-
Notifications
You must be signed in to change notification settings - Fork 861
/
subr_witness.c
2598 lines (2287 loc) · 69 KB
/
subr_witness.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/* $OpenBSD: subr_witness.c,v 1.47 2021/03/23 10:22:20 mpi Exp $ */
/*-
* Copyright (c) 2008 Isilon Systems, Inc.
* Copyright (c) 2008 Ilya Maykov <ivmaykov@gmail.com>
* Copyright (c) 1998 Berkeley Software Design, Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Berkeley Software Design Inc's name may not be used to endorse or
* promote products derived from this software without specific prior
* written permission.
*
* THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL BERKELEY SOFTWARE DESIGN INC BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* from BSDI Id: mutex_witness.c,v 1.1.2.20 2000/04/27 03:10:27 cp Exp
* and BSDI Id: synch_machdep.c,v 2.3.2.39 2000/04/27 03:10:25 cp Exp
*/
/*
* Implementation of the `witness' lock verifier. Originally implemented for
* mutexes in BSD/OS. Extended to handle generic lock objects and lock
* classes in FreeBSD.
*/
/*
* Main Entry: witness
* Pronunciation: 'wit-n&s
* Function: noun
* Etymology: Middle English witnesse, from Old English witnes knowledge,
* testimony, witness, from 2wit
* Date: before 12th century
* 1 : attestation of a fact or event : TESTIMONY
* 2 : one that gives evidence; specifically : one who testifies in
* a cause or before a judicial tribunal
* 3 : one asked to be present at a transaction so as to be able to
* testify to its having taken place
* 4 : one who has personal knowledge of something
* 5 a : something serving as evidence or proof : SIGN
* b : public affirmation by word or example of usually
* religious faith or conviction <the heroic witness to divine
* life -- Pilot>
* 6 capitalized : a member of the Jehovah's Witnesses
*/
/*
* Special rules concerning Giant and lock orders:
*
* 1) Giant must be acquired before any other mutexes. Stated another way,
* no other mutex may be held when Giant is acquired.
*
* 2) Giant must be released when blocking on a sleepable lock.
*
* This rule is less obvious, but is a result of Giant providing the same
* semantics as spl(). Basically, when a thread sleeps, it must release
* Giant. When a thread blocks on a sleepable lock, it sleeps. Hence rule
* 2).
*
* 3) Giant may be acquired before or after sleepable locks.
*
* This rule is also not quite as obvious. Giant may be acquired after
* a sleepable lock because it is a non-sleepable lock and non-sleepable
* locks may always be acquired while holding a sleepable lock. The second
* case, Giant before a sleepable lock, follows from rule 2) above. Suppose
* you have two threads T1 and T2 and a sleepable lock X. Suppose that T1
* acquires X and blocks on Giant. Then suppose that T2 acquires Giant and
* blocks on X. When T2 blocks on X, T2 will release Giant allowing T1 to
* execute. Thus, acquiring Giant both before and after a sleepable lock
* will not result in a lock order reversal.
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#ifdef MULTIPROCESSOR
#include <sys/mplock.h>
#endif
#include <sys/mutex.h>
#include <sys/percpu.h>
#include <sys/proc.h>
#include <sys/sched.h>
#include <sys/stacktrace.h>
#include <sys/stdint.h>
#include <sys/sysctl.h>
#include <sys/syslog.h>
#include <sys/witness.h>
#include <machine/cpu.h>
#include <uvm/uvm_extern.h> /* uvm_pageboot_alloc */
#ifndef DDB
#error "DDB is required for WITNESS"
#endif
#include <machine/db_machdep.h>
#include <ddb/db_access.h>
#include <ddb/db_var.h>
#include <ddb/db_output.h>
#define LI_RECURSEMASK 0x0000ffff /* Recursion depth of lock instance. */
#define LI_EXCLUSIVE 0x00010000 /* Exclusive lock instance. */
#define LI_NORELEASE 0x00020000 /* Lock not allowed to be released. */
#ifndef WITNESS_COUNT
#define WITNESS_COUNT 1536
#endif
#define WITNESS_HASH_SIZE 251 /* Prime, gives load factor < 2 */
#define WITNESS_PENDLIST (1024 + MAXCPUS)
/* Allocate 256 KB of stack data space */
#define WITNESS_LO_DATA_COUNT 2048
/* Prime, gives load factor of ~2 at full load */
#define WITNESS_LO_HASH_SIZE 1021
/*
* XXX: This is somewhat bogus, as we assume here that at most 2048 threads
* will hold LOCK_NCHILDREN locks. We handle failure ok, and we should
* probably be safe for the most part, but it's still a SWAG.
*/
#define LOCK_NCHILDREN 5
#define LOCK_CHILDCOUNT 2048
#define FULLGRAPH_SBUF_SIZE 512
/*
* These flags go in the witness relationship matrix and describe the
* relationship between any two struct witness objects.
*/
#define WITNESS_UNRELATED 0x00 /* No lock order relation. */
#define WITNESS_PARENT 0x01 /* Parent, aka direct ancestor. */
#define WITNESS_ANCESTOR 0x02 /* Direct or indirect ancestor. */
#define WITNESS_CHILD 0x04 /* Child, aka direct descendant. */
#define WITNESS_DESCENDANT 0x08 /* Direct or indirect descendant. */
#define WITNESS_ANCESTOR_MASK (WITNESS_PARENT | WITNESS_ANCESTOR)
#define WITNESS_DESCENDANT_MASK (WITNESS_CHILD | WITNESS_DESCENDANT)
#define WITNESS_RELATED_MASK \
(WITNESS_ANCESTOR_MASK | WITNESS_DESCENDANT_MASK)
#define WITNESS_REVERSAL 0x10 /* A lock order reversal has been
* observed. */
#define WITNESS_RESERVED1 0x20 /* Unused flag, reserved. */
#define WITNESS_RESERVED2 0x40 /* Unused flag, reserved. */
#define WITNESS_LOCK_ORDER_KNOWN 0x80 /* This lock order is known. */
/* Descendant to ancestor flags */
#define WITNESS_DTOA(x) (((x) & WITNESS_RELATED_MASK) >> 2)
/* Ancestor to descendant flags */
#define WITNESS_ATOD(x) (((x) & WITNESS_RELATED_MASK) << 2)
#define WITNESS_INDEX_ASSERT(i) \
KASSERT((i) > 0 && (i) <= w_max_used_index && (i) < witness_count)
/*
* Lock classes. Each lock has a class which describes characteristics
* common to all types of locks of a given class.
*
* Spin locks in general must always protect against preemption, as it is
* an error to perform any type of context switch while holding a spin lock.
* Also, for an individual lock to be recursable, its class must allow
* recursion and the lock itself must explicitly allow recursion.
*/
struct lock_class {
const char *lc_name;
u_int lc_flags;
};
union lock_stack {
union lock_stack *ls_next;
struct stacktrace ls_stack;
};
#define LC_SLEEPLOCK 0x00000001 /* Sleep lock. */
#define LC_SPINLOCK 0x00000002 /* Spin lock. */
#define LC_SLEEPABLE 0x00000004 /* Sleeping allowed with this lock. */
#define LC_RECURSABLE 0x00000008 /* Locks of this type may recurse. */
#define LC_UPGRADABLE 0x00000010 /* Upgrades and downgrades permitted. */
/*
* Lock instances. A lock instance is the data associated with a lock while
* it is held by witness. For example, a lock instance will hold the
* recursion count of a lock. Lock instances are held in lists. Spin locks
* are held in a per-cpu list while sleep locks are held in per-thread list.
*/
struct lock_instance {
struct lock_object *li_lock;
union lock_stack *li_stack;
u_int li_flags;
};
/*
* A simple list type used to build the list of locks held by a thread
* or CPU. We can't simply embed the list in struct lock_object since a
* lock may be held by more than one thread if it is a shared lock. Locks
* are added to the head of the list, so we fill up each list entry from
* "the back" logically. To ease some of the arithmetic, we actually fill
* in each list entry the normal way (children[0] then children[1], etc.) but
* when we traverse the list we read children[count-1] as the first entry
* down to children[0] as the final entry.
*/
struct lock_list_entry {
struct lock_list_entry *ll_next;
struct lock_instance ll_children[LOCK_NCHILDREN];
int ll_count;
};
/*
* The main witness structure. One of these per named lock type in the system
* (for example, "vnode interlock").
*/
struct witness {
const struct lock_type *w_type;
const char *w_subtype;
uint32_t w_index; /* Index in the relationship matrix */
struct lock_class *w_class;
SLIST_ENTRY(witness) w_list; /* List of all witnesses. */
SLIST_ENTRY(witness) w_typelist; /* Witnesses of a type. */
SLIST_ENTRY(witness) w_hash_next; /* Linked list in
* hash buckets. */
uint16_t w_num_ancestors; /* direct/indirect
* ancestor count */
uint16_t w_num_descendants; /* direct/indirect
* descendant count */
int16_t w_ddb_level;
unsigned w_acquired:1;
unsigned w_displayed:1;
unsigned w_reversed:1;
};
SLIST_HEAD(witness_list, witness);
/*
* The witness hash table. Keys are witness names (const char *), elements are
* witness objects (struct witness *).
*/
struct witness_hash {
struct witness_list wh_array[WITNESS_HASH_SIZE];
uint32_t wh_size;
uint32_t wh_count;
};
/*
* Key type for the lock order data hash table.
*/
struct witness_lock_order_key {
uint16_t from;
uint16_t to;
};
struct witness_lock_order_data {
struct stacktrace wlod_stack;
struct witness_lock_order_key wlod_key;
struct witness_lock_order_data *wlod_next;
};
/*
* The witness lock order data hash table. Keys are witness index tuples
* (struct witness_lock_order_key), elements are lock order data objects
* (struct witness_lock_order_data).
*/
struct witness_lock_order_hash {
struct witness_lock_order_data *wloh_array[WITNESS_LO_HASH_SIZE];
u_int wloh_size;
u_int wloh_count;
};
struct witness_pendhelp {
const struct lock_type *wh_type;
struct lock_object *wh_lock;
};
struct witness_cpu {
struct lock_list_entry *wc_spinlocks;
struct lock_list_entry *wc_lle_cache;
union lock_stack *wc_stk_cache;
unsigned int wc_lle_count;
unsigned int wc_stk_count;
} __aligned(CACHELINESIZE);
#define WITNESS_LLE_CACHE_MAX 8
#define WITNESS_STK_CACHE_MAX (WITNESS_LLE_CACHE_MAX * LOCK_NCHILDREN)
struct witness_cpu witness_cpu[MAXCPUS];
/*
* Returns 0 if one of the locks is a spin lock and the other is not.
* Returns 1 otherwise.
*/
static __inline int
witness_lock_type_equal(struct witness *w1, struct witness *w2)
{
return ((w1->w_class->lc_flags & (LC_SLEEPLOCK | LC_SPINLOCK)) ==
(w2->w_class->lc_flags & (LC_SLEEPLOCK | LC_SPINLOCK)));
}
static __inline int
witness_lock_order_key_equal(const struct witness_lock_order_key *a,
const struct witness_lock_order_key *b)
{
return (a->from == b->from && a->to == b->to);
}
static int _isitmyx(struct witness *w1, struct witness *w2, int rmask,
const char *fname);
static void adopt(struct witness *parent, struct witness *child);
static struct witness *enroll(const struct lock_type *, const char *,
struct lock_class *);
static struct lock_instance *find_instance(struct lock_list_entry *list,
const struct lock_object *lock);
static int isitmychild(struct witness *parent, struct witness *child);
static int isitmydescendant(struct witness *parent, struct witness *child);
static void itismychild(struct witness *parent, struct witness *child);
#ifdef DDB
static void db_witness_add_fullgraph(struct witness *parent);
static void witness_ddb_compute_levels(void);
static void witness_ddb_display(int(*)(const char *fmt, ...));
static void witness_ddb_display_descendants(int(*)(const char *fmt, ...),
struct witness *, int indent);
static void witness_ddb_display_list(int(*prnt)(const char *fmt, ...),
struct witness_list *list);
static void witness_ddb_level_descendants(struct witness *parent, int l);
static void witness_ddb_list(struct proc *td);
#endif
static int witness_alloc_stacks(void);
static void witness_debugger(int dump);
static void witness_free(struct witness *m);
static struct witness *witness_get(void);
static uint32_t witness_hash_djb2(const uint8_t *key, uint32_t size);
static struct witness *witness_hash_get(const struct lock_type *,
const char *);
static void witness_hash_put(struct witness *w);
static void witness_init_hash_tables(void);
static void witness_increment_graph_generation(void);
static int witness_list_locks(struct lock_list_entry **,
int (*)(const char *, ...));
static void witness_lock_list_free(struct lock_list_entry *lle);
static struct lock_list_entry *witness_lock_list_get(void);
static void witness_lock_stack_free(union lock_stack *stack);
static union lock_stack *witness_lock_stack_get(void);
static int witness_lock_order_add(struct witness *parent,
struct witness *child);
static int witness_lock_order_check(struct witness *parent,
struct witness *child);
static struct witness_lock_order_data *witness_lock_order_get(
struct witness *parent,
struct witness *child);
static void witness_list_lock(struct lock_instance *instance,
int (*prnt)(const char *fmt, ...));
static void witness_setflag(struct lock_object *lock, int flag, int set);
/*
* If set to 0, lock order checking is disabled. If set to -1,
* witness is completely disabled. Otherwise witness performs full
* lock order checking for all locks. At runtime, lock order checking
* may be toggled. However, witness cannot be reenabled once it is
* completely disabled.
*/
#ifdef WITNESS_WATCH
static int witness_watch = 3;
#else
static int witness_watch = 2;
#endif
#ifdef WITNESS_LOCKTRACE
static int witness_locktrace = 1;
#else
static int witness_locktrace = 0;
#endif
int witness_count = WITNESS_COUNT;
int witness_uninitialized_report = 5;
static struct mutex w_mtx;
static struct rwlock w_ctlock = RWLOCK_INITIALIZER("w_ctlock");
/* w_list */
static struct witness_list w_free = SLIST_HEAD_INITIALIZER(w_free);
static struct witness_list w_all = SLIST_HEAD_INITIALIZER(w_all);
/* w_typelist */
static struct witness_list w_spin = SLIST_HEAD_INITIALIZER(w_spin);
static struct witness_list w_sleep = SLIST_HEAD_INITIALIZER(w_sleep);
/* lock list */
static struct lock_list_entry *w_lock_list_free = NULL;
static struct witness_pendhelp pending_locks[WITNESS_PENDLIST];
static u_int pending_cnt;
static int w_free_cnt, w_spin_cnt, w_sleep_cnt;
static struct witness *w_data;
static uint8_t **w_rmatrix;
static struct lock_list_entry w_locklistdata[LOCK_CHILDCOUNT];
static struct witness_hash w_hash; /* The witness hash table. */
/* The lock order data hash */
static struct witness_lock_order_data w_lodata[WITNESS_LO_DATA_COUNT];
static struct witness_lock_order_data *w_lofree = NULL;
static struct witness_lock_order_hash w_lohash;
static int w_max_used_index = 0;
static unsigned int w_generation = 0;
static union lock_stack *w_lock_stack_free;
static unsigned int w_lock_stack_num;
static struct lock_class lock_class_kernel_lock = {
.lc_name = "kernel_lock",
.lc_flags = LC_SLEEPLOCK | LC_RECURSABLE | LC_SLEEPABLE
};
static struct lock_class lock_class_sched_lock = {
.lc_name = "sched_lock",
.lc_flags = LC_SPINLOCK | LC_RECURSABLE
};
static struct lock_class lock_class_mutex = {
.lc_name = "mutex",
.lc_flags = LC_SPINLOCK
};
static struct lock_class lock_class_rwlock = {
.lc_name = "rwlock",
.lc_flags = LC_SLEEPLOCK | LC_SLEEPABLE | LC_UPGRADABLE
};
static struct lock_class lock_class_rrwlock = {
.lc_name = "rrwlock",
.lc_flags = LC_SLEEPLOCK | LC_RECURSABLE | LC_SLEEPABLE |
LC_UPGRADABLE
};
static struct lock_class *lock_classes[] = {
&lock_class_kernel_lock,
&lock_class_sched_lock,
&lock_class_mutex,
&lock_class_rwlock,
&lock_class_rrwlock,
};
/*
* This global is set to 0 once it becomes safe to use the witness code.
*/
static int witness_cold = 1;
/*
* This global is set to 1 once the static lock orders have been enrolled
* so that a warning can be issued for any spin locks enrolled later.
*/
static int witness_spin_warn = 0;
/*
* The WITNESS-enabled diagnostic code. Note that the witness code does
* assume that the early boot is single-threaded at least until after this
* routine is completed.
*/
void
witness_initialize(void)
{
struct lock_object *lock;
union lock_stack *stacks;
struct witness *w;
int i, s;
w_data = (void *)uvm_pageboot_alloc(sizeof(struct witness) *
witness_count);
memset(w_data, 0, sizeof(struct witness) * witness_count);
w_rmatrix = (void *)uvm_pageboot_alloc(sizeof(*w_rmatrix) *
(witness_count + 1));
for (i = 0; i < witness_count + 1; i++) {
w_rmatrix[i] = (void *)uvm_pageboot_alloc(
sizeof(*w_rmatrix[i]) * (witness_count + 1));
memset(w_rmatrix[i], 0, sizeof(*w_rmatrix[i]) *
(witness_count + 1));
}
mtx_init_flags(&w_mtx, IPL_HIGH, "witness lock", MTX_NOWITNESS);
for (i = witness_count - 1; i >= 0; i--) {
w = &w_data[i];
memset(w, 0, sizeof(*w));
w_data[i].w_index = i; /* Witness index never changes. */
witness_free(w);
}
KASSERTMSG(SLIST_FIRST(&w_free)->w_index == 0,
"%s: Invalid list of free witness objects", __func__);
/* Witness with index 0 is not used to aid in debugging. */
SLIST_REMOVE_HEAD(&w_free, w_list);
w_free_cnt--;
for (i = 0; i < witness_count; i++) {
memset(w_rmatrix[i], 0, sizeof(*w_rmatrix[i]) *
(witness_count + 1));
}
if (witness_locktrace) {
w_lock_stack_num = LOCK_CHILDCOUNT * LOCK_NCHILDREN;
stacks = (void *)uvm_pageboot_alloc(sizeof(*stacks) *
w_lock_stack_num);
}
s = splhigh();
for (i = 0; i < w_lock_stack_num; i++)
witness_lock_stack_free(&stacks[i]);
for (i = 0; i < LOCK_CHILDCOUNT; i++)
witness_lock_list_free(&w_locklistdata[i]);
splx(s);
witness_init_hash_tables();
witness_spin_warn = 1;
/* Iterate through all locks and add them to witness. */
for (i = 0; pending_locks[i].wh_lock != NULL; i++) {
lock = pending_locks[i].wh_lock;
KASSERTMSG(lock->lo_flags & LO_WITNESS,
"%s: lock %s is on pending list but not LO_WITNESS",
__func__, lock->lo_name);
lock->lo_witness = enroll(pending_locks[i].wh_type,
lock->lo_name, LOCK_CLASS(lock));
}
/* Mark the witness code as being ready for use. */
witness_cold = 0;
}
void
witness_init(struct lock_object *lock, const struct lock_type *type)
{
struct lock_class *class;
/* Various sanity checks. */
class = LOCK_CLASS(lock);
if ((lock->lo_flags & LO_RECURSABLE) != 0 &&
(class->lc_flags & LC_RECURSABLE) == 0)
panic("%s: lock (%s) %s can not be recursable",
__func__, class->lc_name, lock->lo_name);
if ((lock->lo_flags & LO_SLEEPABLE) != 0 &&
(class->lc_flags & LC_SLEEPABLE) == 0)
panic("%s: lock (%s) %s can not be sleepable",
__func__, class->lc_name, lock->lo_name);
if ((lock->lo_flags & LO_UPGRADABLE) != 0 &&
(class->lc_flags & LC_UPGRADABLE) == 0)
panic("%s: lock (%s) %s can not be upgradable",
__func__, class->lc_name, lock->lo_name);
/*
* If we shouldn't watch this lock, then just clear lo_witness.
* Record the type in case the lock becomes watched later.
* Otherwise, if witness_cold is set, then it is too early to
* enroll this lock, so defer it to witness_initialize() by adding
* it to the pending_locks list. If it is not too early, then enroll
* the lock now.
*/
if (witness_watch < 1 || panicstr != NULL || db_active ||
(lock->lo_flags & LO_WITNESS) == 0) {
lock->lo_witness = NULL;
lock->lo_type = type;
} else if (witness_cold) {
pending_locks[pending_cnt].wh_lock = lock;
pending_locks[pending_cnt++].wh_type = type;
if (pending_cnt > WITNESS_PENDLIST)
panic("%s: pending locks list is too small, "
"increase WITNESS_PENDLIST",
__func__);
} else
lock->lo_witness = enroll(type, lock->lo_name, class);
}
static inline int
is_kernel_lock(const struct lock_object *lock)
{
#ifdef MULTIPROCESSOR
return (lock == &kernel_lock.mpl_lock_obj);
#else
return (0);
#endif
}
#ifdef DDB
static void
witness_ddb_compute_levels(void)
{
struct witness *w;
/*
* First clear all levels.
*/
SLIST_FOREACH(w, &w_all, w_list)
w->w_ddb_level = -1;
/*
* Look for locks with no parents and level all their descendants.
*/
SLIST_FOREACH(w, &w_all, w_list) {
/* If the witness has ancestors (is not a root), skip it. */
if (w->w_num_ancestors > 0)
continue;
witness_ddb_level_descendants(w, 0);
}
}
static void
witness_ddb_level_descendants(struct witness *w, int l)
{
int i;
if (w->w_ddb_level >= l)
return;
w->w_ddb_level = l;
l++;
for (i = 1; i <= w_max_used_index; i++) {
if (w_rmatrix[w->w_index][i] & WITNESS_PARENT)
witness_ddb_level_descendants(&w_data[i], l);
}
}
static void
witness_ddb_display_descendants(int(*prnt)(const char *fmt, ...),
struct witness *w, int indent)
{
int i;
for (i = 0; i < indent; i++)
prnt(" ");
prnt("%s (type: %s, depth: %d)",
w->w_type->lt_name, w->w_class->lc_name, w->w_ddb_level);
if (w->w_displayed) {
prnt(" -- (already displayed)\n");
return;
}
w->w_displayed = 1;
if (!w->w_acquired)
prnt(" -- never acquired\n");
else
prnt("\n");
indent++;
WITNESS_INDEX_ASSERT(w->w_index);
for (i = 1; i <= w_max_used_index; i++) {
if (w_rmatrix[w->w_index][i] & WITNESS_PARENT)
witness_ddb_display_descendants(prnt, &w_data[i],
indent);
}
}
static void
witness_ddb_display_list(int(*prnt)(const char *fmt, ...),
struct witness_list *list)
{
struct witness *w;
SLIST_FOREACH(w, list, w_typelist) {
if (!w->w_acquired || w->w_ddb_level > 0)
continue;
/* This lock has no anscestors - display its descendants. */
witness_ddb_display_descendants(prnt, w, 0);
}
}
static void
witness_ddb_display(int(*prnt)(const char *fmt, ...))
{
struct witness *w;
KASSERTMSG(witness_cold == 0, "%s: witness_cold", __func__);
witness_ddb_compute_levels();
/* Clear all the displayed flags. */
SLIST_FOREACH(w, &w_all, w_list)
w->w_displayed = 0;
/*
* First, handle sleep locks which have been acquired at least
* once.
*/
prnt("Sleep locks:\n");
witness_ddb_display_list(prnt, &w_sleep);
/*
* Now do spin locks which have been acquired at least once.
*/
prnt("\nSpin locks:\n");
witness_ddb_display_list(prnt, &w_spin);
/*
* Finally, any locks which have not been acquired yet.
*/
prnt("\nLocks which were never acquired:\n");
SLIST_FOREACH(w, &w_all, w_list) {
if (w->w_acquired)
continue;
prnt("%s (type: %s, depth: %d)\n", w->w_type->lt_name,
w->w_class->lc_name, w->w_ddb_level);
}
}
#endif /* DDB */
int
witness_defineorder(struct lock_object *lock1, struct lock_object *lock2)
{
if (witness_watch < 0 || panicstr != NULL || db_active)
return (0);
/* Require locks that witness knows about. */
if (lock1 == NULL || lock1->lo_witness == NULL || lock2 == NULL ||
lock2->lo_witness == NULL)
return (EINVAL);
MUTEX_ASSERT_UNLOCKED(&w_mtx);
mtx_enter(&w_mtx);
/*
* If we already have either an explicit or implied lock order that
* is the other way around, then return an error.
*/
if (witness_watch &&
isitmydescendant(lock2->lo_witness, lock1->lo_witness)) {
mtx_leave(&w_mtx);
return (EINVAL);
}
/* Try to add the new order. */
itismychild(lock1->lo_witness, lock2->lo_witness);
mtx_leave(&w_mtx);
return (0);
}
void
witness_checkorder(struct lock_object *lock, int flags,
struct lock_object *interlock)
{
struct lock_list_entry *lock_list, *lle;
struct lock_instance *lock1, *lock2, *plock;
struct lock_class *class, *iclass;
struct proc *p;
struct witness *w, *w1;
int i, j, s;
if (witness_cold || witness_watch < 1 || panicstr != NULL || db_active)
return;
if ((lock->lo_flags & LO_INITIALIZED) == 0) {
if (witness_uninitialized_report > 0) {
witness_uninitialized_report--;
printf("witness: lock_object uninitialized: %p\n", lock);
witness_debugger(1);
}
lock->lo_flags |= LO_INITIALIZED;
}
if ((lock->lo_flags & LO_WITNESS) == 0)
return;
w = lock->lo_witness;
class = LOCK_CLASS(lock);
if (w == NULL)
w = lock->lo_witness =
enroll(lock->lo_type, lock->lo_name, class);
p = curproc;
if (class->lc_flags & LC_SLEEPLOCK) {
/*
* Since spin locks include a critical section, this check
* implicitly enforces a lock order of all sleep locks before
* all spin locks.
*/
lock_list = witness_cpu[cpu_number()].wc_spinlocks;
if (lock_list != NULL && lock_list->ll_count > 0) {
panic("acquiring blockable sleep lock with "
"spinlock or critical section held (%s) %s",
class->lc_name, lock->lo_name);
}
/*
* If this is the first lock acquired then just return as
* no order checking is needed.
*/
lock_list = p->p_sleeplocks;
if (lock_list == NULL || lock_list->ll_count == 0)
return;
} else {
/*
* If this is the first lock, just return as no order
* checking is needed.
*/
lock_list = witness_cpu[cpu_number()].wc_spinlocks;
if (lock_list == NULL || lock_list->ll_count == 0)
return;
}
s = splhigh();
/*
* Check to see if we are recursing on a lock we already own. If
* so, make sure that we don't mismatch exclusive and shared lock
* acquires.
*/
lock1 = find_instance(lock_list, lock);
if (lock1 != NULL) {
if ((lock1->li_flags & LI_EXCLUSIVE) != 0 &&
(flags & LOP_EXCLUSIVE) == 0) {
printf("witness: shared lock of (%s) %s "
"while exclusively locked\n",
class->lc_name, lock->lo_name);
panic("excl->share");
}
if ((lock1->li_flags & LI_EXCLUSIVE) == 0 &&
(flags & LOP_EXCLUSIVE) != 0) {
printf("witness: exclusive lock of (%s) %s "
"while share locked\n",
class->lc_name, lock->lo_name);
panic("share->excl");
}
goto out_splx;
}
/* Warn if the interlock is not locked exactly once. */
if (interlock != NULL) {
iclass = LOCK_CLASS(interlock);
lock1 = find_instance(lock_list, interlock);
if (lock1 == NULL)
panic("interlock (%s) %s not locked",
iclass->lc_name, interlock->lo_name);
else if ((lock1->li_flags & LI_RECURSEMASK) != 0)
panic("interlock (%s) %s recursed",
iclass->lc_name, interlock->lo_name);
}
/*
* Find the previously acquired lock, but ignore interlocks.
*/
plock = &lock_list->ll_children[lock_list->ll_count - 1];
if (interlock != NULL && plock->li_lock == interlock) {
if (lock_list->ll_count > 1)
plock =
&lock_list->ll_children[lock_list->ll_count - 2];
else {
lle = lock_list->ll_next;
/*
* The interlock is the only lock we hold, so
* simply return.
*/
if (lle == NULL)
goto out_splx;
plock = &lle->ll_children[lle->ll_count - 1];
}
}
/*
* Try to perform most checks without a lock. If this succeeds we
* can skip acquiring the lock and return success. Otherwise we redo
* the check with the lock held to handle races with concurrent updates.
*/
w1 = plock->li_lock->lo_witness;
if (witness_lock_order_check(w1, w))
goto out_splx;
mtx_enter(&w_mtx);
if (witness_lock_order_check(w1, w))
goto out;
witness_lock_order_add(w1, w);
/*
* Check for duplicate locks of the same type. Note that we only
* have to check for this on the last lock we just acquired. Any
* other cases will be caught as lock order violations.
*/
if (w1 == w) {
i = w->w_index;
if (!(lock->lo_flags & LO_DUPOK) && !(flags & LOP_DUPOK) &&
!(w_rmatrix[i][i] & WITNESS_REVERSAL)) {
w_rmatrix[i][i] |= WITNESS_REVERSAL;
w->w_reversed = 1;
mtx_leave(&w_mtx);
printf("witness: acquiring duplicate lock of "
"same type: \"%s\"\n", w->w_type->lt_name);
printf(" 1st %s\n", plock->li_lock->lo_name);
printf(" 2nd %s\n", lock->lo_name);
witness_debugger(1);
} else
mtx_leave(&w_mtx);
goto out_splx;
}
MUTEX_ASSERT_LOCKED(&w_mtx);
/*
* If we know that the lock we are acquiring comes after
* the lock we most recently acquired in the lock order tree,
* then there is no need for any further checks.
*/
if (isitmychild(w1, w))
goto out;
for (j = 0, lle = lock_list; lle != NULL; lle = lle->ll_next) {
for (i = lle->ll_count - 1; i >= 0; i--, j++) {
KASSERT(j < LOCK_CHILDCOUNT * LOCK_NCHILDREN);
lock1 = &lle->ll_children[i];
/*
* Ignore the interlock.
*/
if (interlock == lock1->li_lock)
continue;
/*
* If this lock doesn't undergo witness checking,
* then skip it.
*/
w1 = lock1->li_lock->lo_witness;
if (w1 == NULL) {
KASSERTMSG((lock1->li_lock->lo_flags &
LO_WITNESS) == 0,
"lock missing witness structure");
continue;
}
/*
* If we are locking Giant and this is a sleepable
* lock, then skip it.
*/
if ((lock1->li_lock->lo_flags & LO_SLEEPABLE) != 0 &&
is_kernel_lock(lock))
continue;
/*
* If we are locking a sleepable lock and this lock
* is Giant, then skip it.
*/
if ((lock->lo_flags & LO_SLEEPABLE) != 0 &&
is_kernel_lock(lock1->li_lock))
continue;
/*
* If we are locking a sleepable lock and this lock
* isn't sleepable, we want to treat it as a lock
* order violation to enfore a general lock order of
* sleepable locks before non-sleepable locks.
*/
if (((lock->lo_flags & LO_SLEEPABLE) != 0 &&
(lock1->li_lock->lo_flags & LO_SLEEPABLE) == 0))
goto reversal;
/*
* If we are locking Giant and this is a non-sleepable
* lock, then treat it as a reversal.
*/
if ((lock1->li_lock->lo_flags & LO_SLEEPABLE) == 0 &&
is_kernel_lock(lock))
goto reversal;
/*
* Check the lock order hierarchy for a reveresal.
*/
if (!isitmydescendant(w, w1))
continue;
reversal:
/*
* We have a lock order violation, check to see if it
* is allowed or has already been yelled about.
*/
/* Bail if this violation is known */
if (w_rmatrix[w1->w_index][w->w_index] & WITNESS_REVERSAL)
goto out;
/* Record this as a violation */
w_rmatrix[w1->w_index][w->w_index] |= WITNESS_REVERSAL;