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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
|
// Copyright 2018 The gVisor Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package kernel
import (
"fmt"
"sync/atomic"
"gvisor.dev/gvisor/pkg/abi/linux"
"gvisor.dev/gvisor/pkg/hostarch"
"gvisor.dev/gvisor/pkg/marshal/primitive"
"gvisor.dev/gvisor/pkg/sentry/arch"
"gvisor.dev/gvisor/pkg/sentry/mm"
"gvisor.dev/gvisor/pkg/syserror"
"gvisor.dev/gvisor/pkg/usermem"
)
// ptraceOptions are the subset of options controlling a task's ptrace behavior
// that are set by ptrace(PTRACE_SETOPTIONS).
//
// +stateify savable
type ptraceOptions struct {
// ExitKill is true if the tracee should be sent SIGKILL when the tracer
// exits.
ExitKill bool
// If SysGood is true, set bit 7 in the signal number for
// syscall-entry-stop and syscall-exit-stop traps delivered to this task's
// tracer.
SysGood bool
// TraceClone is true if the tracer wants to receive PTRACE_EVENT_CLONE
// events.
TraceClone bool
// TraceExec is true if the tracer wants to receive PTRACE_EVENT_EXEC
// events.
TraceExec bool
// TraceExit is true if the tracer wants to receive PTRACE_EVENT_EXIT
// events.
TraceExit bool
// TraceFork is true if the tracer wants to receive PTRACE_EVENT_FORK
// events.
TraceFork bool
// TraceSeccomp is true if the tracer wants to receive PTRACE_EVENT_SECCOMP
// events.
TraceSeccomp bool
// TraceVfork is true if the tracer wants to receive PTRACE_EVENT_VFORK
// events.
TraceVfork bool
// TraceVforkDone is true if the tracer wants to receive
// PTRACE_EVENT_VFORK_DONE events.
TraceVforkDone bool
}
// ptraceSyscallMode controls the behavior of a ptraced task at syscall entry
// and exit.
type ptraceSyscallMode int
const (
// ptraceSyscallNone indicates that the task has never ptrace-stopped, or
// that it was resumed from its last ptrace-stop by PTRACE_CONT or
// PTRACE_DETACH. The task's syscalls will not be intercepted.
ptraceSyscallNone ptraceSyscallMode = iota
// ptraceSyscallIntercept indicates that the task was resumed from its last
// ptrace-stop by PTRACE_SYSCALL. The next time the task enters or exits a
// syscall, a ptrace-stop will occur.
ptraceSyscallIntercept
// ptraceSyscallEmu indicates that the task was resumed from its last
// ptrace-stop by PTRACE_SYSEMU or PTRACE_SYSEMU_SINGLESTEP. The next time
// the task enters a syscall, the syscall will be skipped, and a
// ptrace-stop will occur.
ptraceSyscallEmu
)
// CanTrace checks that t is permitted to access target's state, as defined by
// ptrace(2), subsection "Ptrace access mode checking". If attach is true, it
// checks for access mode PTRACE_MODE_ATTACH; otherwise, it checks for access
// mode PTRACE_MODE_READ.
//
// In Linux, ptrace access restrictions may be configured by LSMs. While we do
// not support LSMs, we do add additional restrictions based on the commoncap
// and YAMA LSMs.
//
// TODO(gvisor.dev/issue/212): The result of CanTrace is immediately stale (e.g., a
// racing setuid(2) may change traceability). This may pose a risk when a task
// changes from traceable to not traceable. This is only problematic across
// execve, where privileges may increase.
//
// We currently do not implement privileged executables (set-user/group-ID bits
// and file capabilities), so that case is not reachable.
func (t *Task) CanTrace(target *Task, attach bool) bool {
// "If the calling thread and the target thread are in the same thread
// group, access is always allowed." - ptrace(2)
//
// Note: Strictly speaking, prior to 73af963f9f30 ("__ptrace_may_access()
// should not deny sub-threads", first released in Linux 3.12), the rule
// only applies if t and target are the same task. But, as that commit
// message puts it, "[any] security check is pointless when the tasks share
// the same ->mm."
if t.tg == target.tg {
return true
}
if !t.canTraceStandard(target, attach) {
return false
}
// YAMA only supported for vfs2.
if !VFS2Enabled {
return true
}
if atomic.LoadInt32(&t.k.YAMAPtraceScope) == linux.YAMA_SCOPE_RELATIONAL {
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if !t.canTraceYAMALocked(target) {
return false
}
}
return true
}
// canTraceLocked is the same as CanTrace, except the caller must already hold
// the TaskSet mutex (for reading or writing).
func (t *Task) canTraceLocked(target *Task, attach bool) bool {
if t.tg == target.tg {
return true
}
if !t.canTraceStandard(target, attach) {
return false
}
// YAMA only supported for vfs2.
if !VFS2Enabled {
return true
}
if atomic.LoadInt32(&t.k.YAMAPtraceScope) == linux.YAMA_SCOPE_RELATIONAL {
if !t.canTraceYAMALocked(target) {
return false
}
}
return true
}
// canTraceStandard performs standard ptrace access checks as defined by
// kernel/ptrace.c:__ptrace_may_access as well as the commoncap LSM
// implementation of the security_ptrace_access_check() interface, which is
// always invoked.
func (t *Task) canTraceStandard(target *Task, attach bool) bool {
// """
// TODO(gvisor.dev/issue/260): 1. If the access mode specifies
// PTRACE_MODE_FSCREDS (ED: snipped, doesn't exist until Linux 4.5).
//
// Otherwise, the access mode specifies PTRACE_MODE_REALCREDS, so use the
// caller's real UID and GID for the checks in the next step. (Most APIs
// that check the caller's UID and GID use the effective IDs. For
// historical reasons, the PTRACE_MODE_REALCREDS check uses the real IDs
// instead.)
//
// 2. Deny access if neither of the following is true:
//
// - The real, effective, and saved-set user IDs of the target match the
// caller's user ID, *and* the real, effective, and saved-set group IDs of
// the target match the caller's group ID.
//
// - The caller has the CAP_SYS_PTRACE capability in the user namespace of
// the target.
//
// 3. Deny access if the target process "dumpable" attribute has a value
// other than 1 (SUID_DUMP_USER; see the discussion of PR_SET_DUMPABLE in
// prctl(2)), and the caller does not have the CAP_SYS_PTRACE capability in
// the user namespace of the target process.
//
// 4. The commoncap LSM performs the following steps:
//
// a) If the access mode includes PTRACE_MODE_FSCREDS, then use the
// caller's effective capability set; otherwise (the access mode specifies
// PTRACE_MODE_REALCREDS, so) use the caller's permitted capability set.
//
// b) Deny access if neither of the following is true:
//
// - The caller and the target process are in the same user namespace, and
// the caller's capabilities are a proper superset of the target process's
// permitted capabilities.
//
// - The caller has the CAP_SYS_PTRACE capability in the target process's
// user namespace.
//
// Note that the commoncap LSM does not distinguish between
// PTRACE_MODE_READ and PTRACE_MODE_ATTACH. (ED: From earlier in this
// section: "the commoncap LSM ... is always invoked".)
// """
callerCreds := t.Credentials()
targetCreds := target.Credentials()
if callerCreds.HasCapabilityIn(linux.CAP_SYS_PTRACE, targetCreds.UserNamespace) {
return true
}
if cuid := callerCreds.RealKUID; cuid != targetCreds.RealKUID || cuid != targetCreds.EffectiveKUID || cuid != targetCreds.SavedKUID {
return false
}
if cgid := callerCreds.RealKGID; cgid != targetCreds.RealKGID || cgid != targetCreds.EffectiveKGID || cgid != targetCreds.SavedKGID {
return false
}
var targetMM *mm.MemoryManager
target.WithMuLocked(func(t *Task) {
targetMM = t.MemoryManager()
})
if targetMM != nil && targetMM.Dumpability() != mm.UserDumpable {
return false
}
if callerCreds.UserNamespace != targetCreds.UserNamespace {
return false
}
if targetCreds.PermittedCaps&^callerCreds.PermittedCaps != 0 {
return false
}
return true
}
// canTraceYAMALocked performs ptrace access checks as defined by the YAMA LSM
// implementation of the security_ptrace_access_check() interface, with YAMA
// configured to mode 1. This is a common default among various Linux
// distributions.
//
// It only permits the tracer to proceed if one of the following conditions is
// met:
//
// a) The tracer is already attached to the tracee.
//
// b) The target is a descendant of the tracer.
//
// c) The target has explicitly given permission to the tracer through the
// PR_SET_PTRACER prctl.
//
// d) The tracer has CAP_SYS_PTRACE.
//
// See security/yama/yama_lsm.c:yama_ptrace_access_check.
//
// Precondition: the TaskSet mutex must be locked (for reading or writing).
func (t *Task) canTraceYAMALocked(target *Task) bool {
if tracer := target.Tracer(); tracer != nil {
if tracer.tg == t.tg {
return true
}
}
if target.isYAMADescendantOfLocked(t) {
return true
}
if target.hasYAMAExceptionForLocked(t) {
return true
}
if t.HasCapabilityIn(linux.CAP_SYS_PTRACE, target.UserNamespace()) {
return true
}
return false
}
// Determines whether t is considered a descendant of ancestor for the purposes
// of YAMA permissions (specifically, whether t's thread group is descended from
// ancestor's).
//
// Precondition: the TaskSet mutex must be locked (for reading or writing).
func (t *Task) isYAMADescendantOfLocked(ancestor *Task) bool {
walker := t
for walker != nil {
if walker.tg.leader == ancestor.tg.leader {
return true
}
walker = walker.parent
}
return false
}
// Precondition: the TaskSet mutex must be locked (for reading or writing).
func (t *Task) hasYAMAExceptionForLocked(tracer *Task) bool {
allowed, ok := t.k.ptraceExceptions[t]
if !ok {
return false
}
return allowed == nil || tracer.isYAMADescendantOfLocked(allowed)
}
// ClearYAMAException removes any YAMA exception with t as the tracee.
func (t *Task) ClearYAMAException() {
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
tracee := t.tg.leader
delete(t.k.ptraceExceptions, tracee)
}
// SetYAMAException creates a YAMA exception allowing all descendants of tracer
// to trace t. If tracer is nil, then any task is allowed to trace t.
//
// If there was an existing exception, it is overwritten with the new one.
func (t *Task) SetYAMAException(tracer *Task) {
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
tracee := t.tg.leader
tracee.ptraceYAMAExceptionAdded = true
if tracer != nil {
tracer.ptraceYAMAExceptionAdded = true
}
t.k.ptraceExceptions[tracee] = tracer
}
// Tracer returns t's ptrace Tracer.
func (t *Task) Tracer() *Task {
return t.ptraceTracer.Load().(*Task)
}
// hasTracer returns true if t has a ptrace tracer attached.
func (t *Task) hasTracer() bool {
// This isn't just inlined into callers so that if Task.Tracer() turns out
// to be too expensive because of e.g. interface conversion, we can switch
// to having a separate atomic flag more easily.
return t.Tracer() != nil
}
// ptraceStop is a TaskStop placed on tasks in a ptrace-stop.
//
// +stateify savable
type ptraceStop struct {
// If frozen is true, the stopped task's tracer is currently operating on
// it, so Task.Kill should not remove the stop.
frozen bool
// If listen is true, the stopped task's tracer invoked PTRACE_LISTEN, so
// ptraceFreeze should fail.
listen bool
}
// Killable implements TaskStop.Killable.
func (s *ptraceStop) Killable() bool {
return !s.frozen
}
// beginPtraceStopLocked initiates an unfrozen ptrace-stop on t. If t has been
// killed, the stop is skipped, and beginPtraceStopLocked returns false.
//
// beginPtraceStopLocked does not signal t's tracer or wake it if it is
// waiting.
//
// Preconditions:
// * The TaskSet mutex must be locked.
// * The caller must be running on the task goroutine.
func (t *Task) beginPtraceStopLocked() bool {
t.tg.signalHandlers.mu.Lock()
defer t.tg.signalHandlers.mu.Unlock()
// This is analogous to Linux's kernel/signal.c:ptrace_stop() => ... =>
// kernel/sched/core.c:__schedule() => signal_pending_state() check, which
// is what prevents tasks from entering ptrace-stops after being killed.
// Note that if t was SIGKILLed and beingPtraceStopLocked is being called
// for PTRACE_EVENT_EXIT, the task will have dequeued the signal before
// entering the exit path, so t.killedLocked() will no longer return true.
// This is consistent with Linux: "Bugs: ... A SIGKILL signal may still
// cause a PTRACE_EVENT_EXIT stop before actual signal death. This may be
// changed in the future; SIGKILL is meant to always immediately kill tasks
// even under ptrace. Last confirmed on Linux 3.13." - ptrace(2)
if t.killedLocked() {
return false
}
t.beginInternalStopLocked(&ptraceStop{})
return true
}
// Preconditions: The TaskSet mutex must be locked.
func (t *Task) ptraceTrapLocked(code int32) {
// This is unconditional in ptrace_stop().
t.tg.signalHandlers.mu.Lock()
t.trapStopPending = false
t.tg.signalHandlers.mu.Unlock()
t.ptraceCode = code
t.ptraceSiginfo = &arch.SignalInfo{
Signo: int32(linux.SIGTRAP),
Code: code,
}
t.ptraceSiginfo.SetPID(int32(t.tg.pidns.tids[t]))
t.ptraceSiginfo.SetUID(int32(t.Credentials().RealKUID.In(t.UserNamespace()).OrOverflow()))
if t.beginPtraceStopLocked() {
tracer := t.Tracer()
tracer.signalStop(t, arch.CLD_TRAPPED, int32(linux.SIGTRAP))
tracer.tg.eventQueue.Notify(EventTraceeStop)
}
}
// ptraceFreeze checks if t is in a ptraceStop. If so, it freezes the
// ptraceStop, temporarily preventing it from being removed by a concurrent
// Task.Kill, and returns true. Otherwise it returns false.
//
// Preconditions:
// * The TaskSet mutex must be locked.
// * The caller must be running on the task goroutine of t's tracer.
func (t *Task) ptraceFreeze() bool {
t.tg.signalHandlers.mu.Lock()
defer t.tg.signalHandlers.mu.Unlock()
if t.stop == nil {
return false
}
s, ok := t.stop.(*ptraceStop)
if !ok {
return false
}
if s.listen {
return false
}
s.frozen = true
return true
}
// ptraceUnfreeze ends the effect of a previous successful call to
// ptraceFreeze.
//
// Preconditions: t must be in a frozen ptraceStop.
func (t *Task) ptraceUnfreeze() {
// t.tg.signalHandlers is stable because t is in a frozen ptrace-stop,
// preventing its thread group from completing execve.
t.tg.signalHandlers.mu.Lock()
defer t.tg.signalHandlers.mu.Unlock()
t.ptraceUnfreezeLocked()
}
// Preconditions:
// * t must be in a frozen ptraceStop.
// * t's signal mutex must be locked.
func (t *Task) ptraceUnfreezeLocked() {
// Do this even if the task has been killed to ensure a panic if t.stop is
// nil or not a ptraceStop.
t.stop.(*ptraceStop).frozen = false
if t.killedLocked() {
t.endInternalStopLocked()
}
}
// ptraceUnstop implements ptrace request PTRACE_CONT, PTRACE_SYSCALL,
// PTRACE_SINGLESTEP, PTRACE_SYSEMU, or PTRACE_SYSEMU_SINGLESTEP depending on
// mode and singlestep.
//
// Preconditions: t must be in a frozen ptrace stop.
//
// Postconditions: If ptraceUnstop returns nil, t will no longer be in a ptrace
// stop.
func (t *Task) ptraceUnstop(mode ptraceSyscallMode, singlestep bool, sig linux.Signal) error {
if sig != 0 && !sig.IsValid() {
return syserror.EIO
}
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
t.ptraceCode = int32(sig)
t.ptraceSyscallMode = mode
t.ptraceSinglestep = singlestep
t.tg.signalHandlers.mu.Lock()
defer t.tg.signalHandlers.mu.Unlock()
t.endInternalStopLocked()
return nil
}
func (t *Task) ptraceTraceme() error {
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
if t.hasTracer() {
return syserror.EPERM
}
if t.parent == nil {
// In Linux, only init can not have a parent, and init is assumed never
// to invoke PTRACE_TRACEME. In the sentry, TGID 1 is an arbitrary user
// application that may invoke PTRACE_TRACEME; having no parent can
// also occur if all tasks in the parent thread group have exited, and
// failed to find a living thread group to reparent to. The former case
// is treated as if TGID 1 has an exited parent in an invisible
// ancestor PID namespace that is an owner of the root user namespace
// (and consequently has CAP_SYS_PTRACE), and the latter case is a
// special form of the exited parent case below. In either case,
// returning nil here is correct.
return nil
}
if !t.parent.canTraceLocked(t, true) {
return syserror.EPERM
}
if t.parent.exitState != TaskExitNone {
// Fail silently, as if we were successfully attached but then
// immediately detached. This is consistent with Linux.
return nil
}
t.ptraceTracer.Store(t.parent)
t.parent.ptraceTracees[t] = struct{}{}
return nil
}
// ptraceAttach implements ptrace(PTRACE_ATTACH, target) if seize is false, and
// ptrace(PTRACE_SEIZE, target, 0, opts) if seize is true. t is the caller.
func (t *Task) ptraceAttach(target *Task, seize bool, opts uintptr) error {
if t.tg == target.tg {
return syserror.EPERM
}
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
if !t.canTraceLocked(target, true) {
return syserror.EPERM
}
if target.hasTracer() {
return syserror.EPERM
}
// Attaching to zombies and dead tasks is not permitted; the exit
// notification logic relies on this. Linux allows attaching to PF_EXITING
// tasks, though.
if target.exitState >= TaskExitZombie {
return syserror.EPERM
}
if seize {
if err := target.ptraceSetOptionsLocked(opts); err != nil {
return syserror.EIO
}
}
target.ptraceTracer.Store(t)
t.ptraceTracees[target] = struct{}{}
target.ptraceSeized = seize
target.tg.signalHandlers.mu.Lock()
// "Unlike PTRACE_ATTACH, PTRACE_SEIZE does not stop the process." -
// ptrace(2)
if !seize {
target.sendSignalLocked(&arch.SignalInfo{
Signo: int32(linux.SIGSTOP),
Code: arch.SignalInfoUser,
}, false /* group */)
}
// Undocumented Linux feature: If the tracee is already group-stopped (and
// consequently will not report the SIGSTOP just sent), force it to leave
// and re-enter the stop so that it will switch to a ptrace-stop.
if target.stop == (*groupStop)(nil) {
target.trapStopPending = true
target.endInternalStopLocked()
// TODO(jamieliu): Linux blocks ptrace_attach() until the task has
// entered the ptrace-stop (or exited) via JOBCTL_TRAPPING.
}
target.tg.signalHandlers.mu.Unlock()
return nil
}
// ptraceDetach implements ptrace(PTRACE_DETACH, target, 0, sig). t is the
// caller.
//
// Preconditions: target must be a tracee of t in a frozen ptrace stop.
//
// Postconditions: If ptraceDetach returns nil, target will no longer be in a
// ptrace stop.
func (t *Task) ptraceDetach(target *Task, sig linux.Signal) error {
if sig != 0 && !sig.IsValid() {
return syserror.EIO
}
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
target.ptraceCode = int32(sig)
target.forgetTracerLocked()
delete(t.ptraceTracees, target)
return nil
}
// exitPtrace is called in the exit path to detach all of t's tracees.
func (t *Task) exitPtrace() {
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
for target := range t.ptraceTracees {
if target.ptraceOpts.ExitKill {
target.tg.signalHandlers.mu.Lock()
target.sendSignalLocked(&arch.SignalInfo{
Signo: int32(linux.SIGKILL),
}, false /* group */)
target.tg.signalHandlers.mu.Unlock()
}
// Leave ptraceCode unchanged so that if the task is ptrace-stopped, it
// observes the ptraceCode it set before it entered the stop. I believe
// this is consistent with Linux.
target.forgetTracerLocked()
}
// "nil maps cannot be saved"
t.ptraceTracees = make(map[*Task]struct{})
if t.ptraceYAMAExceptionAdded {
delete(t.k.ptraceExceptions, t)
for tracee, tracer := range t.k.ptraceExceptions {
if tracer == t {
delete(t.k.ptraceExceptions, tracee)
}
}
}
}
// forgetTracerLocked detaches t's tracer and ensures that t is no longer
// ptrace-stopped.
//
// Preconditions: The TaskSet mutex must be locked for writing.
func (t *Task) forgetTracerLocked() {
t.ptraceSeized = false
t.ptraceOpts = ptraceOptions{}
t.ptraceSyscallMode = ptraceSyscallNone
t.ptraceSinglestep = false
t.ptraceTracer.Store((*Task)(nil))
if t.exitTracerNotified && !t.exitTracerAcked {
t.exitTracerAcked = true
t.exitNotifyLocked(true)
}
t.tg.signalHandlers.mu.Lock()
defer t.tg.signalHandlers.mu.Unlock()
// Unset t.trapStopPending, which might have been set by PTRACE_INTERRUPT. If
// it wasn't, it will be reset via t.groupStopPending after the following.
t.trapStopPending = false
// If t's thread group is in a group stop and t is eligible to participate,
// make it do so. This is essentially the reverse of the special case in
// ptraceAttach, which converts a group stop to a ptrace stop. ("Handling
// of restart from group-stop is currently buggy, but the "as planned"
// behavior is to leave tracee stopped and waiting for SIGCONT." -
// ptrace(2))
if (t.tg.groupStopComplete || t.tg.groupStopPendingCount != 0) && !t.groupStopPending && t.exitState < TaskExitInitiated {
t.groupStopPending = true
// t already participated in the group stop when it unset
// groupStopPending.
t.groupStopAcknowledged = true
t.interrupt()
}
if _, ok := t.stop.(*ptraceStop); ok {
t.endInternalStopLocked()
}
}
// ptraceSignalLocked is called after signal dequeueing to check if t should
// enter ptrace signal-delivery-stop.
//
// Preconditions:
// * The signal mutex must be locked.
// * The caller must be running on the task goroutine.
func (t *Task) ptraceSignalLocked(info *arch.SignalInfo) bool {
if linux.Signal(info.Signo) == linux.SIGKILL {
return false
}
if !t.hasTracer() {
return false
}
// The tracer might change this signal into a stop signal, in which case
// any SIGCONT received after the signal was originally dequeued should
// cancel it. This is consistent with Linux.
t.tg.groupStopDequeued = true
// This is unconditional in ptrace_stop().
t.trapStopPending = false
// Can't lock the TaskSet mutex while holding a signal mutex.
t.tg.signalHandlers.mu.Unlock()
defer t.tg.signalHandlers.mu.Lock()
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
tracer := t.Tracer()
if tracer == nil {
return false
}
t.ptraceCode = info.Signo
t.ptraceSiginfo = info
t.Debugf("Entering signal-delivery-stop for signal %d", info.Signo)
if t.beginPtraceStopLocked() {
tracer.signalStop(t, arch.CLD_TRAPPED, info.Signo)
tracer.tg.eventQueue.Notify(EventTraceeStop)
}
return true
}
// ptraceSeccomp is called when a seccomp-bpf filter returns action
// SECCOMP_RET_TRACE to check if t should enter PTRACE_EVENT_SECCOMP stop. data
// is the lower 16 bits of the filter's return value.
func (t *Task) ptraceSeccomp(data uint16) bool {
if !t.hasTracer() {
return false
}
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if !t.ptraceOpts.TraceSeccomp {
return false
}
t.Debugf("Entering PTRACE_EVENT_SECCOMP stop")
t.ptraceEventLocked(linux.PTRACE_EVENT_SECCOMP, uint64(data))
return true
}
// ptraceSyscallEnter is called immediately before entering a syscall to check
// if t should enter ptrace syscall-enter-stop.
func (t *Task) ptraceSyscallEnter() (taskRunState, bool) {
if !t.hasTracer() {
return nil, false
}
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
switch t.ptraceSyscallMode {
case ptraceSyscallNone:
return nil, false
case ptraceSyscallIntercept:
t.Debugf("Entering syscall-enter-stop from PTRACE_SYSCALL")
t.ptraceSyscallStopLocked()
return (*runSyscallAfterSyscallEnterStop)(nil), true
case ptraceSyscallEmu:
t.Debugf("Entering syscall-enter-stop from PTRACE_SYSEMU")
t.ptraceSyscallStopLocked()
return (*runSyscallAfterSysemuStop)(nil), true
}
panic(fmt.Sprintf("Unknown ptraceSyscallMode: %v", t.ptraceSyscallMode))
}
// ptraceSyscallExit is called immediately after leaving a syscall to check if
// t should enter ptrace syscall-exit-stop.
func (t *Task) ptraceSyscallExit() {
if !t.hasTracer() {
return
}
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if t.ptraceSyscallMode != ptraceSyscallIntercept {
return
}
t.Debugf("Entering syscall-exit-stop")
t.ptraceSyscallStopLocked()
}
// Preconditions: The TaskSet mutex must be locked.
func (t *Task) ptraceSyscallStopLocked() {
code := int32(linux.SIGTRAP)
if t.ptraceOpts.SysGood {
code |= 0x80
}
t.ptraceTrapLocked(code)
}
type ptraceCloneKind int32
const (
// ptraceCloneKindClone represents a call to Task.Clone where
// TerminationSignal is not SIGCHLD and Vfork is false.
ptraceCloneKindClone ptraceCloneKind = iota
// ptraceCloneKindFork represents a call to Task.Clone where
// TerminationSignal is SIGCHLD and Vfork is false.
ptraceCloneKindFork
// ptraceCloneKindVfork represents a call to Task.Clone where Vfork is
// true.
ptraceCloneKindVfork
)
// ptraceClone is called at the end of a clone or fork syscall to check if t
// should enter PTRACE_EVENT_CLONE, PTRACE_EVENT_FORK, or PTRACE_EVENT_VFORK
// stop. child is the new task.
func (t *Task) ptraceClone(kind ptraceCloneKind, child *Task, opts *CloneOptions) bool {
if !t.hasTracer() {
return false
}
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
event := false
if !opts.Untraced {
switch kind {
case ptraceCloneKindClone:
if t.ptraceOpts.TraceClone {
t.Debugf("Entering PTRACE_EVENT_CLONE stop")
t.ptraceEventLocked(linux.PTRACE_EVENT_CLONE, uint64(t.tg.pidns.tids[child]))
event = true
}
case ptraceCloneKindFork:
if t.ptraceOpts.TraceFork {
t.Debugf("Entering PTRACE_EVENT_FORK stop")
t.ptraceEventLocked(linux.PTRACE_EVENT_FORK, uint64(t.tg.pidns.tids[child]))
event = true
}
case ptraceCloneKindVfork:
if t.ptraceOpts.TraceVfork {
t.Debugf("Entering PTRACE_EVENT_VFORK stop")
t.ptraceEventLocked(linux.PTRACE_EVENT_VFORK, uint64(t.tg.pidns.tids[child]))
event = true
}
default:
panic(fmt.Sprintf("Unknown ptraceCloneKind: %v", kind))
}
}
// "If the PTRACE_O_TRACEFORK, PTRACE_O_TRACEVFORK, or PTRACE_O_TRACECLONE
// options are in effect, then children created by, respectively, vfork(2)
// or clone(2) with the CLONE_VFORK flag, fork(2) or clone(2) with the exit
// signal set to SIGCHLD, and other kinds of clone(2), are automatically
// attached to the same tracer which traced their parent. SIGSTOP is
// delivered to the children, causing them to enter signal-delivery-stop
// after they exit the system call which created them." - ptrace(2)
//
// clone(2)'s documentation of CLONE_UNTRACED and CLONE_PTRACE is
// confusingly wrong; see kernel/fork.c:_do_fork() => copy_process() =>
// include/linux/ptrace.h:ptrace_init_task().
if event || opts.InheritTracer {
tracer := t.Tracer()
if tracer != nil {
child.ptraceTracer.Store(tracer)
tracer.ptraceTracees[child] = struct{}{}
// "The "seized" behavior ... is inherited by children that are
// automatically attached using PTRACE_O_TRACEFORK,
// PTRACE_O_TRACEVFORK, and PTRACE_O_TRACECLONE." - ptrace(2)
child.ptraceSeized = t.ptraceSeized
// "Flags are inherited by new tracees created and "auto-attached"
// via active PTRACE_O_TRACEFORK, PTRACE_O_TRACEVFORK, or
// PTRACE_O_TRACECLONE options." - ptrace(2)
child.ptraceOpts = t.ptraceOpts
child.tg.signalHandlers.mu.Lock()
// "PTRACE_SEIZE: ... Automatically attached children stop with
// PTRACE_EVENT_STOP and WSTOPSIG(status) returns SIGTRAP instead
// of having SIGSTOP signal delivered to them." - ptrace(2)
if child.ptraceSeized {
child.trapStopPending = true
} else {
child.pendingSignals.enqueue(&arch.SignalInfo{
Signo: int32(linux.SIGSTOP),
}, nil)
}
// The child will self-interrupt() when its task goroutine starts
// running, so we don't have to.
child.tg.signalHandlers.mu.Unlock()
}
}
return event
}
// ptraceVforkDone is called after the end of a vfork stop to check if t should
// enter PTRACE_EVENT_VFORK_DONE stop. child is the new task's thread ID in t's
// PID namespace.
func (t *Task) ptraceVforkDone(child ThreadID) bool {
if !t.hasTracer() {
return false
}
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if !t.ptraceOpts.TraceVforkDone {
return false
}
t.Debugf("Entering PTRACE_EVENT_VFORK_DONE stop")
t.ptraceEventLocked(linux.PTRACE_EVENT_VFORK_DONE, uint64(child))
return true
}
// ptraceExec is called at the end of an execve syscall to check if t should
// enter PTRACE_EVENT_EXEC stop. oldTID is t's thread ID, in its *tracer's* PID
// namespace, prior to the execve. (If t did not have a tracer at the time
// oldTID was read, oldTID may be 0. This is consistent with Linux.)
func (t *Task) ptraceExec(oldTID ThreadID) {
if !t.hasTracer() {
return
}
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
// Recheck with the TaskSet mutex locked. Most ptrace points don't need to
// do this because detaching resets ptrace options, but PTRACE_EVENT_EXEC
// is special because both TraceExec and !TraceExec do something if a
// tracer is attached.
if !t.hasTracer() {
return
}
if t.ptraceOpts.TraceExec {
t.Debugf("Entering PTRACE_EVENT_EXEC stop")
t.ptraceEventLocked(linux.PTRACE_EVENT_EXEC, uint64(oldTID))
return
}
// "If the PTRACE_O_TRACEEXEC option is not in effect for the execing
// tracee, and if the tracee was PTRACE_ATTACHed rather that [sic]
// PTRACE_SEIZEd, the kernel delivers an extra SIGTRAP to the tracee after
// execve(2) returns. This is an ordinary signal (similar to one which can
// be generated by `kill -TRAP`, not a special kind of ptrace-stop.
// Employing PTRACE_GETSIGINFO for this signal returns si_code set to 0
// (SI_USER). This signal may be blocked by signal mask, and thus may be
// delivered (much) later." - ptrace(2)
if t.ptraceSeized {
return
}
t.tg.signalHandlers.mu.Lock()
defer t.tg.signalHandlers.mu.Unlock()
t.sendSignalLocked(&arch.SignalInfo{
Signo: int32(linux.SIGTRAP),
Code: arch.SignalInfoUser,
}, false /* group */)
}
// ptraceExit is called early in the task exit path to check if t should enter
// PTRACE_EVENT_EXIT stop.
func (t *Task) ptraceExit() {
if !t.hasTracer() {
return
}
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if !t.ptraceOpts.TraceExit {
return
}
t.tg.signalHandlers.mu.Lock()
status := t.exitStatus.Status()
t.tg.signalHandlers.mu.Unlock()
t.Debugf("Entering PTRACE_EVENT_EXIT stop")
t.ptraceEventLocked(linux.PTRACE_EVENT_EXIT, uint64(status))
}
// Preconditions: The TaskSet mutex must be locked.
func (t *Task) ptraceEventLocked(event int32, msg uint64) {
t.ptraceEventMsg = msg
// """
// PTRACE_EVENT stops are observed by the tracer as waitpid(2) returning
// with WIFSTOPPED(status), and WSTOPSIG(status) returns SIGTRAP. An
// additional bit is set in the higher byte of the status word: the value
// status>>8 will be
//
// (SIGTRAP | PTRACE_EVENT_foo << 8).
//
// ...
//
// """ - ptrace(2)
t.ptraceTrapLocked(int32(linux.SIGTRAP) | (event << 8))
}
// ptraceKill implements ptrace(PTRACE_KILL, target). t is the caller.
func (t *Task) ptraceKill(target *Task) error {
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
if target.Tracer() != t {
return syserror.ESRCH
}
target.tg.signalHandlers.mu.Lock()
defer target.tg.signalHandlers.mu.Unlock()
// "This operation is deprecated; do not use it! Instead, send a SIGKILL
// directly using kill(2) or tgkill(2). The problem with PTRACE_KILL is
// that it requires the tracee to be in signal-delivery-stop, otherwise it
// may not work (i.e., may complete successfully but won't kill the
// tracee)." - ptrace(2)
if target.stop == nil {
return nil
}
if _, ok := target.stop.(*ptraceStop); !ok {
return nil
}
target.ptraceCode = int32(linux.SIGKILL)
target.endInternalStopLocked()
return nil
}
func (t *Task) ptraceInterrupt(target *Task) error {
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
if target.Tracer() != t {
return syserror.ESRCH
}
if !target.ptraceSeized {
return syserror.EIO
}
target.tg.signalHandlers.mu.Lock()
defer target.tg.signalHandlers.mu.Unlock()
if target.killedLocked() || target.exitState >= TaskExitInitiated {
return nil
}
target.trapStopPending = true
if s, ok := target.stop.(*ptraceStop); ok && s.listen {
target.endInternalStopLocked()
}
target.interrupt()
return nil
}
// Preconditions:
// * The TaskSet mutex must be locked for writing.
// * t must have a tracer.
func (t *Task) ptraceSetOptionsLocked(opts uintptr) error {
const valid = uintptr(linux.PTRACE_O_EXITKILL |
linux.PTRACE_O_TRACESYSGOOD |
linux.PTRACE_O_TRACECLONE |
linux.PTRACE_O_TRACEEXEC |
linux.PTRACE_O_TRACEEXIT |
linux.PTRACE_O_TRACEFORK |
linux.PTRACE_O_TRACESECCOMP |
linux.PTRACE_O_TRACEVFORK |
linux.PTRACE_O_TRACEVFORKDONE)
if opts&^valid != 0 {
return syserror.EINVAL
}
t.ptraceOpts = ptraceOptions{
ExitKill: opts&linux.PTRACE_O_EXITKILL != 0,
SysGood: opts&linux.PTRACE_O_TRACESYSGOOD != 0,
TraceClone: opts&linux.PTRACE_O_TRACECLONE != 0,
TraceExec: opts&linux.PTRACE_O_TRACEEXEC != 0,
TraceExit: opts&linux.PTRACE_O_TRACEEXIT != 0,
TraceFork: opts&linux.PTRACE_O_TRACEFORK != 0,
TraceSeccomp: opts&linux.PTRACE_O_TRACESECCOMP != 0,
TraceVfork: opts&linux.PTRACE_O_TRACEVFORK != 0,
TraceVforkDone: opts&linux.PTRACE_O_TRACEVFORKDONE != 0,
}
return nil
}
// Ptrace implements the ptrace system call.
func (t *Task) Ptrace(req int64, pid ThreadID, addr, data hostarch.Addr) error {
// PTRACE_TRACEME ignores all other arguments.
if req == linux.PTRACE_TRACEME {
return t.ptraceTraceme()
}
// All other ptrace requests operate on a current or future tracee
// specified by pid.
target := t.tg.pidns.TaskWithID(pid)
if target == nil {
return syserror.ESRCH
}
// PTRACE_ATTACH and PTRACE_SEIZE do not require that target is not already
// a tracee.
if req == linux.PTRACE_ATTACH || req == linux.PTRACE_SEIZE {
seize := req == linux.PTRACE_SEIZE
if seize && addr != 0 {
return syserror.EIO
}
return t.ptraceAttach(target, seize, uintptr(data))
}
// PTRACE_KILL and PTRACE_INTERRUPT require that the target is a tracee,
// but does not require that it is ptrace-stopped.
if req == linux.PTRACE_KILL {
return t.ptraceKill(target)
}
if req == linux.PTRACE_INTERRUPT {
return t.ptraceInterrupt(target)
}
// All other ptrace requests require that the target is a ptrace-stopped
// tracee, and freeze the ptrace-stop so the tracee can be operated on.
t.tg.pidns.owner.mu.RLock()
if target.Tracer() != t {
t.tg.pidns.owner.mu.RUnlock()
return syserror.ESRCH
}
if !target.ptraceFreeze() {
t.tg.pidns.owner.mu.RUnlock()
// "Most ptrace commands (all except PTRACE_ATTACH, PTRACE_SEIZE,
// PTRACE_TRACEME, PTRACE_INTERRUPT, and PTRACE_KILL) require the
// tracee to be in a ptrace-stop, otherwise they fail with ESRCH." -
// ptrace(2)
return syserror.ESRCH
}
t.tg.pidns.owner.mu.RUnlock()
// Even if the target has a ptrace-stop active, the tracee's task goroutine
// may not yet have reached Task.doStop; wait for it to do so. This is safe
// because there's no way for target to initiate a ptrace-stop and then
// block (by calling Task.block) before entering it.
//
// Caveat: If tasks were just restored, the tracee's first call to
// Task.Activate (in Task.run) occurs before its first call to Task.doStop,
// which may block if the tracer's address space is active.
t.UninterruptibleSleepStart(true)
target.waitGoroutineStoppedOrExited()
t.UninterruptibleSleepFinish(true)
// Resuming commands end the ptrace stop, but only if successful.
// PTRACE_LISTEN ends the ptrace stop if trapNotifyPending is already set on the
// target.
switch req {
case linux.PTRACE_DETACH:
if err := t.ptraceDetach(target, linux.Signal(data)); err != nil {
target.ptraceUnfreeze()
return err
}
return nil
case linux.PTRACE_CONT:
if err := target.ptraceUnstop(ptraceSyscallNone, false, linux.Signal(data)); err != nil {
target.ptraceUnfreeze()
return err
}
return nil
case linux.PTRACE_SYSCALL:
if err := target.ptraceUnstop(ptraceSyscallIntercept, false, linux.Signal(data)); err != nil {
target.ptraceUnfreeze()
return err
}
return nil
case linux.PTRACE_SINGLESTEP:
if err := target.ptraceUnstop(ptraceSyscallNone, true, linux.Signal(data)); err != nil {
target.ptraceUnfreeze()
return err
}
return nil
case linux.PTRACE_SYSEMU:
if err := target.ptraceUnstop(ptraceSyscallEmu, false, linux.Signal(data)); err != nil {
target.ptraceUnfreeze()
return err
}
return nil
case linux.PTRACE_SYSEMU_SINGLESTEP:
if err := target.ptraceUnstop(ptraceSyscallEmu, true, linux.Signal(data)); err != nil {
target.ptraceUnfreeze()
return err
}
return nil
case linux.PTRACE_LISTEN:
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if !target.ptraceSeized {
return syserror.EIO
}
if target.ptraceSiginfo == nil {
return syserror.EIO
}
if target.ptraceSiginfo.Code>>8 != linux.PTRACE_EVENT_STOP {
return syserror.EIO
}
target.tg.signalHandlers.mu.Lock()
defer target.tg.signalHandlers.mu.Unlock()
if target.trapNotifyPending {
target.endInternalStopLocked()
} else {
target.stop.(*ptraceStop).listen = true
target.ptraceUnfreezeLocked()
}
return nil
}
// All other ptrace requests expect us to unfreeze the stop.
defer target.ptraceUnfreeze()
switch req {
case linux.PTRACE_PEEKTEXT, linux.PTRACE_PEEKDATA:
// "At the system call level, the PTRACE_PEEKTEXT, PTRACE_PEEKDATA, and
// PTRACE_PEEKUSER requests have a different API: they store the result
// at the address specified by the data parameter, and the return value
// is the error flag." - ptrace(2)
word := t.Arch().Native(0)
if _, err := word.CopyIn(target.CopyContext(t, usermem.IOOpts{IgnorePermissions: true}), addr); err != nil {
return err
}
_, err := word.CopyOut(t, data)
return err
case linux.PTRACE_POKETEXT, linux.PTRACE_POKEDATA:
word := t.Arch().Native(uintptr(data))
_, err := word.CopyOut(target.CopyContext(t, usermem.IOOpts{IgnorePermissions: true}), addr)
return err
case linux.PTRACE_GETREGSET:
// "Read the tracee's registers. addr specifies, in an
// architecture-dependent way, the type of registers to be read. ...
// data points to a struct iovec, which describes the destination
// buffer's location and length. On return, the kernel modifies iov.len
// to indicate the actual number of bytes returned." - ptrace(2)
ars, err := t.CopyInIovecs(data, 1)
if err != nil {
return err
}
t.p.PullFullState(t.MemoryManager().AddressSpace(), t.Arch())
ar := ars.Head()
n, err := target.Arch().PtraceGetRegSet(uintptr(addr), &usermem.IOReadWriter{
Ctx: t,
IO: t.MemoryManager(),
Addr: ar.Start,
Opts: usermem.IOOpts{
AddressSpaceActive: true,
},
}, int(ar.Length()))
if err != nil {
return err
}
// Update iovecs to represent the range of the written register set.
end, ok := ar.Start.AddLength(uint64(n))
if !ok {
panic(fmt.Sprintf("%#x + %#x overflows. Invalid reg size > %#x", ar.Start, n, ar.Length()))
}
ar.End = end
return t.CopyOutIovecs(data, hostarch.AddrRangeSeqOf(ar))
case linux.PTRACE_SETREGSET:
ars, err := t.CopyInIovecs(data, 1)
if err != nil {
return err
}
mm := t.MemoryManager()
t.p.PullFullState(mm.AddressSpace(), t.Arch())
ar := ars.Head()
n, err := target.Arch().PtraceSetRegSet(uintptr(addr), &usermem.IOReadWriter{
Ctx: t,
IO: mm,
Addr: ar.Start,
Opts: usermem.IOOpts{
AddressSpaceActive: true,
},
}, int(ar.Length()))
if err != nil {
return err
}
t.p.FullStateChanged()
ar.End -= hostarch.Addr(n)
return t.CopyOutIovecs(data, hostarch.AddrRangeSeqOf(ar))
case linux.PTRACE_GETSIGINFO:
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if target.ptraceSiginfo == nil {
return syserror.EINVAL
}
_, err := target.ptraceSiginfo.CopyOut(t, data)
return err
case linux.PTRACE_SETSIGINFO:
var info arch.SignalInfo
if _, err := info.CopyIn(t, data); err != nil {
return err
}
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
if target.ptraceSiginfo == nil {
return syserror.EINVAL
}
target.ptraceSiginfo = &info
return nil
case linux.PTRACE_GETSIGMASK:
if addr != linux.SignalSetSize {
return syserror.EINVAL
}
mask := target.SignalMask()
_, err := mask.CopyOut(t, data)
return err
case linux.PTRACE_SETSIGMASK:
if addr != linux.SignalSetSize {
return syserror.EINVAL
}
var mask linux.SignalSet
if _, err := mask.CopyIn(t, data); err != nil {
return err
}
// The target's task goroutine is stopped, so this is safe:
target.SetSignalMask(mask &^ UnblockableSignals)
return nil
case linux.PTRACE_SETOPTIONS:
t.tg.pidns.owner.mu.Lock()
defer t.tg.pidns.owner.mu.Unlock()
return target.ptraceSetOptionsLocked(uintptr(data))
case linux.PTRACE_GETEVENTMSG:
t.tg.pidns.owner.mu.RLock()
defer t.tg.pidns.owner.mu.RUnlock()
_, err := primitive.CopyUint64Out(t, hostarch.Addr(data), target.ptraceEventMsg)
return err
// PEEKSIGINFO is unimplemented but seems to have no users anywhere.
default:
return t.ptraceArch(target, req, addr, data)
}
}
|