr = current;
3730 int i;
3731
a24fc60d
PZ
3732 if (unlikely(!debug_locks))
3733 return;
3734
f607c668 3735 for (i = 0; i < curr->lockdep_depth; i++) {
bb97a91e
PZ
3736 struct held_lock *hlock = curr->held_locks + i;
3737
a24fc60d
PZ
3738 if (match_held_lock(hlock, lock)) {
3739 if (WARN(!hlock->pin_count, "unpinning an unpinned lock\n"))
3740 return;
3741
e7904a28
PZ
3742 hlock->pin_count -= cookie.val;
3743
3744 if (WARN((int)hlock->pin_count < 0, "pin count corrupted\n"))
3745 hlock->pin_count = 0;
3746
a24fc60d
PZ
3747 return;
3748 }
f607c668
PZ
3749 }
3750
a24fc60d 3751 WARN(1, "unpinning an unheld lock\n");
f607c668
PZ
3752}
3753
fbb9ce95
IM
3754/*
3755 * Check whether we follow the irq-flags state precisely:
3756 */
1d09daa5 3757static void check_flags(unsigned long flags)
fbb9ce95 3758{
992860e9
IM
3759#if defined(CONFIG_PROVE_LOCKING) && defined(CONFIG_DEBUG_LOCKDEP) && \
3760 defined(CONFIG_TRACE_IRQFLAGS)
fbb9ce95
IM
3761 if (!debug_locks)
3762 return;
3763
5f9fa8a6
IM
3764 if (irqs_disabled_flags(flags)) {
3765 if (DEBUG_LOCKS_WARN_ON(current->hardirqs_enabled)) {
3766 printk("possible reason: unannotated irqs-off.\n");
3767 }
3768 } else {
3769 if (DEBUG_LOCKS_WARN_ON(!current->hardirqs_enabled)) {
3770 printk("possible reason: unannotated irqs-on.\n");
3771 }
3772 }
fbb9ce95
IM
3773
3774 /*
3775 * We dont accurately track softirq state in e.g.
3776 * hardirq contexts (such as on 4KSTACKS), so only
3777 * check if not in hardirq contexts:
3778 */
3779 if (!hardirq_count()) {
0119fee4
PZ
3780 if (softirq_count()) {
3781 /* like the above, but with softirqs */
fbb9ce95 3782 DEBUG_LOCKS_WARN_ON(current->softirqs_enabled);
0119fee4
PZ
3783 } else {
3784 /* lick the above, does it taste good? */
fbb9ce95 3785 DEBUG_LOCKS_WARN_ON(!current->softirqs_enabled);
0119fee4 3786 }
fbb9ce95
IM
3787 }
3788
3789 if (!debug_locks)
3790 print_irqtrace_events(current);
3791#endif
3792}
3793
00ef9f73
PZ
3794void lock_set_class(struct lockdep_map *lock, const char *name,
3795 struct lock_class_key *key, unsigned int subclass,
3796 unsigned long ip)
64aa348e
PZ
3797{
3798 unsigned long flags;
3799
3800 if (unlikely(current->lockdep_recursion))
3801 return;
3802
3803 raw_local_irq_save(flags);
3804 current->lockdep_recursion = 1;
3805 check_flags(flags);
00ef9f73 3806 if (__lock_set_class(lock, name, key, subclass, ip))
64aa348e
PZ
3807 check_chain_key(current);
3808 current->lockdep_recursion = 0;
3809 raw_local_irq_restore(flags);
3810}
00ef9f73 3811EXPORT_SYMBOL_GPL(lock_set_class);
64aa348e 3812
6419c4af
O
3813void lock_downgrade(struct lockdep_map *lock, unsigned long ip)
3814{
3815 unsigned long flags;
3816
3817 if (unlikely(current->lockdep_recursion))
3818 return;
3819
3820 raw_local_irq_save(flags);
3821 current->lockdep_recursion = 1;
3822 check_flags(flags);
3823 if (__lock_downgrade(lock, ip))
3824 check_chain_key(current);
3825 current->lockdep_recursion = 0;
3826 raw_local_irq_restore(flags);
3827}
3828EXPORT_SYMBOL_GPL(lock_downgrade);
3829
fbb9ce95
IM
3830/*
3831 * We are not always called with irqs disabled - do that here,
3832 * and also avoid lockdep recursion:
3833 */
1d09daa5 3834void lock_acquire(struct lockdep_map *lock, unsigned int subclass,
7531e2f3
PZ
3835 int trylock, int read, int check,
3836 struct lockdep_map *nest_lock, unsigned long ip)
fbb9ce95
IM
3837{
3838 unsigned long flags;
3839
3840 if (unlikely(current->lockdep_recursion))
3841 return;
3842
3843 raw_local_irq_save(flags);
3844 check_flags(flags);
3845
3846 current->lockdep_recursion = 1;
db2c4c77 3847 trace_lock_acquire(lock, subclass, trylock, read, check, nest_lock, ip);
fbb9ce95 3848 __lock_acquire(lock, subclass, trylock, read, check,
21199f27 3849 irqs_disabled_flags(flags), nest_lock, ip, 0, 0);
fbb9ce95
IM
3850 current->lockdep_recursion = 0;
3851 raw_local_irq_restore(flags);
3852}
fbb9ce95
IM
3853EXPORT_SYMBOL_GPL(lock_acquire);
3854
1d09daa5 3855void lock_release(struct lockdep_map *lock, int nested,
0764d23c 3856 unsigned long ip)
fbb9ce95
IM
3857{
3858 unsigned long flags;
3859
3860 if (unlikely(current->lockdep_recursion))
3861 return;
3862
3863 raw_local_irq_save(flags);
3864 check_flags(flags);
3865 current->lockdep_recursion = 1;
93135439 3866 trace_lock_release(lock, ip);
e0f56fd7
PZ
3867 if (__lock_release(lock, nested, ip))
3868 check_chain_key(current);
fbb9ce95
IM
3869 current->lockdep_recursion = 0;
3870 raw_local_irq_restore(flags);
3871}
fbb9ce95
IM
3872EXPORT_SYMBOL_GPL(lock_release);
3873
f8319483 3874int lock_is_held_type(struct lockdep_map *lock, int read)
f607c668
PZ
3875{
3876 unsigned long flags;
3877 int ret = 0;
3878
3879 if (unlikely(current->lockdep_recursion))
f2513cde 3880 return 1; /* avoid false negative lockdep_assert_held() */
f607c668
PZ
3881
3882 raw_local_irq_save(flags);
3883 check_flags(flags);
3884
3885 current->lockdep_recursion = 1;
f8319483 3886 ret = __lock_is_held(lock, read);
f607c668
PZ
3887 current->lockdep_recursion = 0;
3888 raw_local_irq_restore(flags);
3889
3890 return ret;
3891}
f8319483 3892EXPORT_SYMBOL_GPL(lock_is_held_type);
f607c668 3893
e7904a28 3894struct pin_cookie lock_pin_lock(struct lockdep_map *lock)
a24fc60d 3895{
e7904a28 3896 struct pin_cookie cookie = NIL_COOKIE;
a24fc60d
PZ
3897 unsigned long flags;
3898
3899 if (unlikely(current->lockdep_recursion))
e7904a28 3900 return cookie;
a24fc60d
PZ
3901
3902 raw_local_irq_save(flags);
3903 check_flags(flags);
3904
3905 current->lockdep_recursion = 1;
e7904a28 3906 cookie = __lock_pin_lock(lock);
a24fc60d
PZ
3907 current->lockdep_recursion = 0;
3908 raw_local_irq_restore(flags);
e7904a28
PZ
3909
3910 return cookie;
a24fc60d
PZ
3911}
3912EXPORT_SYMBOL_GPL(lock_pin_lock);
3913
e7904a28
PZ
3914void lock_repin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
3915{
3916 unsigned long flags;
3917
3918 if (unlikely(current->lockdep_recursion))
3919 return;
3920
3921 raw_local_irq_save(flags);
3922 check_flags(flags);
3923
3924 current->lockdep_recursion = 1;
3925 __lock_repin_lock(lock, cookie);
3926 current->lockdep_recursion = 0;
3927 raw_local_irq_restore(flags);
3928}
3929EXPORT_SYMBOL_GPL(lock_repin_lock);
3930
3931void lock_unpin_lock(struct lockdep_map *lock, struct pin_cookie cookie)
a24fc60d
PZ
3932{
3933 unsigned long flags;
3934
3935 if (unlikely(current->lockdep_recursion))
3936 return;
3937
3938 raw_local_irq_save(flags);
3939 check_flags(flags);
3940
3941 current->lockdep_recursion = 1;
e7904a28 3942 __lock_unpin_lock(lock, cookie);
a24fc60d
PZ
3943 current->lockdep_recursion = 0;
3944 raw_local_irq_restore(flags);
3945}
3946EXPORT_SYMBOL_GPL(lock_unpin_lock);
3947
cf40bd16
NP
3948void lockdep_set_current_reclaim_state(gfp_t gfp_mask)
3949{
3950 current->lockdep_reclaim_gfp = gfp_mask;
3951}
3952
3953void lockdep_clear_current_reclaim_state(void)
3954{
3955 current->lockdep_reclaim_gfp = 0;
3956}
3957
f20786ff
PZ
3958#ifdef CONFIG_LOCK_STAT
3959static int
3960print_lock_contention_bug(struct task_struct *curr, struct lockdep_map *lock,
3961 unsigned long ip)
3962{
3963 if (!debug_locks_off())
3964 return 0;
3965 if (debug_locks_silent)
3966 return 0;
3967
b3fbab05
PM
3968 printk("\n");
3969 printk("=================================\n");
3970 printk("[ BUG: bad contention detected! ]\n");
fbdc4b9a 3971 print_kernel_ident();
b3fbab05 3972 printk("---------------------------------\n");
f20786ff 3973 printk("%s/%d is trying to contend lock (",
ba25f9dc 3974 curr->comm, task_pid_nr(curr));
f20786ff 3975 print_lockdep_cache(lock);
f943fe0f 3976 printk(KERN_CONT ") at:\n");
f20786ff
PZ
3977 print_ip_sym(ip);
3978 printk("but there are no locks held!\n");
3979 printk("\nother info that might help us debug this:\n");
3980 lockdep_print_held_locks(curr);
3981
3982 printk("\nstack backtrace:\n");
3983 dump_stack();
3984
3985 return 0;
3986}
3987
3988static void
3989__lock_contended(struct lockdep_map *lock, unsigned long ip)
3990{
3991 struct task_struct *curr = current;
41c2c5b8 3992 struct held_lock *hlock;
f20786ff
PZ
3993 struct lock_class_stats *stats;
3994 unsigned int depth;
c7e78cff 3995 int i, contention_point, contending_point;
f20786ff
PZ
3996
3997 depth = curr->lockdep_depth;
0119fee4
PZ
3998 /*
3999 * Whee, we contended on this lock, except it seems we're not
4000 * actually trying to acquire anything much at all..
4001 */
f20786ff
PZ
4002 if (DEBUG_LOCKS_WARN_ON(!depth))
4003 return;
4004
41c2c5b8
O
4005 hlock = find_held_lock(curr, lock, depth, &i);
4006 if (!hlock) {
4007 print_lock_contention_bug(curr, lock, ip);
4008 return;
f20786ff 4009 }
f20786ff 4010
bb97a91e
PZ
4011 if (hlock->instance != lock)
4012 return;
4013
3365e779 4014 hlock->waittime_stamp = lockstat_clock();
f20786ff 4015
c7e78cff
PZ
4016 contention_point = lock_point(hlock_class(hlock)->contention_point, ip);
4017 contending_point = lock_point(hlock_class(hlock)->contending_point,
4018 lock->ip);
f20786ff 4019
f82b217e 4020 stats = get_lock_stats(hlock_class(hlock));
c7e78cff
PZ
4021 if (contention_point < LOCKSTAT_POINTS)
4022 stats->contention_point[contention_point]++;
4023 if (contending_point < LOCKSTAT_POINTS)
4024 stats->contending_point[contending_point]++;
96645678
PZ
4025 if (lock->cpu != smp_processor_id())
4026 stats->bounces[bounce_contended + !!hlock->read]++;
f20786ff
PZ
4027 put_lock_stats(stats);
4028}
4029
4030static void
c7e78cff 4031__lock_acquired(struct lockdep_map *lock, unsigned long ip)
f20786ff
PZ
4032{
4033 struct task_struct *curr = current;
41c2c5b8 4034 struct held_lock *hlock;
f20786ff
PZ
4035 struct lock_class_stats *stats;
4036 unsigned int depth;
3365e779 4037 u64 now, waittime = 0;
96645678 4038 int i, cpu;
f20786ff
PZ
4039
4040 depth = curr->lockdep_depth;
0119fee4
PZ
4041 /*
4042 * Yay, we acquired ownership of this lock we didn't try to
4043 * acquire, how the heck did that happen?
4044 */
f20786ff
PZ
4045 if (DEBUG_LOCKS_WARN_ON(!depth))
4046 return;
4047
41c2c5b8
O
4048 hlock = find_held_lock(curr, lock, depth, &i);
4049 if (!hlock) {
4050 print_lock_contention_bug(curr, lock, _RET_IP_);
4051 return;
f20786ff 4052 }
f20786ff 4053
bb97a91e
PZ
4054 if (hlock->instance != lock)
4055 return;
4056
96645678
PZ
4057 cpu = smp_processor_id();
4058 if (hlock->waittime_stamp) {
3365e779 4059 now = lockstat_clock();
96645678
PZ
4060 waittime = now - hlock->waittime_stamp;
4061 hlock->holdtime_stamp = now;
4062 }
f20786ff 4063
883a2a31 4064 trace_lock_acquired(lock, ip);
2062501a 4065
f82b217e 4066 stats = get_lock_stats(hlock_class(hlock));
96645678
PZ
4067 if (waittime) {
4068 if (hlock->read)
4069 lock_time_inc(&stats->read_waittime, waittime);
4070 else
4071 lock_time_inc(&stats->write_waittime, waittime);
4072 }
4073 if (lock->cpu != cpu)
4074 stats->bounces[bounce_acquired + !!hlock->read]++;
f20786ff 4075 put_lock_stats(stats);
96645678
PZ
4076
4077 lock->cpu = cpu;
c7e78cff 4078 lock->ip = ip;
f20786ff
PZ
4079}
4080
4081void lock_contended(struct lockdep_map *lock, unsigned long ip)
4082{
4083 unsigned long flags;
4084
4085 if (unlikely(!lock_stat))
4086 return;
4087
4088 if (unlikely(current->lockdep_recursion))
4089 return;
4090
4091 raw_local_irq_save(flags);
4092 check_flags(flags);
4093 current->lockdep_recursion = 1;
db2c4c77 4094 trace_lock_contended(lock, ip);
f20786ff
PZ
4095 __lock_contended(lock, ip);
4096 current->lockdep_recursion = 0;
4097 raw_local_irq_restore(flags);
4098}
4099EXPORT_SYMBOL_GPL(lock_contended);
4100
c7e78cff 4101void lock_acquired(struct lockdep_map *lock, unsigned long ip)
f20786ff
PZ
4102{
4103 unsigned long flags;
4104
4105 if (unlikely(!lock_stat))
4106 return;
4107
4108 if (unlikely(current->lockdep_recursion))
4109 return;
4110
4111 raw_local_irq_save(flags);
4112 check_flags(flags);
4113 current->lockdep_recursion = 1;
c7e78cff 4114 __lock_acquired(lock, ip);
f20786ff
PZ
4115 current->lockdep_recursion = 0;
4116 raw_local_irq_restore(flags);
4117}
4118EXPORT_SYMBOL_GPL(lock_acquired);
4119#endif
4120
fbb9ce95
IM
4121/*
4122 * Used by the testsuite, sanitize the validator state
4123 * after a simulated failure:
4124 */
4125
4126void lockdep_reset(void)
4127{
4128 unsigned long flags;
23d95a03 4129 int i;
fbb9ce95
IM
4130
4131 raw_local_irq_save(flags);
4132 current->curr_chain_key = 0;
4133 current->lockdep_depth = 0;
4134 current->lockdep_recursion = 0;
4135 memset(current->held_locks, 0, MAX_LOCK_DEPTH*sizeof(struct held_lock));
4136 nr_hardirq_chains = 0;
4137 nr_softirq_chains = 0;
4138 nr_process_chains = 0;
4139 debug_locks = 1;
23d95a03 4140 for (i = 0; i < CHAINHASH_SIZE; i++)
a63f38cc 4141 INIT_HLIST_HEAD(chainhash_table + i);
fbb9ce95
IM
4142 raw_local_irq_restore(flags);
4143}
4144
4145static void zap_class(struct lock_class *class)
4146{
4147 int i;
4148
4149 /*
4150 * Remove all dependencies this lock is
4151 * involved in:
4152 */
4153 for (i = 0; i < nr_list_entries; i++) {
4154 if (list_entries[i].class == class)
4155 list_del_rcu(&list_entries[i].entry);
4156 }
4157 /*
4158 * Unhash the class and remove it from the all_lock_classes list:
4159 */
a63f38cc 4160 hlist_del_rcu(&class->hash_entry);
fbb9ce95
IM
4161 list_del_rcu(&class->lock_entry);
4162
cee34d88
PZ
4163 RCU_INIT_POINTER(class->key, NULL);
4164 RCU_INIT_POINTER(class->name, NULL);
fbb9ce95
IM
4165}
4166
fabe874a 4167static inline int within(const void *addr, void *start, unsigned long size)
fbb9ce95
IM
4168{
4169 return addr >= start && addr < start + size;
4170}
4171
35a9393c
PZ
4172/*
4173 * Used in module.c to remove lock classes from memory that is going to be
4174 * freed; and possibly re-used by other modules.
4175 *
4176 * We will have had one sync_sched() before getting here, so we're guaranteed
4177 * nobody will look up these exact classes -- they're properly dead but still
4178 * allocated.
4179 */
fbb9ce95
IM
4180void lockdep_free_key_range(void *start, unsigned long size)
4181{
35a9393c 4182 struct lock_class *class;
a63f38cc 4183 struct hlist_head *head;
fbb9ce95
IM
4184 unsigned long flags;
4185 int i;
5a26db5b 4186 int locked;
fbb9ce95
IM
4187
4188 raw_local_irq_save(flags);
5a26db5b 4189 locked = graph_lock();
fbb9ce95
IM
4190
4191 /*
4192 * Unhash all classes that were created by this module:
4193 */
4194 for (i = 0; i < CLASSHASH_SIZE; i++) {
4195 head = classhash_table + i;
a63f38cc 4196 hlist_for_each_entry_rcu(class, head, hash_entry) {
fbb9ce95
IM
4197 if (within(class->key, start, size))
4198 zap_class(class);
fabe874a
AV
4199 else if (within(class->name, start, size))
4200 zap_class(class);
4201 }
fbb9ce95
IM
4202 }
4203
5a26db5b
NP
4204 if (locked)
4205 graph_unlock();
fbb9ce95 4206 raw_local_irq_restore(flags);
35a9393c
PZ
4207
4208 /*
4209 * Wait for any possible iterators from look_up_lock_class() to pass
4210 * before continuing to free the memory they refer to.
4211 *
4212 * sync_sched() is sufficient because the read-side is IRQ disable.
4213 */
4214 synchronize_sched();
4215
4216 /*
4217 * XXX at this point we could return the resources to the pool;
4218 * instead we leak them. We would need to change to bitmap allocators
4219 * instead of the linear allocators we have now.
4220 */
fbb9ce95
IM
4221}
4222
4223void lockdep_reset_lock(struct lockdep_map *lock)
4224{
35a9393c 4225 struct lock_class *class;
a63f38cc 4226 struct hlist_head *head;
fbb9ce95
IM
4227 unsigned long flags;
4228 int i, j;
5a26db5b 4229 int locked;
fbb9ce95
IM
4230
4231 raw_local_irq_save(flags);
fbb9ce95
IM
4232
4233 /*
d6d897ce
IM
4234 * Remove all classes this lock might have:
4235 */
4236 for (j = 0; j < MAX_LOCKDEP_SUBCLASSES; j++) {
4237 /*
4238 * If the class exists we look it up and zap it:
4239 */
4240 class = look_up_lock_class(lock, j);
383776fa 4241 if (!IS_ERR_OR_NULL(class))
d6d897ce
IM
4242 zap_class(class);
4243 }
4244 /*
4245 * Debug check: in the end all mapped classes should
4246 * be gone.
fbb9ce95 4247 */
5a26db5b 4248 locked = graph_lock();
fbb9ce95
IM
4249 for (i = 0; i < CLASSHASH_SIZE; i++) {
4250 head = classhash_table + i;
a63f38cc 4251 hlist_for_each_entry_rcu(class, head, hash_entry) {
62016250
HM
4252 int match = 0;
4253
4254 for (j = 0; j < NR_LOCKDEP_CACHING_CLASSES; j++)
4255 match |= class == lock->class_cache[j];
4256
4257 if (unlikely(match)) {
0119fee4
PZ
4258 if (debug_locks_off_graph_unlock()) {
4259 /*
4260 * We all just reset everything, how did it match?
4261 */
74c383f1 4262 WARN_ON(1);
0119fee4 4263 }
d6d897ce 4264 goto out_restore;
fbb9ce95
IM
4265 }
4266 }
4267 }
5a26db5b
NP
4268 if (locked)
4269 graph_unlock();
d6d897ce
IM
4270
4271out_restore:
fbb9ce95
IM
4272 raw_local_irq_restore(flags);
4273}
4274
fbb9ce95
IM
4275void __init lockdep_info(void)
4276{
4277 printk("Lock dependency validator: Copyright (c) 2006 Red Hat, Inc., Ingo Molnar\n");
4278
b0788caf 4279 printk("... MAX_LOCKDEP_SUBCLASSES: %lu\n", MAX_LOCKDEP_SUBCLASSES);
fbb9ce95
IM
4280 printk("... MAX_LOCK_DEPTH: %lu\n", MAX_LOCK_DEPTH);
4281 printk("... MAX_LOCKDEP_KEYS: %lu\n", MAX_LOCKDEP_KEYS);
b0788caf 4282 printk("... CLASSHASH_SIZE: %lu\n", CLASSHASH_SIZE);
fbb9ce95
IM
4283 printk("... MAX_LOCKDEP_ENTRIES: %lu\n", MAX_LOCKDEP_ENTRIES);
4284 printk("... MAX_LOCKDEP_CHAINS: %lu\n", MAX_LOCKDEP_CHAINS);
4285 printk("... CHAINHASH_SIZE: %lu\n", CHAINHASH_SIZE);
4286
4287 printk(" memory used by lock dependency info: %lu kB\n",
4288 (sizeof(struct lock_class) * MAX_LOCKDEP_KEYS +
4289 sizeof(struct list_head) * CLASSHASH_SIZE +
4290 sizeof(struct lock_list) * MAX_LOCKDEP_ENTRIES +
4291 sizeof(struct lock_chain) * MAX_LOCKDEP_CHAINS +
90629209 4292 sizeof(struct list_head) * CHAINHASH_SIZE
4dd861d6 4293#ifdef CONFIG_PROVE_LOCKING
e351b660 4294 + sizeof(struct circular_queue)
4dd861d6 4295#endif
90629209 4296 ) / 1024
4dd861d6 4297 );
fbb9ce95
IM
4298
4299 printk(" per task-struct memory footprint: %lu bytes\n",
4300 sizeof(struct held_lock) * MAX_LOCK_DEPTH);
fbb9ce95
IM
4301}
4302
fbb9ce95
IM
4303static void
4304print_freed_lock_bug(struct task_struct *curr, const void *mem_from,
55794a41 4305 const void *mem_to, struct held_lock *hlock)
fbb9ce95
IM
4306{
4307 if (!debug_locks_off())
4308 return;
4309 if (debug_locks_silent)
4310 return;
4311
b3fbab05
PM
4312 printk("\n");
4313 printk("=========================\n");
4314 printk("[ BUG: held lock freed! ]\n");
fbdc4b9a 4315 print_kernel_ident();
b3fbab05 4316 printk("-------------------------\n");
fbb9ce95 4317 printk("%s/%d is freeing memory %p-%p, with a lock still held there!\n",
ba25f9dc 4318 curr->comm, task_pid_nr(curr), mem_from, mem_to-1);
55794a41 4319 print_lock(hlock);
fbb9ce95
IM
4320 lockdep_print_held_locks(curr);
4321
4322 printk("\nstack backtrace:\n");
4323 dump_stack();
4324}
4325
54561783
ON
4326static inline int not_in_range(const void* mem_from, unsigned long mem_len,
4327 const void* lock_from, unsigned long lock_len)
4328{
4329 return lock_from + lock_len <= mem_from ||
4330 mem_from + mem_len <= lock_from;
4331}
4332
fbb9ce95
IM
4333/*
4334 * Called when kernel memory is freed (or unmapped), or if a lock
4335 * is destroyed or reinitialized - this code checks whether there is
4336 * any held lock in the memory range of <from> to <to>:
4337 */
4338void debug_check_no_locks_freed(const void *mem_from, unsigned long mem_len)
4339{
fbb9ce95
IM
4340 struct task_struct *curr = current;
4341 struct held_lock *hlock;
4342 unsigned long flags;
4343 int i;
4344
4345 if (unlikely(!debug_locks))
4346 return;
4347
4348 local_irq_save(flags);
4349 for (i = 0; i < curr->lockdep_depth; i++) {
4350 hlock = curr->held_locks + i;
4351
54561783
ON
4352 if (not_in_range(mem_from, mem_len, hlock->instance,
4353 sizeof(*hlock->instance)))
fbb9ce95
IM
4354 continue;
4355
54561783 4356 print_freed_lock_bug(curr, mem_from, mem_from + mem_len, hlock);
fbb9ce95
IM
4357 break;
4358 }
4359 local_irq_restore(flags);
4360}
ed07536e 4361EXPORT_SYMBOL_GPL(debug_check_no_locks_freed);
fbb9ce95 4362
1b1d2fb4 4363static void print_held_locks_bug(void)
fbb9ce95
IM
4364{
4365 if (!debug_locks_off())
4366 return;
4367 if (debug_locks_silent)
4368 return;
4369
b3fbab05
PM
4370 printk("\n");
4371 printk("=====================================\n");
1b1d2fb4
CC
4372 printk("[ BUG: %s/%d still has locks held! ]\n",
4373 current->comm, task_pid_nr(current));
fbdc4b9a 4374 print_kernel_ident();
b3fbab05 4375 printk("-------------------------------------\n");
1b1d2fb4 4376 lockdep_print_held_locks(current);
fbb9ce95
IM
4377 printk("\nstack backtrace:\n");
4378 dump_stack();
4379}
4380
1b1d2fb4 4381void debug_check_no_locks_held(void)
fbb9ce95 4382{
1b1d2fb4
CC
4383 if (unlikely(current->lockdep_depth > 0))
4384 print_held_locks_bug();
fbb9ce95 4385}
1b1d2fb4 4386EXPORT_SYMBOL_GPL(debug_check_no_locks_held);
fbb9ce95 4387
8dce7a9a 4388#ifdef __KERNEL__
fbb9ce95
IM
4389void debug_show_all_locks(void)
4390{
4391 struct task_struct *g, *p;
4392 int count = 10;
4393 int unlock = 1;
4394
9c35dd7f
JP
4395 if (unlikely(!debug_locks)) {
4396 printk("INFO: lockdep is turned off.\n");
4397 return;
4398 }
fbb9ce95
IM
4399 printk("\nShowing all locks held in the system:\n");
4400
4401 /*
4402 * Here we try to get the tasklist_lock as hard as possible,
4403 * if not successful after 2 seconds we ignore it (but keep
4404 * trying). This is to enable a debug printout even if a
4405 * tasklist_lock-holding task deadlocks or crashes.
4406 */
4407retry:
4408 if (!read_trylock(&tasklist_lock)) {
4409 if (count == 10)
4410 printk("hm, tasklist_lock locked, retrying... ");
4411 if (count) {
4412 count--;
4413 printk(" #%d", 10-count);
4414 mdelay(200);
4415 goto retry;
4416 }
4417 printk(" ignoring it.\n");
4418 unlock = 0;
46fec7ac 4419 } else {
4420 if (count != 10)
4421 printk(KERN_CONT " locked it.\n");
fbb9ce95 4422 }
fbb9ce95
IM
4423
4424 do_each_thread(g, p) {
85684873
IM
4425 /*
4426 * It's not reliable to print a task's held locks
4427 * if it's not sleeping (or if it's not the current
4428 * task):
4429 */
4430 if (p->state == TASK_RUNNING && p != current)
4431 continue;
fbb9ce95
IM
4432 if (p->lockdep_depth)
4433 lockdep_print_held_locks(p);
4434 if (!unlock)
4435 if (read_trylock(&tasklist_lock))
4436 unlock = 1;
4437 } while_each_thread(g, p);
4438
4439 printk("\n");
4440 printk("=============================================\n\n");
4441
4442 if (unlock)
4443 read_unlock(&tasklist_lock);
4444}
fbb9ce95 4445EXPORT_SYMBOL_GPL(debug_show_all_locks);
8dce7a9a 4446#endif
fbb9ce95 4447
82a1fcb9
IM
4448/*
4449 * Careful: only use this function if you are sure that
4450 * the task cannot run in parallel!
4451 */
f1b499f0 4452void debug_show_held_locks(struct task_struct *task)
fbb9ce95 4453{
9c35dd7f
JP
4454 if (unlikely(!debug_locks)) {
4455 printk("INFO: lockdep is turned off.\n");
4456 return;
4457 }
fbb9ce95
IM
4458 lockdep_print_held_locks(task);
4459}
fbb9ce95 4460EXPORT_SYMBOL_GPL(debug_show_held_locks);
b351d164 4461
722a9f92 4462asmlinkage __visible void lockdep_sys_exit(void)
b351d164
PZ
4463{
4464 struct task_struct *curr = current;
4465
4466 if (unlikely(curr->lockdep_depth)) {
4467 if (!debug_locks_off())
4468 return;
b3fbab05
PM
4469 printk("\n");
4470 printk("================================================\n");
4471 printk("[ BUG: lock held when returning to user space! ]\n");
fbdc4b9a 4472 print_kernel_ident();
b3fbab05 4473 printk("------------------------------------------------\n");
b351d164
PZ
4474 printk("%s/%d is leaving the kernel with locks still held!\n",
4475 curr->comm, curr->pid);
4476 lockdep_print_held_locks(curr);
4477 }
4478}
0632eb3d 4479
b3fbab05 4480void lockdep_rcu_suspicious(const char *file, const int line, const char *s)
0632eb3d
PM
4481{
4482 struct task_struct *curr = current;
4483
2b3fc35f 4484#ifndef CONFIG_PROVE_RCU_REPEATEDLY
0632eb3d
PM
4485 if (!debug_locks_off())
4486 return;
2b3fc35f
LJ
4487#endif /* #ifdef CONFIG_PROVE_RCU_REPEATEDLY */
4488 /* Note: the following can be executed concurrently, so be careful. */
b3fbab05 4489 printk("\n");
4d4f88fa
PM
4490 pr_err("===============================\n");
4491 pr_err("[ ERR: suspicious RCU usage. ]\n");
fbdc4b9a 4492 print_kernel_ident();
4d4f88fa
PM
4493 pr_err("-------------------------------\n");
4494 pr_err("%s:%d %s!\n", file, line, s);
4495 pr_err("\nother info that might help us debug this:\n\n");
4496 pr_err("\n%srcu_scheduler_active = %d, debug_locks = %d\n",
c5fdcec9
PM
4497 !rcu_lockdep_current_cpu_online()
4498 ? "RCU used illegally from offline CPU!\n"
5c173eb8 4499 : !rcu_is_watching()
c5fdcec9
PM
4500 ? "RCU used illegally from idle CPU!\n"
4501 : "",
4502 rcu_scheduler_active, debug_locks);
0464e937
FW
4503
4504 /*
4505 * If a CPU is in the RCU-free window in idle (ie: in the section
4506 * between rcu_idle_enter() and rcu_idle_exit(), then RCU
4507 * considers that CPU to be in an "extended quiescent state",
4508 * which means that RCU will be completely ignoring that CPU.
4509 * Therefore, rcu_read_lock() and friends have absolutely no
4510 * effect on a CPU running in that state. In other words, even if
4511 * such an RCU-idle CPU has called rcu_read_lock(), RCU might well
4512 * delete data structures out from under it. RCU really has no
4513 * choice here: we need to keep an RCU-free window in idle where
4514 * the CPU may possibly enter into low power mode. This way we can
4515 * notice an extended quiescent state to other CPUs that started a grace
4516 * period. Otherwise we would delay any grace period as long as we run
4517 * in the idle task.
4518 *
4519 * So complain bitterly if someone does call rcu_read_lock(),
4520 * rcu_read_lock_bh() and so on from extended quiescent states.
4521 */
5c173eb8 4522 if (!rcu_is_watching())
0464e937
FW
4523 printk("RCU used illegally from extended quiescent state!\n");
4524
0632eb3d
PM
4525 lockdep_print_held_locks(curr);
4526 printk("\nstack backtrace:\n");
4527 dump_stack();
4528}
b3fbab05 4529EXPORT_SYMBOL_GPL(lockdep_rcu_suspicious);