-
Notifications
You must be signed in to change notification settings - Fork 3k
/
Copy pathdtrace.c
18457 lines (15494 loc) · 463 KB
/
dtrace.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
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2016, Joyent, Inc. All rights reserved.
* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
*/
/*
* DTrace - Dynamic Tracing for Solaris
*
* This is the implementation of the Solaris Dynamic Tracing framework
* (DTrace). The user-visible interface to DTrace is described at length in
* the "Solaris Dynamic Tracing Guide". The interfaces between the libdtrace
* library, the in-kernel DTrace framework, and the DTrace providers are
* described in the block comments in the <sys/dtrace.h> header file. The
* internal architecture of DTrace is described in the block comments in the
* <sys/dtrace_impl.h> header file. The comments contained within the DTrace
* implementation very much assume mastery of all of these sources; if one has
* an unanswered question about the implementation, one should consult them
* first.
*
* The functions here are ordered roughly as follows:
*
* - Probe context functions
* - Probe hashing functions
* - Non-probe context utility functions
* - Matching functions
* - Provider-to-Framework API functions
* - Probe management functions
* - DIF object functions
* - Format functions
* - Predicate functions
* - ECB functions
* - Buffer functions
* - Enabling functions
* - DOF functions
* - Anonymous enabling functions
* - Consumer state functions
* - Helper functions
* - Hook functions
* - Driver cookbook functions
*
* Each group of functions begins with a block comment labelled the "DTrace
* [Group] Functions", allowing one to find each block by searching forward
* on capital-f functions.
*/
#include <sys/errno.h>
#include <sys/param.h>
#include <sys/types.h>
#ifndef illumos
#include <sys/time.h>
#endif
#include <sys/stat.h>
#include <sys/conf.h>
#include <sys/systm.h>
#include <sys/endian.h>
#ifdef illumos
#include <sys/ddi.h>
#include <sys/sunddi.h>
#endif
#include <sys/cpuvar.h>
#include <sys/kmem.h>
#ifdef illumos
#include <sys/strsubr.h>
#endif
#include <sys/sysmacros.h>
#include <sys/dtrace_impl.h>
#include <sys/atomic.h>
#include <sys/cmn_err.h>
#ifdef illumos
#include <sys/mutex_impl.h>
#include <sys/rwlock_impl.h>
#endif
#include <sys/ctf_api.h>
#ifdef illumos
#include <sys/panic.h>
#include <sys/priv_impl.h>
#endif
#ifdef illumos
#include <sys/cred_impl.h>
#include <sys/procfs_isa.h>
#endif
#include <sys/taskq.h>
#ifdef illumos
#include <sys/mkdev.h>
#include <sys/kdi.h>
#endif
#include <sys/zone.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include "strtolctype.h"
/* FreeBSD includes: */
#ifndef illumos
#include <sys/callout.h>
#include <sys/ctype.h>
#include <sys/eventhandler.h>
#include <sys/limits.h>
#include <sys/linker.h>
#include <sys/kdb.h>
#include <sys/jail.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/ptrace.h>
#include <sys/random.h>
#include <sys/rwlock.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/mount.h>
#undef AT_UID
#undef AT_GID
#include <sys/vnode.h>
#include <sys/cred.h>
#include <sys/dtrace_bsd.h>
#include <netinet/in.h>
#include "dtrace_cddl.h"
#include "dtrace_debug.c"
#endif
#include "dtrace_xoroshiro128_plus.h"
/*
* DTrace Tunable Variables
*
* The following variables may be tuned by adding a line to /etc/system that
* includes both the name of the DTrace module ("dtrace") and the name of the
* variable. For example:
*
* set dtrace:dtrace_destructive_disallow = 1
*
* In general, the only variables that one should be tuning this way are those
* that affect system-wide DTrace behavior, and for which the default behavior
* is undesirable. Most of these variables are tunable on a per-consumer
* basis using DTrace options, and need not be tuned on a system-wide basis.
* When tuning these variables, avoid pathological values; while some attempt
* is made to verify the integrity of these variables, they are not considered
* part of the supported interface to DTrace, and they are therefore not
* checked comprehensively. Further, these variables should not be tuned
* dynamically via "mdb -kw" or other means; they should only be tuned via
* /etc/system.
*/
int dtrace_destructive_disallow = 0;
#ifndef illumos
/* Positive logic version of dtrace_destructive_disallow for loader tunable */
int dtrace_allow_destructive = 1;
#endif
dtrace_optval_t dtrace_nonroot_maxsize = (16 * 1024 * 1024);
size_t dtrace_difo_maxsize = (256 * 1024);
dtrace_optval_t dtrace_dof_maxsize = (8 * 1024 * 1024);
size_t dtrace_statvar_maxsize = (16 * 1024);
size_t dtrace_actions_max = (16 * 1024);
size_t dtrace_retain_max = 1024;
dtrace_optval_t dtrace_helper_actions_max = 128;
dtrace_optval_t dtrace_helper_providers_max = 32;
dtrace_optval_t dtrace_dstate_defsize = (1 * 1024 * 1024);
size_t dtrace_strsize_default = 256;
dtrace_optval_t dtrace_cleanrate_default = 9900990; /* 101 hz */
dtrace_optval_t dtrace_cleanrate_min = 200000; /* 5000 hz */
dtrace_optval_t dtrace_cleanrate_max = (uint64_t)60 * NANOSEC; /* 1/minute */
dtrace_optval_t dtrace_aggrate_default = NANOSEC; /* 1 hz */
dtrace_optval_t dtrace_statusrate_default = NANOSEC; /* 1 hz */
dtrace_optval_t dtrace_statusrate_max = (hrtime_t)10 * NANOSEC; /* 6/minute */
dtrace_optval_t dtrace_switchrate_default = NANOSEC; /* 1 hz */
dtrace_optval_t dtrace_nspec_default = 1;
dtrace_optval_t dtrace_specsize_default = 32 * 1024;
dtrace_optval_t dtrace_stackframes_default = 20;
dtrace_optval_t dtrace_ustackframes_default = 20;
dtrace_optval_t dtrace_jstackframes_default = 50;
dtrace_optval_t dtrace_jstackstrsize_default = 512;
int dtrace_msgdsize_max = 128;
hrtime_t dtrace_chill_max = MSEC2NSEC(500); /* 500 ms */
hrtime_t dtrace_chill_interval = NANOSEC; /* 1000 ms */
int dtrace_devdepth_max = 32;
int dtrace_err_verbose;
hrtime_t dtrace_deadman_interval = NANOSEC;
hrtime_t dtrace_deadman_timeout = (hrtime_t)10 * NANOSEC;
hrtime_t dtrace_deadman_user = (hrtime_t)30 * NANOSEC;
hrtime_t dtrace_unregister_defunct_reap = (hrtime_t)60 * NANOSEC;
#ifndef illumos
int dtrace_memstr_max = 4096;
int dtrace_bufsize_max_frac = 128;
#endif
/*
* DTrace External Variables
*
* As dtrace(7D) is a kernel module, any DTrace variables are obviously
* available to DTrace consumers via the backtick (`) syntax. One of these,
* dtrace_zero, is made deliberately so: it is provided as a source of
* well-known, zero-filled memory. While this variable is not documented,
* it is used by some translators as an implementation detail.
*/
const char dtrace_zero[256] = { 0 }; /* zero-filled memory */
/*
* DTrace Internal Variables
*/
#ifdef illumos
static dev_info_t *dtrace_devi; /* device info */
#endif
#ifdef illumos
static vmem_t *dtrace_arena; /* probe ID arena */
static vmem_t *dtrace_minor; /* minor number arena */
#else
static taskq_t *dtrace_taskq; /* task queue */
static struct unrhdr *dtrace_arena; /* Probe ID number. */
#endif
static dtrace_probe_t **dtrace_probes; /* array of all probes */
static int dtrace_nprobes; /* number of probes */
static dtrace_provider_t *dtrace_provider; /* provider list */
static dtrace_meta_t *dtrace_meta_pid; /* user-land meta provider */
static int dtrace_opens; /* number of opens */
static int dtrace_helpers; /* number of helpers */
static int dtrace_getf; /* number of unpriv getf()s */
#ifdef illumos
static void *dtrace_softstate; /* softstate pointer */
#endif
static dtrace_hash_t *dtrace_bymod; /* probes hashed by module */
static dtrace_hash_t *dtrace_byfunc; /* probes hashed by function */
static dtrace_hash_t *dtrace_byname; /* probes hashed by name */
static dtrace_toxrange_t *dtrace_toxrange; /* toxic range array */
static int dtrace_toxranges; /* number of toxic ranges */
static int dtrace_toxranges_max; /* size of toxic range array */
static dtrace_anon_t dtrace_anon; /* anonymous enabling */
static kmem_cache_t *dtrace_state_cache; /* cache for dynamic state */
static uint64_t dtrace_vtime_references; /* number of vtimestamp refs */
static kthread_t *dtrace_panicked; /* panicking thread */
static dtrace_ecb_t *dtrace_ecb_create_cache; /* cached created ECB */
static dtrace_genid_t dtrace_probegen; /* current probe generation */
static dtrace_helpers_t *dtrace_deferred_pid; /* deferred helper list */
static dtrace_enabling_t *dtrace_retained; /* list of retained enablings */
static dtrace_genid_t dtrace_retained_gen; /* current retained enab gen */
static dtrace_dynvar_t dtrace_dynhash_sink; /* end of dynamic hash chains */
static int dtrace_dynvar_failclean; /* dynvars failed to clean */
#ifndef illumos
static struct mtx dtrace_unr_mtx;
MTX_SYSINIT(dtrace_unr_mtx, &dtrace_unr_mtx, "Unique resource identifier", MTX_DEF);
static eventhandler_tag dtrace_kld_load_tag;
static eventhandler_tag dtrace_kld_unload_try_tag;
#endif
/*
* DTrace Locking
* DTrace is protected by three (relatively coarse-grained) locks:
*
* (1) dtrace_lock is required to manipulate essentially any DTrace state,
* including enabling state, probes, ECBs, consumer state, helper state,
* etc. Importantly, dtrace_lock is _not_ required when in probe context;
* probe context is lock-free -- synchronization is handled via the
* dtrace_sync() cross call mechanism.
*
* (2) dtrace_provider_lock is required when manipulating provider state, or
* when provider state must be held constant.
*
* (3) dtrace_meta_lock is required when manipulating meta provider state, or
* when meta provider state must be held constant.
*
* The lock ordering between these three locks is dtrace_meta_lock before
* dtrace_provider_lock before dtrace_lock. (In particular, there are
* several places where dtrace_provider_lock is held by the framework as it
* calls into the providers -- which then call back into the framework,
* grabbing dtrace_lock.)
*
* There are two other locks in the mix: mod_lock and cpu_lock. With respect
* to dtrace_provider_lock and dtrace_lock, cpu_lock continues its historical
* role as a coarse-grained lock; it is acquired before both of these locks.
* With respect to dtrace_meta_lock, its behavior is stranger: cpu_lock must
* be acquired _between_ dtrace_meta_lock and any other DTrace locks.
* mod_lock is similar with respect to dtrace_provider_lock in that it must be
* acquired _between_ dtrace_provider_lock and dtrace_lock.
*/
static kmutex_t dtrace_lock; /* probe state lock */
static kmutex_t dtrace_provider_lock; /* provider state lock */
static kmutex_t dtrace_meta_lock; /* meta-provider state lock */
#ifndef illumos
/* XXX FreeBSD hacks. */
#define cr_suid cr_svuid
#define cr_sgid cr_svgid
#define ipaddr_t in_addr_t
#define mod_modname pathname
#define vuprintf vprintf
#ifndef crgetzoneid
#define crgetzoneid(_a) 0
#endif
#define ttoproc(_a) ((_a)->td_proc)
#define SNOCD 0
#define CPU_ON_INTR(_a) 0
#define PRIV_EFFECTIVE (1 << 0)
#define PRIV_DTRACE_KERNEL (1 << 1)
#define PRIV_DTRACE_PROC (1 << 2)
#define PRIV_DTRACE_USER (1 << 3)
#define PRIV_PROC_OWNER (1 << 4)
#define PRIV_PROC_ZONE (1 << 5)
#define PRIV_ALL ~0
SYSCTL_DECL(_debug_dtrace);
SYSCTL_DECL(_kern_dtrace);
#endif
#ifdef illumos
#define curcpu CPU->cpu_id
#endif
/*
* DTrace Provider Variables
*
* These are the variables relating to DTrace as a provider (that is, the
* provider of the BEGIN, END, and ERROR probes).
*/
static dtrace_pattr_t dtrace_provider_attr = {
{ DTRACE_STABILITY_STABLE, DTRACE_STABILITY_STABLE, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
{ DTRACE_STABILITY_STABLE, DTRACE_STABILITY_STABLE, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_STABLE, DTRACE_STABILITY_STABLE, DTRACE_CLASS_COMMON },
};
static void
dtrace_nullop(void)
{}
static dtrace_pops_t dtrace_provider_ops = {
.dtps_provide = (void (*)(void *, dtrace_probedesc_t *))dtrace_nullop,
.dtps_provide_module = (void (*)(void *, modctl_t *))dtrace_nullop,
.dtps_enable = (void (*)(void *, dtrace_id_t, void *))dtrace_nullop,
.dtps_disable = (void (*)(void *, dtrace_id_t, void *))dtrace_nullop,
.dtps_suspend = (void (*)(void *, dtrace_id_t, void *))dtrace_nullop,
.dtps_resume = (void (*)(void *, dtrace_id_t, void *))dtrace_nullop,
.dtps_getargdesc = NULL,
.dtps_getargval = NULL,
.dtps_usermode = NULL,
.dtps_destroy = (void (*)(void *, dtrace_id_t, void *))dtrace_nullop,
};
static dtrace_id_t dtrace_probeid_begin; /* special BEGIN probe */
static dtrace_id_t dtrace_probeid_end; /* special END probe */
dtrace_id_t dtrace_probeid_error; /* special ERROR probe */
/*
* DTrace Helper Tracing Variables
*
* These variables should be set dynamically to enable helper tracing. The
* only variables that should be set are dtrace_helptrace_enable (which should
* be set to a non-zero value to allocate helper tracing buffers on the next
* open of /dev/dtrace) and dtrace_helptrace_disable (which should be set to a
* non-zero value to deallocate helper tracing buffers on the next close of
* /dev/dtrace). When (and only when) helper tracing is disabled, the
* buffer size may also be set via dtrace_helptrace_bufsize.
*/
int dtrace_helptrace_enable = 0;
int dtrace_helptrace_disable = 0;
int dtrace_helptrace_bufsize = 16 * 1024 * 1024;
uint32_t dtrace_helptrace_nlocals;
static dtrace_helptrace_t *dtrace_helptrace_buffer;
static uint32_t dtrace_helptrace_next = 0;
static int dtrace_helptrace_wrapped = 0;
/*
* DTrace Error Hashing
*
* On DEBUG kernels, DTrace will track the errors that has seen in a hash
* table. This is very useful for checking coverage of tests that are
* expected to induce DIF or DOF processing errors, and may be useful for
* debugging problems in the DIF code generator or in DOF generation . The
* error hash may be examined with the ::dtrace_errhash MDB dcmd.
*/
#ifdef DEBUG
static dtrace_errhash_t dtrace_errhash[DTRACE_ERRHASHSZ];
static const char *dtrace_errlast;
static kthread_t *dtrace_errthread;
static kmutex_t dtrace_errlock;
#endif
/*
* DTrace Macros and Constants
*
* These are various macros that are useful in various spots in the
* implementation, along with a few random constants that have no meaning
* outside of the implementation. There is no real structure to this cpp
* mishmash -- but is there ever?
*/
#define DTRACE_HASHSTR(hash, probe) \
dtrace_hash_str(*((char **)((uintptr_t)(probe) + (hash)->dth_stroffs)))
#define DTRACE_HASHNEXT(hash, probe) \
(dtrace_probe_t **)((uintptr_t)(probe) + (hash)->dth_nextoffs)
#define DTRACE_HASHPREV(hash, probe) \
(dtrace_probe_t **)((uintptr_t)(probe) + (hash)->dth_prevoffs)
#define DTRACE_HASHEQ(hash, lhs, rhs) \
(strcmp(*((char **)((uintptr_t)(lhs) + (hash)->dth_stroffs)), \
*((char **)((uintptr_t)(rhs) + (hash)->dth_stroffs))) == 0)
#define DTRACE_AGGHASHSIZE_SLEW 17
#define DTRACE_V4MAPPED_OFFSET (sizeof (uint32_t) * 3)
/*
* The key for a thread-local variable consists of the lower 61 bits of the
* t_did, plus the 3 bits of the highest active interrupt above LOCK_LEVEL.
* We add DIF_VARIABLE_MAX to t_did to assure that the thread key is never
* equal to a variable identifier. This is necessary (but not sufficient) to
* assure that global associative arrays never collide with thread-local
* variables. To guarantee that they cannot collide, we must also define the
* order for keying dynamic variables. That order is:
*
* [ key0 ] ... [ keyn ] [ variable-key ] [ tls-key ]
*
* Because the variable-key and the tls-key are in orthogonal spaces, there is
* no way for a global variable key signature to match a thread-local key
* signature.
*/
#ifdef illumos
#define DTRACE_TLS_THRKEY(where) { \
uint_t intr = 0; \
uint_t actv = CPU->cpu_intr_actv >> (LOCK_LEVEL + 1); \
for (; actv; actv >>= 1) \
intr++; \
ASSERT(intr < (1 << 3)); \
(where) = ((curthread->t_did + DIF_VARIABLE_MAX) & \
(((uint64_t)1 << 61) - 1)) | ((uint64_t)intr << 61); \
}
#else
#define DTRACE_TLS_THRKEY(where) { \
solaris_cpu_t *_c = &solaris_cpu[curcpu]; \
uint_t intr = 0; \
uint_t actv = _c->cpu_intr_actv; \
for (; actv; actv >>= 1) \
intr++; \
ASSERT(intr < (1 << 3)); \
(where) = ((curthread->td_tid + DIF_VARIABLE_MAX) & \
(((uint64_t)1 << 61) - 1)) | ((uint64_t)intr << 61); \
}
#endif
#define DT_BSWAP_8(x) ((x) & 0xff)
#define DT_BSWAP_16(x) ((DT_BSWAP_8(x) << 8) | DT_BSWAP_8((x) >> 8))
#define DT_BSWAP_32(x) ((DT_BSWAP_16(x) << 16) | DT_BSWAP_16((x) >> 16))
#define DT_BSWAP_64(x) ((DT_BSWAP_32(x) << 32) | DT_BSWAP_32((x) >> 32))
#define DT_MASK_LO 0x00000000FFFFFFFFULL
#define DTRACE_STORE(type, tomax, offset, what) \
*((type *)((uintptr_t)(tomax) + (size_t)offset)) = (type)(what);
#if !defined(__x86) && !defined(__aarch64__)
#define DTRACE_ALIGNCHECK(addr, size, flags) \
if (addr & (size - 1)) { \
*flags |= CPU_DTRACE_BADALIGN; \
cpu_core[curcpu].cpuc_dtrace_illval = addr; \
return (0); \
}
#else
#define DTRACE_ALIGNCHECK(addr, size, flags)
#endif
/*
* Test whether a range of memory starting at testaddr of size testsz falls
* within the range of memory described by addr, sz. We take care to avoid
* problems with overflow and underflow of the unsigned quantities, and
* disallow all negative sizes. Ranges of size 0 are allowed.
*/
#define DTRACE_INRANGE(testaddr, testsz, baseaddr, basesz) \
((testaddr) - (uintptr_t)(baseaddr) < (basesz) && \
(testaddr) + (testsz) - (uintptr_t)(baseaddr) <= (basesz) && \
(testaddr) + (testsz) >= (testaddr))
#define DTRACE_RANGE_REMAIN(remp, addr, baseaddr, basesz) \
do { \
if ((remp) != NULL) { \
*(remp) = (uintptr_t)(baseaddr) + (basesz) - (addr); \
} \
} while (0)
/*
* Test whether alloc_sz bytes will fit in the scratch region. We isolate
* alloc_sz on the righthand side of the comparison in order to avoid overflow
* or underflow in the comparison with it. This is simpler than the INRANGE
* check above, because we know that the dtms_scratch_ptr is valid in the
* range. Allocations of size zero are allowed.
*/
#define DTRACE_INSCRATCH(mstate, alloc_sz) \
((mstate)->dtms_scratch_base + (mstate)->dtms_scratch_size - \
(mstate)->dtms_scratch_ptr >= (alloc_sz))
#define DTRACE_INSCRATCHPTR(mstate, ptr, howmany) \
((ptr) >= (mstate)->dtms_scratch_base && \
(ptr) <= \
((mstate)->dtms_scratch_base + (mstate)->dtms_scratch_size - (howmany)))
#define DTRACE_LOADFUNC(bits) \
/*CSTYLED*/ \
uint##bits##_t \
dtrace_load##bits(uintptr_t addr) \
{ \
size_t size = bits / NBBY; \
/*CSTYLED*/ \
uint##bits##_t rval; \
int i; \
volatile uint16_t *flags = (volatile uint16_t *) \
&cpu_core[curcpu].cpuc_dtrace_flags; \
\
DTRACE_ALIGNCHECK(addr, size, flags); \
\
for (i = 0; i < dtrace_toxranges; i++) { \
if (addr >= dtrace_toxrange[i].dtt_limit) \
continue; \
\
if (addr + size <= dtrace_toxrange[i].dtt_base) \
continue; \
\
/* \
* This address falls within a toxic region; return 0. \
*/ \
*flags |= CPU_DTRACE_BADADDR; \
cpu_core[curcpu].cpuc_dtrace_illval = addr; \
return (0); \
} \
\
*flags |= CPU_DTRACE_NOFAULT; \
/*CSTYLED*/ \
rval = *((volatile uint##bits##_t *)addr); \
*flags &= ~CPU_DTRACE_NOFAULT; \
\
return (!(*flags & CPU_DTRACE_FAULT) ? rval : 0); \
}
#ifdef _LP64
#define dtrace_loadptr dtrace_load64
#else
#define dtrace_loadptr dtrace_load32
#endif
#define DTRACE_DYNHASH_FREE 0
#define DTRACE_DYNHASH_SINK 1
#define DTRACE_DYNHASH_VALID 2
#define DTRACE_MATCH_NEXT 0
#define DTRACE_MATCH_DONE 1
#define DTRACE_ANCHORED(probe) ((probe)->dtpr_func[0] != '\0')
#define DTRACE_STATE_ALIGN 64
#define DTRACE_FLAGS2FLT(flags) \
(((flags) & CPU_DTRACE_BADADDR) ? DTRACEFLT_BADADDR : \
((flags) & CPU_DTRACE_ILLOP) ? DTRACEFLT_ILLOP : \
((flags) & CPU_DTRACE_DIVZERO) ? DTRACEFLT_DIVZERO : \
((flags) & CPU_DTRACE_KPRIV) ? DTRACEFLT_KPRIV : \
((flags) & CPU_DTRACE_UPRIV) ? DTRACEFLT_UPRIV : \
((flags) & CPU_DTRACE_TUPOFLOW) ? DTRACEFLT_TUPOFLOW : \
((flags) & CPU_DTRACE_BADALIGN) ? DTRACEFLT_BADALIGN : \
((flags) & CPU_DTRACE_NOSCRATCH) ? DTRACEFLT_NOSCRATCH : \
((flags) & CPU_DTRACE_BADSTACK) ? DTRACEFLT_BADSTACK : \
DTRACEFLT_UNKNOWN)
#define DTRACEACT_ISSTRING(act) \
((act)->dta_kind == DTRACEACT_DIFEXPR && \
(act)->dta_difo->dtdo_rtype.dtdt_kind == DIF_TYPE_STRING)
/* Function prototype definitions: */
static size_t dtrace_strlen(const char *, size_t);
static dtrace_probe_t *dtrace_probe_lookup_id(dtrace_id_t id);
static void dtrace_enabling_provide(dtrace_provider_t *);
static int dtrace_enabling_match(dtrace_enabling_t *, int *);
static void dtrace_enabling_matchall(void);
static void dtrace_enabling_matchall_task(void *);
static void dtrace_enabling_reap(void *);
static dtrace_state_t *dtrace_anon_grab(void);
static uint64_t dtrace_helper(int, dtrace_mstate_t *,
dtrace_state_t *, uint64_t, uint64_t);
static dtrace_helpers_t *dtrace_helpers_create(proc_t *);
static void dtrace_buffer_drop(dtrace_buffer_t *);
static int dtrace_buffer_consumed(dtrace_buffer_t *, hrtime_t when);
static ssize_t dtrace_buffer_reserve(dtrace_buffer_t *, size_t, size_t,
dtrace_state_t *, dtrace_mstate_t *);
static int dtrace_state_option(dtrace_state_t *, dtrace_optid_t,
dtrace_optval_t);
static int dtrace_ecb_create_enable(dtrace_probe_t *, void *);
static void dtrace_helper_provider_destroy(dtrace_helper_provider_t *);
uint16_t dtrace_load16(uintptr_t);
uint32_t dtrace_load32(uintptr_t);
uint64_t dtrace_load64(uintptr_t);
uint8_t dtrace_load8(uintptr_t);
void dtrace_dynvar_clean(dtrace_dstate_t *);
dtrace_dynvar_t *dtrace_dynvar(dtrace_dstate_t *, uint_t, dtrace_key_t *,
size_t, dtrace_dynvar_op_t, dtrace_mstate_t *, dtrace_vstate_t *);
uintptr_t dtrace_dif_varstr(uintptr_t, dtrace_state_t *, dtrace_mstate_t *);
static int dtrace_priv_proc(dtrace_state_t *);
static void dtrace_getf_barrier(void);
static int dtrace_canload_remains(uint64_t, size_t, size_t *,
dtrace_mstate_t *, dtrace_vstate_t *);
static int dtrace_canstore_remains(uint64_t, size_t, size_t *,
dtrace_mstate_t *, dtrace_vstate_t *);
/*
* DTrace Probe Context Functions
*
* These functions are called from probe context. Because probe context is
* any context in which C may be called, arbitrarily locks may be held,
* interrupts may be disabled, we may be in arbitrary dispatched state, etc.
* As a result, functions called from probe context may only call other DTrace
* support functions -- they may not interact at all with the system at large.
* (Note that the ASSERT macro is made probe-context safe by redefining it in
* terms of dtrace_assfail(), a probe-context safe function.) If arbitrary
* loads are to be performed from probe context, they _must_ be in terms of
* the safe dtrace_load*() variants.
*
* Some functions in this block are not actually called from probe context;
* for these functions, there will be a comment above the function reading
* "Note: not called from probe context."
*/
void
dtrace_panic(const char *format, ...)
{
va_list alist;
va_start(alist, format);
#ifdef __FreeBSD__
vpanic(format, alist);
#else
dtrace_vpanic(format, alist);
#endif
va_end(alist);
}
int
dtrace_assfail(const char *a, const char *f, int l)
{
dtrace_panic("assertion failed: %s, file: %s, line: %d", a, f, l);
/*
* We just need something here that even the most clever compiler
* cannot optimize away.
*/
return (a[(uintptr_t)f]);
}
/*
* Atomically increment a specified error counter from probe context.
*/
static void
dtrace_error(uint32_t *counter)
{
/*
* Most counters stored to in probe context are per-CPU counters.
* However, there are some error conditions that are sufficiently
* arcane that they don't merit per-CPU storage. If these counters
* are incremented concurrently on different CPUs, scalability will be
* adversely affected -- but we don't expect them to be white-hot in a
* correctly constructed enabling...
*/
uint32_t oval, nval;
do {
oval = *counter;
if ((nval = oval + 1) == 0) {
/*
* If the counter would wrap, set it to 1 -- assuring
* that the counter is never zero when we have seen
* errors. (The counter must be 32-bits because we
* aren't guaranteed a 64-bit compare&swap operation.)
* To save this code both the infamy of being fingered
* by a priggish news story and the indignity of being
* the target of a neo-puritan witch trial, we're
* carefully avoiding any colorful description of the
* likelihood of this condition -- but suffice it to
* say that it is only slightly more likely than the
* overflow of predicate cache IDs, as discussed in
* dtrace_predicate_create().
*/
nval = 1;
}
} while (dtrace_cas32(counter, oval, nval) != oval);
}
/*
* Use the DTRACE_LOADFUNC macro to define functions for each of loading a
* uint8_t, a uint16_t, a uint32_t and a uint64_t.
*/
/* BEGIN CSTYLED */
DTRACE_LOADFUNC(8)
DTRACE_LOADFUNC(16)
DTRACE_LOADFUNC(32)
DTRACE_LOADFUNC(64)
/* END CSTYLED */
static int
dtrace_inscratch(uintptr_t dest, size_t size, dtrace_mstate_t *mstate)
{
if (dest < mstate->dtms_scratch_base)
return (0);
if (dest + size < dest)
return (0);
if (dest + size > mstate->dtms_scratch_ptr)
return (0);
return (1);
}
static int
dtrace_canstore_statvar(uint64_t addr, size_t sz, size_t *remain,
dtrace_statvar_t **svars, int nsvars)
{
int i;
size_t maxglobalsize, maxlocalsize;
if (nsvars == 0)
return (0);
maxglobalsize = dtrace_statvar_maxsize + sizeof (uint64_t);
maxlocalsize = maxglobalsize * (mp_maxid + 1);
for (i = 0; i < nsvars; i++) {
dtrace_statvar_t *svar = svars[i];
uint8_t scope;
size_t size;
if (svar == NULL || (size = svar->dtsv_size) == 0)
continue;
scope = svar->dtsv_var.dtdv_scope;
/*
* We verify that our size is valid in the spirit of providing
* defense in depth: we want to prevent attackers from using
* DTrace to escalate an orthogonal kernel heap corruption bug
* into the ability to store to arbitrary locations in memory.
*/
VERIFY((scope == DIFV_SCOPE_GLOBAL && size <= maxglobalsize) ||
(scope == DIFV_SCOPE_LOCAL && size <= maxlocalsize));
if (DTRACE_INRANGE(addr, sz, svar->dtsv_data,
svar->dtsv_size)) {
DTRACE_RANGE_REMAIN(remain, addr, svar->dtsv_data,
svar->dtsv_size);
return (1);
}
}
return (0);
}
/*
* Check to see if the address is within a memory region to which a store may
* be issued. This includes the DTrace scratch areas, and any DTrace variable
* region. The caller of dtrace_canstore() is responsible for performing any
* alignment checks that are needed before stores are actually executed.
*/
static int
dtrace_canstore(uint64_t addr, size_t sz, dtrace_mstate_t *mstate,
dtrace_vstate_t *vstate)
{
return (dtrace_canstore_remains(addr, sz, NULL, mstate, vstate));
}
/*
* Implementation of dtrace_canstore which communicates the upper bound of the
* allowed memory region.
*/
static int
dtrace_canstore_remains(uint64_t addr, size_t sz, size_t *remain,
dtrace_mstate_t *mstate, dtrace_vstate_t *vstate)
{
/*
* First, check to see if the address is in scratch space...
*/
if (DTRACE_INRANGE(addr, sz, mstate->dtms_scratch_base,
mstate->dtms_scratch_size)) {
DTRACE_RANGE_REMAIN(remain, addr, mstate->dtms_scratch_base,
mstate->dtms_scratch_size);
return (1);
}
/*
* Now check to see if it's a dynamic variable. This check will pick
* up both thread-local variables and any global dynamically-allocated
* variables.
*/
if (DTRACE_INRANGE(addr, sz, vstate->dtvs_dynvars.dtds_base,
vstate->dtvs_dynvars.dtds_size)) {
dtrace_dstate_t *dstate = &vstate->dtvs_dynvars;
uintptr_t base = (uintptr_t)dstate->dtds_base +
(dstate->dtds_hashsize * sizeof (dtrace_dynhash_t));
uintptr_t chunkoffs;
dtrace_dynvar_t *dvar;
/*
* Before we assume that we can store here, we need to make
* sure that it isn't in our metadata -- storing to our
* dynamic variable metadata would corrupt our state. For
* the range to not include any dynamic variable metadata,
* it must:
*
* (1) Start above the hash table that is at the base of
* the dynamic variable space
*
* (2) Have a starting chunk offset that is beyond the
* dtrace_dynvar_t that is at the base of every chunk
*
* (3) Not span a chunk boundary
*
* (4) Not be in the tuple space of a dynamic variable
*
*/
if (addr < base)
return (0);
chunkoffs = (addr - base) % dstate->dtds_chunksize;
if (chunkoffs < sizeof (dtrace_dynvar_t))
return (0);
if (chunkoffs + sz > dstate->dtds_chunksize)
return (0);
dvar = (dtrace_dynvar_t *)((uintptr_t)addr - chunkoffs);
if (dvar->dtdv_hashval == DTRACE_DYNHASH_FREE)
return (0);
if (chunkoffs < sizeof (dtrace_dynvar_t) +
((dvar->dtdv_tuple.dtt_nkeys - 1) * sizeof (dtrace_key_t)))
return (0);
DTRACE_RANGE_REMAIN(remain, addr, dvar, dstate->dtds_chunksize);
return (1);
}
/*
* Finally, check the static local and global variables. These checks
* take the longest, so we perform them last.
*/
if (dtrace_canstore_statvar(addr, sz, remain,
vstate->dtvs_locals, vstate->dtvs_nlocals))
return (1);
if (dtrace_canstore_statvar(addr, sz, remain,
vstate->dtvs_globals, vstate->dtvs_nglobals))
return (1);
return (0);
}
/*
* Convenience routine to check to see if the address is within a memory
* region in which a load may be issued given the user's privilege level;
* if not, it sets the appropriate error flags and loads 'addr' into the
* illegal value slot.
*
* DTrace subroutines (DIF_SUBR_*) should use this helper to implement
* appropriate memory access protection.
*/
static int
dtrace_canload(uint64_t addr, size_t sz, dtrace_mstate_t *mstate,
dtrace_vstate_t *vstate)
{
return (dtrace_canload_remains(addr, sz, NULL, mstate, vstate));
}
/*
* Implementation of dtrace_canload which communicates the uppoer bound of the
* allowed memory region.
*/
static int
dtrace_canload_remains(uint64_t addr, size_t sz, size_t *remain,
dtrace_mstate_t *mstate, dtrace_vstate_t *vstate)
{
volatile uintptr_t *illval = &cpu_core[curcpu].cpuc_dtrace_illval;
file_t *fp;
/*
* If we hold the privilege to read from kernel memory, then
* everything is readable.
*/
if ((mstate->dtms_access & DTRACE_ACCESS_KERNEL) != 0) {
DTRACE_RANGE_REMAIN(remain, addr, addr, sz);
return (1);
}
/*
* You can obviously read that which you can store.
*/
if (dtrace_canstore_remains(addr, sz, remain, mstate, vstate))
return (1);
/*
* We're allowed to read from our own string table.
*/
if (DTRACE_INRANGE(addr, sz, mstate->dtms_difo->dtdo_strtab,
mstate->dtms_difo->dtdo_strlen)) {
DTRACE_RANGE_REMAIN(remain, addr,
mstate->dtms_difo->dtdo_strtab,
mstate->dtms_difo->dtdo_strlen);
return (1);
}
if (vstate->dtvs_state != NULL &&
dtrace_priv_proc(vstate->dtvs_state)) {
proc_t *p;
/*
* When we have privileges to the current process, there are
* several context-related kernel structures that are safe to
* read, even absent the privilege to read from kernel memory.
* These reads are safe because these structures contain only
* state that (1) we're permitted to read, (2) is harmless or
* (3) contains pointers to additional kernel state that we're
* not permitted to read (and as such, do not present an
* opportunity for privilege escalation). Finally (and
* critically), because of the nature of their relation with
* the current thread context, the memory associated with these
* structures cannot change over the duration of probe context,
* and it is therefore impossible for this memory to be
* deallocated and reallocated as something else while it's
* being operated upon.
*/
if (DTRACE_INRANGE(addr, sz, curthread, sizeof (kthread_t))) {
DTRACE_RANGE_REMAIN(remain, addr, curthread,
sizeof (kthread_t));
return (1);
}
if ((p = curthread->t_procp) != NULL && DTRACE_INRANGE(addr,
sz, curthread->t_procp, sizeof (proc_t))) {
DTRACE_RANGE_REMAIN(remain, addr, curthread->t_procp,
sizeof (proc_t));
return (1);
}
if (curthread->t_cred != NULL && DTRACE_INRANGE(addr, sz,
curthread->t_cred, sizeof (cred_t))) {
DTRACE_RANGE_REMAIN(remain, addr, curthread->t_cred,
sizeof (cred_t));
return (1);
}
#ifdef illumos
if (p != NULL && p->p_pidp != NULL && DTRACE_INRANGE(addr, sz,
&(p->p_pidp->pid_id), sizeof (pid_t))) {
DTRACE_RANGE_REMAIN(remain, addr, &(p->p_pidp->pid_id),
sizeof (pid_t));
return (1);
}
if (curthread->t_cpu != NULL && DTRACE_INRANGE(addr, sz,
curthread->t_cpu, offsetof(cpu_t, cpu_pause_thread))) {
DTRACE_RANGE_REMAIN(remain, addr, curthread->t_cpu,
offsetof(cpu_t, cpu_pause_thread));
return (1);
}
#endif
}
if ((fp = mstate->dtms_getf) != NULL) {
uintptr_t psz = sizeof (void *);
vnode_t *vp;
vnodeops_t *op;
/*
* When getf() returns a file_t, the enabling is implicitly
* granted the (transient) right to read the returned file_t
* as well as the v_path and v_op->vnop_name of the underlying
* vnode. These accesses are allowed after a successful
* getf() because the members that they refer to cannot change
* once set -- and the barrier logic in the kernel's closef()