-
Notifications
You must be signed in to change notification settings - Fork 5.9k
Expand file tree
/
Copy pathSubprocess.cpp
More file actions
1477 lines (1312 loc) · 45 KB
/
Subprocess.cpp
File metadata and controls
1477 lines (1312 loc) · 45 KB
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
/*
* Copyright (c) Meta Platforms, Inc. and affiliates.
*
* 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.
*/
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
#include <folly/Subprocess.h>
#if defined(__linux__)
#include <sys/prctl.h>
#endif
#include <dlfcn.h>
#include <fcntl.h>
#include <algorithm>
#include <array>
#include <system_error>
#include <thread>
#include <boost/container/flat_set.hpp>
#include <boost/range/adaptors.hpp>
#include <folly/Conv.h>
#include <folly/Exception.h>
#include <folly/ScopeGuard.h>
#include <folly/String.h>
#include <folly/io/Cursor.h>
#include <folly/lang/Assume.h>
#include <folly/logging/xlog.h>
#include <folly/portability/Dirent.h>
#include <folly/portability/Fcntl.h>
#include <folly/portability/Sockets.h>
#include <folly/portability/Stdlib.h>
#include <folly/portability/SysSyscall.h>
#include <folly/portability/Unistd.h>
#include <folly/system/AtFork.h>
#include <folly/system/Shell.h>
/// interceptors to work around:
///
/// https://github.com/llvm/llvm-project/blob/llvmorg-19.1.7/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp
/// https://github.com/llvm/llvm-project/blob/llvmorg-19.1.7/compiler-rt/lib/sanitizer_common/sanitizer_signal_interceptors.inc
/// In sanitized builds, explicitly disable sanitization in the child process.
/// * Disable sanitizer function transformation, including __tsan_func_entry and
/// __tsan_func_exit hooks (under clang).
/// * Bypass sanitizer interceptors of libc/posix functions, including of vfork
/// and of all other libc/posix callees in the child process. Interceptors
/// look like __interceptor_trampoline_{name} for libc/posix function {name}.
#if __has_attribute(disable_sanitizer_instrumentation)
#define FOLLY_DETAIL_SUBPROCESS_RAW \
__attribute__(( \
noinline, \
no_sanitize("address", "undefined", "thread"), \
disable_sanitizer_instrumentation))
#else
#define FOLLY_DETAIL_SUBPROCESS_RAW \
__attribute__((noinline, no_sanitize("address", "undefined", "thread")))
#endif
constexpr int kExecFailure = 127;
constexpr int kChildFailure = 126;
namespace folly {
using detail::linux_syscall;
namespace detail {
SubprocessFdActionsList::SubprocessFdActionsList(
span<value_type const> rep) noexcept
: begin_{rep.data()}, end_{rep.data() + rep.size()} {
[[maybe_unused]] auto lt = [](auto a, auto b) { return a.first < b.first; };
assert(std::is_sorted(begin_, end_, lt));
[[maybe_unused]] auto eq = [](auto a, auto b) { return a.first == b.first; };
assert(std::adjacent_find(begin_, end_, eq) == end_);
}
FOLLY_DETAIL_SUBPROCESS_RAW
auto SubprocessFdActionsList::begin() const noexcept -> value_type const* {
return begin_;
}
FOLLY_DETAIL_SUBPROCESS_RAW
auto SubprocessFdActionsList::end() const noexcept -> value_type const* {
return end_;
}
FOLLY_DETAIL_SUBPROCESS_RAW
auto SubprocessFdActionsList::find(int fd) const noexcept -> int const* {
auto lo = begin_;
auto hi = end_;
while (lo < hi) {
auto mid = lo + (hi - lo) / 2;
if (mid->first == fd) {
return &mid->second;
}
if (mid->first < fd) {
lo = mid + 1;
} else {
hi = mid;
}
}
return nullptr;
}
// clang-format off
static inline constexpr auto subprocess_libc_soname =
kIsLinux ? "libc.so.6" :
kIsFreeBSD ? "libc.so.7" :
kIsApple ? "/usr/lib/libSystem.B.dylib" :
nullptr;
// clang-format on
template <typename Ret>
static Ret subprocess_libc_load(
void* const handle, Ret const ptr, char const* const name) {
return !kIsSanitize ? ptr : reinterpret_cast<Ret>(::dlsym(handle, name));
}
#define FOLLY_DETAIL_SUBPROCESS_LIBC_X_BASE(X) \
X(_exit, _exit) \
X(close, close) \
X(dup2, dup2) \
X(fcntl, fcntl) \
X(pthread_sigmask, pthread_sigmask) \
X(signal, signal) \
X(sprintf, sprintf) \
X(strtol, strtol) \
X(vfork, vfork) \
X(write, write)
#if defined(__BIONIC_INCLUDE_FORTIFY_HEADERS)
#define FOLLY_DETAIL_SUBPROCESS_LIBC_X_OPEN(X) \
X(open, __open_real) \
X(openat, __openat_real)
#else
#define FOLLY_DETAIL_SUBPROCESS_LIBC_X_OPEN(X) \
X(open, open) \
X(openat, openat)
#endif
#if defined(__linux__)
#define FOLLY_DETAIL_SUBPROCESS_LIBC_X_PRCTL(X) X(prctl, prctl)
#else
#define FOLLY_DETAIL_SUBPROCESS_LIBC_X_PRCTL(X)
#endif
#define FOLLY_DETAIL_SUBPROCESS_LIBC_X(X) \
FOLLY_DETAIL_SUBPROCESS_LIBC_X_BASE(X) \
FOLLY_DETAIL_SUBPROCESS_LIBC_X_OPEN(X) \
FOLLY_DETAIL_SUBPROCESS_LIBC_X_PRCTL(X)
#define FOLLY_DETAIL_SUBPROCESS_LIBC_FIELD_DECL(name, func) \
static decltype(&::func) name;
#define FOLLY_DETAIL_SUBPROCESS_LIBC_FIELD_DEFN(name, func) \
decltype(&::func) subprocess_libc::name;
#define FOLLY_DETAIL_SUBPROCESS_LIBC_INIT(name, func) \
subprocess_libc::name = subprocess_libc_load(handle, &::func, #name);
struct subprocess_libc {
FOLLY_DETAIL_SUBPROCESS_LIBC_X(FOLLY_DETAIL_SUBPROCESS_LIBC_FIELD_DECL)
};
FOLLY_DETAIL_SUBPROCESS_LIBC_X(FOLLY_DETAIL_SUBPROCESS_LIBC_FIELD_DEFN)
__attribute__((constructor(101))) static void subprocess_libc_init() {
auto handle = !kIsSanitize
? nullptr
: ::dlopen(subprocess_libc_soname, RTLD_LAZY | RTLD_LOCAL | RTLD_NOLOAD);
assert(!kIsSanitize || !!handle);
FOLLY_DETAIL_SUBPROCESS_LIBC_X(FOLLY_DETAIL_SUBPROCESS_LIBC_INIT)
}
#undef FOLLY_DETAIL_SUBPROCESS_LIBC_FIELD_DECL
#undef FOLLY_DETAIL_SUBPROCESS_LIBC_FIELD_DEFN
#undef FOLLY_DETAIL_SUBPROCESS_LIBC_INIT
#undef FOLLY_DETAIL_SUBPROCESS_LIBC_X
#undef FOLLY_DETAIL_SUBPROCESS_LIBC_X_PRCTL
#undef FOLLY_DETAIL_SUBPROCESS_LIBC_X_OPEN
#undef FOLLY_DETAIL_SUBPROCESS_LIBC_X_BASE
} // namespace detail
struct Subprocess::SpawnRawArgs {
struct Scratch {
std::vector<std::pair<int, int>> fdActions;
std::vector<char*> setPrintPidToBuffer;
std::vector<std::pair<int, Options::AttrWithMeta<rlimit>>> rlimits;
explicit Scratch(Options const& options)
: fdActions{options.fdActions_.begin(), options.fdActions_.end()},
setPrintPidToBuffer{
options.setPrintPidToBuffer_.begin(),
options.setPrintPidToBuffer_.end()},
rlimits{options.rlimits_.begin(), options.rlimits_.end()} {
std::sort(fdActions.begin(), fdActions.end());
}
};
template <typename T>
struct AttrWithMeta {
T value{};
int* errout{};
};
static char const* getCStrForNonEmpty(std::string const& str) {
return str.empty() ? nullptr : str.c_str();
}
// from options
char const* childDir{};
AttrWithMeta<int> linuxCGroupFd{-1, nullptr};
AttrWithMeta<char const*> linuxCGroupPath{nullptr, nullptr};
bool closeOtherFds{};
#if defined(__linux__)
Options::AttrWithMeta<cpu_set_t> const* cpuSet{};
#endif
bool detach{};
detail::SubprocessFdActionsList fdActions;
int parentDeathSignal{};
bool processGroupLeader{};
bool usePath{};
Options::AttrWithMeta<uid_t> const* uid{};
Options::AttrWithMeta<gid_t> const* gid{};
Options::AttrWithMeta<uid_t> const* euid{};
Options::AttrWithMeta<gid_t> const* egid{};
char* const* setPrintPidToBufferData{};
size_t setPrintPidToBufferSize{};
std::pair<int, Options::AttrWithMeta<rlimit>> const* rlimitsData{};
size_t rlimitsSize{};
// assigned explicitly
char const* const* argv{};
char const* const* envv{};
char const* executable{};
ChildErrorInfo* err{};
sigset_t oldSignals{};
explicit SpawnRawArgs(Scratch const& scratch, Options const& options)
: childDir{getCStrForNonEmpty(options.childDir_)},
linuxCGroupFd{
options.linuxCGroupFd_.value, options.linuxCGroupFd_.errout},
linuxCGroupPath{
getCStrForNonEmpty(options.linuxCGroupPath_.value),
options.linuxCGroupPath_.errout},
closeOtherFds{options.closeOtherFds_},
#if defined(__linux__)
cpuSet{get_pointer(options.cpuSet_)},
#endif
detach{options.detach_},
fdActions{scratch.fdActions},
#if defined(__linux__)
parentDeathSignal{options.parentDeathSignal_},
#endif
processGroupLeader{options.processGroupLeader_},
usePath{options.usePath_},
uid{options.uid_.get_pointer()},
gid{options.gid_.get_pointer()},
euid{options.euid_.get_pointer()},
egid{options.egid_.get_pointer()},
setPrintPidToBufferData{scratch.setPrintPidToBuffer.data()},
setPrintPidToBufferSize{scratch.setPrintPidToBuffer.size()},
rlimitsData{scratch.rlimits.data()},
rlimitsSize{scratch.rlimits.size()} {
static_assert(std::is_standard_layout_v<Subprocess::SpawnRawArgs>);
static_assert(std::is_trivially_destructible_v<Subprocess::SpawnRawArgs>);
}
};
ProcessReturnCode ProcessReturnCode::make(int status) {
if (!WIFEXITED(status) && !WIFSIGNALED(status)) {
throw std::runtime_error(
to<std::string>("Invalid ProcessReturnCode: ", status));
}
return ProcessReturnCode(status);
}
ProcessReturnCode::ProcessReturnCode(ProcessReturnCode&& p) noexcept
: rawStatus_(p.rawStatus_) {
p.rawStatus_ = ProcessReturnCode::RV_NOT_STARTED;
}
ProcessReturnCode& ProcessReturnCode::operator=(
ProcessReturnCode&& p) noexcept {
rawStatus_ = p.rawStatus_;
p.rawStatus_ = ProcessReturnCode::RV_NOT_STARTED;
return *this;
}
ProcessReturnCode::State ProcessReturnCode::state() const {
if (rawStatus_ == RV_NOT_STARTED) {
return NOT_STARTED;
}
if (rawStatus_ == RV_RUNNING) {
return RUNNING;
}
if (WIFEXITED(rawStatus_)) {
return EXITED;
}
if (WIFSIGNALED(rawStatus_)) {
return KILLED;
}
assume_unreachable();
}
void ProcessReturnCode::enforce(State expected) const {
State s = state();
if (s != expected) {
throw std::logic_error(
to<std::string>(
"Bad use of ProcessReturnCode; state is ",
s,
" expected ",
expected));
}
}
int ProcessReturnCode::exitStatus() const {
enforce(EXITED);
return WEXITSTATUS(rawStatus_);
}
int ProcessReturnCode::killSignal() const {
enforce(KILLED);
return WTERMSIG(rawStatus_);
}
bool ProcessReturnCode::coreDumped() const {
enforce(KILLED);
return WCOREDUMP(rawStatus_);
}
bool ProcessReturnCode::succeeded() const {
return exited() && exitStatus() == 0;
}
std::string ProcessReturnCode::str() const {
switch (state()) {
case NOT_STARTED:
return "not started";
case RUNNING:
return "running";
case EXITED:
return to<std::string>("exited with status ", exitStatus());
case KILLED:
return to<std::string>(
"killed by signal ",
killSignal(),
(coreDumped() ? " (core dumped)" : ""));
}
assume_unreachable();
}
CalledProcessError::CalledProcessError(ProcessReturnCode rc)
: SubprocessError(rc.str()), returnCode_(rc) {}
static inline std::string toSubprocessSpawnErrorMessage(
char const* executable, int errCode, int errnoValue) {
auto prefix = errCode == kExecFailure
? "failed to execute "
: "error preparing to execute ";
return to<std::string>(prefix, executable, ": ", errnoStr(errnoValue));
}
SubprocessSpawnError::SubprocessSpawnError(
const char* executable, int errCode, int errnoValue)
: SubprocessError(
toSubprocessSpawnErrorMessage(executable, errCode, errnoValue)),
errnoValue_(errnoValue) {}
namespace {
// Copy pointers to the given strings in a format suitable for posix_spawn
std::unique_ptr<const char*[]> cloneStrings(const std::vector<std::string>& s) {
std::unique_ptr<const char*[]> d(new const char*[s.size() + 1]);
for (size_t i = 0; i < s.size(); i++) {
d[i] = s[i].c_str();
}
d[s.size()] = nullptr;
return d;
}
// Check a wait() status, throw on non-successful
void checkStatus(ProcessReturnCode returnCode) {
if (returnCode.state() != ProcessReturnCode::EXITED ||
returnCode.exitStatus() != 0) {
throw CalledProcessError(returnCode);
}
}
} // namespace
Subprocess::Options& Subprocess::Options::fd(int fd, int action) {
if (fdActions_.contains(fd)) {
throw std::invalid_argument("fd already added");
}
if (action == Subprocess::PIPE) {
if (fd == 0) {
action = Subprocess::PIPE_IN;
} else if (fd == 1 || fd == 2) {
action = Subprocess::PIPE_OUT;
} else {
throw std::invalid_argument(
to<std::string>("Only fds 0, 1, 2 are valid for action=PIPE: ", fd));
}
}
fdActions_[fd] = action;
return *this;
}
#if defined(__linux__)
Subprocess::Options& Subprocess::Options::setLinuxCGroupFd(
int cgroupFd, std::shared_ptr<int> errout) {
if (linuxCGroupFd_.value >= 0 || !linuxCGroupPath_.value.empty()) {
throw std::runtime_error("setLinuxCGroup* called more than once");
}
linuxCGroupFd_ = {cgroupFd, std::move(errout)};
return *this;
}
Subprocess::Options& Subprocess::Options::setLinuxCGroupPath(
const std::string& cgroupPath, std::shared_ptr<int> errout) {
if (linuxCGroupFd_.value >= 0 || !linuxCGroupPath_.value.empty()) {
throw std::runtime_error("setLinuxCGroup* called more than once");
}
linuxCGroupPath_ = {cgroupPath, std::move(errout)};
return *this;
}
#endif
Subprocess::Options& Subprocess::Options::addPrintPidToBuffer(span<char> buf) {
if (buf.size() < kPidBufferMinSize) {
throw std::invalid_argument("buf size too small");
}
setPrintPidToBuffer_.insert(buf.data());
return *this;
}
Subprocess::Options& Subprocess::Options::addRLimit(
int resource, rlimit limit, std::shared_ptr<int> errout) {
if (rlimits_.count(resource)) {
throw std::runtime_error("addRLimit called with same limit more than once");
}
rlimits_[resource] = AttrWithMeta<rlimit>{limit, std::move(errout)};
return *this;
}
Subprocess::Subprocess() = default;
Subprocess::Subprocess(
const std::vector<std::string>& argv,
const Options& options,
const char* executable,
const std::vector<std::string>* env)
: destroyBehavior_(options.destroyBehavior_) {
if (argv.empty()) {
throw std::invalid_argument("argv must not be empty");
}
if (!executable) {
executable = argv[0].c_str();
}
spawn(cloneStrings(argv), executable, options, env);
}
Subprocess::Subprocess(
const std::string& cmd,
const Options& options,
const std::vector<std::string>* env)
: destroyBehavior_(options.destroyBehavior_) {
if (options.usePath_) {
throw std::invalid_argument("usePath() not allowed when running in shell");
}
std::vector<std::string> argv = {"/bin/sh", "-c", cmd};
spawn(cloneStrings(argv), argv[0].c_str(), options, env);
}
Subprocess Subprocess::fromExistingProcess(pid_t pid) {
Subprocess sp;
sp.pid_ = pid;
sp.destroyBehavior_ = DestroyBehaviorLeak;
sp.returnCode_ = ProcessReturnCode::makeRunning();
return sp;
}
Subprocess::~Subprocess() {
if (returnCode_.state() == ProcessReturnCode::RUNNING) {
if (destroyBehavior_ == DestroyBehaviorFatal) {
// Explicitly crash if we are destroyed without reaping the child process.
//
// If you are running into this crash, you are destroying a Subprocess
// without cleaning up the child process first, which can leave behind a
// zombie process on the system until the current process exits. You may
// want to use one of the following options instead when creating the
// Subprocess:
// - Options::detach()
// If you do not want to wait on the child process to complete, and do
// not care about its exit status, use detach().
// - Options::killChildOnDestruction()
// If you want the child process to be automatically killed when the
// Subprocess is destroyed, use killChildOnDestruction() or
// terminateChildOnDestruction()
XLOG(FATAL) << "Subprocess destroyed without reaping child";
} else if (destroyBehavior_ == DestroyBehaviorLeak) {
// Do nothing if we are destroyed without reaping the child process.
XLOG(DBG) << "Subprocess destroyed without reaping child process";
} else {
// If we are killed without reaping the child process, explicitly
// terminate/kill it and wait for it to exit.
try {
TimeoutDuration timeout(destroyBehavior_);
terminateOrKill(timeout);
} catch (const std::exception& ex) {
XLOG(WARN) << "error terminating process in Subprocess destructor: "
<< ex.what();
}
}
}
}
struct Subprocess::ChildErrorInfo {
int errCode;
int errnoValue;
};
[[noreturn]]
FOLLY_DETAIL_SUBPROCESS_RAW void Subprocess::childError(
SpawnRawArgs const& args, int errCode, int errnoValue) {
*args.err = {errCode, errnoValue};
detail::subprocess_libc::_exit(errCode);
__builtin_unreachable();
}
void Subprocess::setAllNonBlocking() {
for (auto& p : pipes_) {
int fd = p.pipe.fd();
int flags = ::fcntl(fd, F_GETFL);
checkUnixError(flags, "fcntl");
int r = ::fcntl(fd, F_SETFL, flags | O_NONBLOCK);
checkUnixError(r, "fcntl");
}
}
void Subprocess::spawn(
std::unique_ptr<const char*[]> argv,
const char* executable,
const Options& optionsIn,
const std::vector<std::string>* env) {
if (optionsIn.usePath_ && env) {
throw std::invalid_argument(
"usePath() not allowed when overriding environment");
}
// Make a copy, we'll mutate options
Options options(optionsIn);
// On error, close all pipes_ (ignoring errors, but that seems fine here).
auto pipesGuard = makeGuard([this] { pipes_.clear(); });
ChildErrorInfo err{};
// Perform the actual work of setting up pipes then forking and
// executing the child.
spawnInternal(std::move(argv), executable, options, env, &err);
// After spawnInternal() returns the child is alive. We have to be very
// careful about throwing after this point. We are inside the constructor,
// so if we throw the Subprocess object will have never existed, and the
// destructor will never be called.
//
// We should only throw if we got an error via the ChildErrorInfo, and we know
// the child has exited and can be immediately waited for. In all other
// cases, we have no way of cleaning up the child.
readChildErrorNum(err, executable);
// If we spawned a detached child, wait on the intermediate child process.
// It always exits immediately.
if (options.detach_) {
wait();
}
// We have fully succeeded now, so release the guard on pipes_
pipesGuard.dismiss();
}
void Subprocess::spawnInternal(
std::unique_ptr<const char*[]> argv,
const char* executable,
Options& options,
const std::vector<std::string>* env,
ChildErrorInfo* err) {
// Parent work, pre-fork: create pipes
std::vector<int> childFds;
// Close all of the childFds as we leave this scope
SCOPE_EXIT {
// These are only pipes, closing them shouldn't fail
for (int cfd : childFds) {
CHECK_ERR(fileops::close(cfd));
}
};
int r;
for (auto& p : options.fdActions_) {
if (p.second == PIPE_IN || p.second == PIPE_OUT) {
int fds[2];
// We're setting both ends of the pipe as close-on-exec. The child
// doesn't need to reset the flag on its end, as we always dup2() the fd,
// and dup2() fds don't share the close-on-exec flag.
#if FOLLY_HAVE_PIPE2
// If possible, set close-on-exec atomically. Otherwise, a concurrent
// Subprocess invocation can fork() between "pipe" and "fnctl",
// causing FDs to leak.
r = ::pipe2(fds, O_CLOEXEC);
checkUnixError(r, "pipe2");
#else
r = fileops::pipe(fds);
checkUnixError(r, "pipe");
r = fcntl(fds[0], F_SETFD, FD_CLOEXEC);
checkUnixError(r, "set FD_CLOEXEC");
r = fcntl(fds[1], F_SETFD, FD_CLOEXEC);
checkUnixError(r, "set FD_CLOEXEC");
#endif
pipes_.emplace_back();
Pipe& pipe = pipes_.back();
pipe.direction = p.second;
int cfd;
if (p.second == PIPE_IN) {
// Child gets reading end
pipe.pipe = folly::File(fds[1], /*ownsFd=*/true);
cfd = fds[0];
} else {
pipe.pipe = folly::File(fds[0], /*ownsFd=*/true);
cfd = fds[1];
}
p.second = cfd; // ensure it gets dup2()ed
pipe.childFd = p.first;
childFds.push_back(cfd);
}
}
// This should already be sorted, as options.fdActions_ is
DCHECK(std::is_sorted(pipes_.begin(), pipes_.end()));
// Note that the const casts below are legit, per
// http://pubs.opengroup.org/onlinepubs/009695399/functions/exec.html
// Set up environment
std::unique_ptr<const char*[]> envHolder;
if (env) {
envHolder = cloneStrings(*env);
}
// Block all signals around vfork; see http://ewontfix.com/7/.
//
// As the child may run in the same address space as the parent until
// the actual execve() system call, any (custom) signal handlers that
// the parent has might alter parent's memory if invoked in the child,
// with undefined results. So we block all signals in the parent before
// vfork(), which will cause them to be blocked in the child as well (we
// rely on the fact that Linux, just like all sane implementations, only
// clones the calling thread). Then, in the child, we reset all signals
// to their default dispositions (while still blocked), and unblock them
// (so the exec()ed process inherits the parent's signal mask)
//
// The parent also unblocks all signals as soon as vfork() returns.
sigset_t allBlocked;
r = sigfillset(&allBlocked);
checkUnixError(r, "sigfillset");
sigset_t oldSignals;
r = pthread_sigmask(SIG_SETMASK, &allBlocked, &oldSignals);
checkPosixError(r, "pthread_sigmask");
SCOPE_EXIT {
// Restore signal mask
r = pthread_sigmask(SIG_SETMASK, &oldSignals, nullptr);
CHECK_EQ(r, 0) << "pthread_sigmask: " << errnoStr(r); // shouldn't fail
};
SpawnRawArgs::Scratch scratch{options};
SpawnRawArgs args{scratch, options};
args.argv = argv.get();
args.envv = env ? envHolder.get() : environ;
args.executable = executable;
args.err = err;
args.oldSignals = options.sigmask_.value_or(oldSignals);
// Child is alive. We have to be very careful about throwing after this
// point. We are inside the constructor, so if we throw the Subprocess
// object will have never existed, and the destructor will never be called.
//
// We should only throw if we got an error via the errFd, and we know the
// child has exited and can be immediately waited for. In all other cases,
// we have no way of cleaning up the child.
pid_ = spawnInternalDoFork(args);
returnCode_ = ProcessReturnCode::makeRunning();
}
// With -Wclobbered, gcc complains about vfork potentially clobbering the
// childDir variable, even though we only use it on the child side of the
// vfork.
FOLLY_PUSH_WARNING
FOLLY_GCC_DISABLE_WARNING("-Wclobbered")
FOLLY_DETAIL_SUBPROCESS_RAW
pid_t Subprocess::spawnInternalDoFork(SpawnRawArgs const& args) {
pid_t pid = detail::subprocess_libc::vfork();
checkUnixError(pid, errno, "failed to fork");
if (pid != 0) {
return pid;
}
// From this point onward, we are in the child.
// Fork a second time if detach_ was requested.
// This must be done before signals are restored in prepareChild()
if (args.detach) {
pid = detail::subprocess_libc::vfork();
if (pid == -1) {
// Inform our parent process of the error so it can throw in the parent.
childError(args, kChildFailure, errno);
} else if (pid != 0) {
// We are the intermediate process. Exit immediately.
// Our child will still inform the original parent of success/failure
// through errFd. The pid of the grandchild process never gets
// propagated back up to the original parent. In the future we could
// potentially send it back using errFd if we needed to.
detail::subprocess_libc::_exit(0);
}
}
int errnoValue = prepareChild(args);
if (errnoValue != 0) {
childError(args, kChildFailure, errnoValue);
}
errnoValue = runChild(args);
// If we get here, exec() failed.
childError(args, kExecFailure, errnoValue);
return 0; // unreachable
}
FOLLY_POP_WARNING
FOLLY_DETAIL_SUBPROCESS_RAW
int Subprocess::prepareChildDoOptionalError(int* errout) {
if (errout) {
*errout = errno;
return 0;
} else {
return errno;
}
}
FOLLY_DETAIL_SUBPROCESS_RAW
int Subprocess::prepareChildDoLinuxCGroup(SpawnRawArgs const& args) {
auto cgroupPath = args.linuxCGroupPath;
auto cgroupFd = args.linuxCGroupFd;
if (nullptr != cgroupPath.value) {
int fd = detail::subprocess_libc::open(
cgroupPath.value, O_RDONLY | O_DIRECTORY | O_CLOEXEC);
if (-1 == fd) {
return prepareChildDoOptionalError(cgroupPath.errout);
}
cgroupFd = {fd, cgroupPath.errout};
}
if (-1 != cgroupFd.value) {
int fd = detail::subprocess_libc::openat(
cgroupFd.value, "cgroup.procs", O_WRONLY | O_CLOEXEC);
if (fd == -1) {
return prepareChildDoOptionalError(cgroupFd.errout);
}
int rc = 0;
do {
constexpr char const buf = '0';
rc = detail::subprocess_libc::write(fd, &buf, 1);
} while (rc == -1 && errno == EINTR);
if (rc == -1) {
return prepareChildDoOptionalError(cgroupFd.errout);
}
}
return 0;
}
// If requested, close all other file descriptors. Don't close
// any fds in options.fdActions_, and don't touch stdin, stdout, stderr.
// Ignore errors.
//
//
// This function is called in the child after fork but before exec so
// there is very little it can do. It cannot allocate memory and
// it cannot lock a mutex, just as if it were running in a signal
// handler.
FOLLY_DETAIL_SUBPROCESS_RAW
void Subprocess::closeInheritedFds(const SpawnRawArgs& args) {
#if defined(__linux__)
int dirfd = detail::subprocess_libc::open("/proc/self/fd", O_RDONLY);
if (dirfd != -1) {
char buffer[32768];
int res;
while ((res = syscall(SYS_getdents64, dirfd, buffer, sizeof(buffer))) > 0) {
// linux_dirent64 is part of the kernel ABI for the getdents64 system
// call. It is currently the same as struct dirent64 in both glibc and
// musl, but those are library specific and could change. linux_dirent64
// is not defined in the standard set of Linux userspace headers
// (/usr/include/linux)
//
// We do not use the POSIX interfaces (opendir, readdir, etc..) for
// reading a directory since they may allocate memory / grab a lock, which
// is unsafe in this context.
FOLLY_PUSH_WARNING
FOLLY_CLANG_DISABLE_WARNING("-Wzero-length-array")
struct linux_dirent64 {
uint64_t d_ino;
int64_t d_off;
uint16_t d_reclen;
unsigned char d_type;
char d_name[0];
} const* entry;
FOLLY_POP_WARNING
for (int offset = 0; offset < res; offset += entry->d_reclen) {
entry = reinterpret_cast<struct linux_dirent64*>(buffer + offset);
if (entry->d_type != DT_LNK) {
continue;
}
char* end_p = nullptr;
errno = 0;
int fd = static_cast<int>(
detail::subprocess_libc::strtol(entry->d_name, &end_p, 10));
if (errno == ERANGE || fd < 3 || end_p == entry->d_name) {
continue;
}
if ((fd != dirfd) && (args.fdActions.find(fd) == nullptr)) {
detail::subprocess_libc::close(fd);
}
}
}
detail::subprocess_libc::close(dirfd);
return;
}
#endif
// If not running on Linux or if we failed to open /proc/self/fd, try to close
// all possible open file descriptors.
for (auto fd = sysconf(_SC_OPEN_MAX) - 1; fd >= 3; --fd) {
if (args.fdActions.find(fd) == nullptr) {
detail::subprocess_libc::close(fd);
}
}
}
FOLLY_DETAIL_SUBPROCESS_RAW
int Subprocess::prepareChild(SpawnRawArgs const& args) {
// While all signals are blocked, we must reset their
// dispositions to default.
for (int sig = 1; sig < NSIG; ++sig) {
detail::subprocess_libc::signal(sig, SIG_DFL);
}
{
// Unblock signals; restore signal mask.
int r = detail::subprocess_libc::pthread_sigmask(
SIG_SETMASK, &args.oldSignals, nullptr);
if (r != 0) {
return r; // pthread_sigmask() returns an errno value
}
}
// Move the child process into a linux cgroup, if one is given
if (auto rc = prepareChildDoLinuxCGroup(args)) {
return rc;
}
for (size_t i = 0; i < args.rlimitsSize; ++i) {
auto const& limit = args.rlimitsData[i];
if (setrlimit(limit.first, &limit.second.value) == -1) {
if (limit.second.errout) {
*limit.second.errout = errno;
} else {
return errno;
}
}
}
// Change the working directory, if one is given
if (args.childDir) {
if (::chdir(args.childDir) == -1) {
return errno;
}
}
#ifdef __linux__
// Best effort
if (args.cpuSet) {
const auto& cpuSet = *args.cpuSet;
if (::sched_setaffinity(0, sizeof(cpuSet.value), &cpuSet.value) == -1) {
if (cpuSet.errout) {
*cpuSet.errout = errno;
} else {
return errno;
}
}
}
#endif
// Change effective/real group/user, if requested
// Call the raw syscall directly for linux, as glibc set*id() is not
// safe after vfork() - Linux is the only kernel where libc set*id is not just
// a call to syscall set*id
// The risk is that some vfork child is terminated within set*id() with
// a lock held, deadlocking the next vfork child
auto idval = [](auto id) { return id ? &id->value : nullptr; };
#if defined(__linux__)
constexpr auto k_sys_setresgid = SYS_setresgid;
constexpr auto k_sys_setresuid = SYS_setresuid;
#else
// Unused
constexpr auto k_sys_setresgid = -1;
constexpr auto k_sys_setresuid = -1;
#endif
if (auto p = idval(args.egid); kIsLinux
? p && 0 != linux_syscall(k_sys_setresgid, -1, *p, -1)
: p && 0 != ::setegid(*p)) {
if (auto out = args.egid->errout) {
*out = errno;
} else {
return errno;
}
}
if (auto p = idval(args.gid); kIsLinux
? p && 0 != linux_syscall(SYS_setgid, *p)
: p && 0 != ::setgid(*p)) {
if (auto out = args.gid->errout) {
*out = errno;
} else {
return errno;
}
}
if (auto p = idval(args.euid); kIsLinux
? p && 0 != linux_syscall(k_sys_setresuid, -1, *p, -1)
: p && 0 != ::seteuid(*p)) {
if (auto out = args.euid->errout) {
*out = errno;
} else {
return errno;
}
}
if (auto p = idval(args.uid); kIsLinux
? p && 0 != linux_syscall(SYS_setuid, *p)
: p && 0 != ::setuid(*p)) {
if (auto out = args.uid->errout) {
*out = errno;
} else {
return errno;
}
}
// We don't have to explicitly close the parent's end of all pipes,
// as they all have the FD_CLOEXEC flag set and will be closed at
// exec time.
// Redirect requested FDs to /dev/null or NUL
// dup2 any explicitly specified FDs
for (auto p : args.fdActions) {
if (p.second == DEV_NULL) {
// folly/portability/Fcntl provides an impl of open that will
// map this to NUL on Windows.
auto devNull =
detail::subprocess_libc::open("/dev/null", O_RDWR | O_CLOEXEC);
if (devNull == -1) {
return errno;
}
// note: dup2 will not set CLOEXEC on the destination
if (detail::subprocess_libc::dup2(devNull, p.first) == -1) {
// explicit close on error to avoid leaking fds
detail::subprocess_libc::close(devNull);
return errno;
}
detail::subprocess_libc::close(devNull);
} else if (p.second != p.first && p.second != NO_CLOEXEC) {
if (detail::subprocess_libc::dup2(p.second, p.first) == -1) {
return errno;
}
} else if (p.second == p.first || p.second == NO_CLOEXEC) {
int flags = detail::subprocess_libc::fcntl(p.first, F_GETFD);
if (flags == -1) {