forked from ioi/isolate
-
Notifications
You must be signed in to change notification settings - Fork 0
/
isolate.c
1407 lines (1255 loc) · 34.1 KB
/
isolate.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
/*
* A Process Isolator based on Linux Containers
*
* (c) 2012-2024 Martin Mares <mj@ucw.cz>
* (c) 2012-2014 Bernard Blackham <bernard@blackham.com.au>
*/
#include "isolate.h"
#include <assert.h>
#include <errno.h>
#include <fcntl.h>
#include <getopt.h>
#include <grp.h>
#include <limits.h>
#include <sched.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <net/if.h>
#include <sys/file.h>
#include <sys/mount.h>
#include <sys/resource.h>
#include <sys/signal.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/time.h>
#include <sys/vfs.h>
#include <sys/wait.h>
#include <time.h>
#include <unistd.h>
/* May not be defined in older glibc headers */
#ifndef MS_PRIVATE
#warning "Working around old glibc: no MS_PRIVATE"
#define MS_PRIVATE (1 << 18)
#endif
#ifndef MS_REC
#warning "Working around old glibc: no MS_REC"
#define MS_REC (1 << 14)
#endif
/*
* Theory of operation
*
* Generally, we want to run a process inside a namespace/cgroup and watch it
* from the outside. However, the reality is a little bit more complicated as we
* do not want the inside process to become the init process of the PID namespace
* (we want to have all signals properly delivered).
*
* We are running three processes:
*
* - Keeper process (root privileges, parent namespace, parent cgroups)
* - Proxy process (UID/GID of the calling user, init process of the child
* namespace, parent cgroups)
* - Inside process (per-box UID/GID, child namespace, child cgroups)
*
* The proxy process just waits for the inside process to exit and then it passes
* the exit status to the keeper.
*
* We use two pipes:
*
* - Error pipe for error messages produced by the proxy process and the early
* stages of the inside process (until exec()). Listened to by the keeper.
* - Status pipe for passing the PID of the inside process and its exit status
* from the proxy to the keeper.
*/
#define TIMER_INTERVAL_US 100000
static int timeout; /* milliseconds */
static int wall_timeout;
static int extra_timeout;
int pass_environ;
int verbose;
static int silent;
static int fsize_limit;
static int memory_limit;
static int stack_limit;
static int open_file_limit = 64;
static int core_limit;
int block_quota;
int inode_quota;
static int max_processes = 1;
static char *redir_stdin, *redir_stdout, *redir_stderr;
static int redir_stderr_to_stdout;
static char *set_cwd;
static int share_net;
static int inherit_fds;
static int default_dirs = 1;
static int tty_hack;
static bool special_files;
static bool wait_if_busy;
static int as_uid = -1;
static int as_gid = -1;
int cg_enable;
int cg_memory_limit;
int box_id;
static char box_dir[1024];
static pid_t box_pid;
static pid_t proxy_pid;
uid_t box_uid;
gid_t box_gid;
uid_t orig_uid;
gid_t orig_gid;
static bool invoked_by_root;
static int partial_line;
static int cleanup_ownership;
static struct timeval start_time;
static int ticks_per_sec;
static int total_ms, wall_ms;
static volatile sig_atomic_t timer_tick, interrupt;
static int error_pipes[2];
static int write_errors_to_fd;
static int read_errors_from_fd;
static int status_pipes[2];
static int get_wall_time_ms(void);
static int get_run_time_ms(struct rusage *rus);
/*** Locks ***/
/*
* Whenever a sandbox is initialized, a lock file is created, which
* records which user owns the sandbox and whether the cgroup mode is used.
* Atempts to use the same sandbox by a different user are refused.
*
* The lock file is locked whenever Isolate runs in that sandbox.
*/
#define LOCK_MAGIC 0x48736f6c
struct lock_record {
uint32_t magic;
uint32_t owner_uid;
unsigned char cg_enabled;
unsigned char is_initialized;
unsigned char rfu[2];
};
static int lock_fd = -1;
static struct lock_record lock;
static void
lock_write(void)
{
int n = pwrite(lock_fd, &lock, sizeof(lock), 0);
if (n != sizeof(lock))
die("Cannot write lock file: %m");
}
static bool
lock_box(bool is_init)
{
if (!dir_exists(cf_lock_root))
make_dir(cf_lock_root);
char lock_name[256];
int name_len = snprintf(lock_name, sizeof(lock_name), "%s/%d", cf_lock_root, box_id);
assert(name_len < (int) sizeof(lock_name));
lock_fd = open(lock_name, O_RDWR | (is_init ? O_CREAT : 0), 0666);
if (lock_fd < 0)
{
if (errno == ENOENT)
return false;
die("Cannot open %s: %m", lock_name);
}
if (flock(lock_fd, LOCK_EX | (wait_if_busy ? 0 : LOCK_NB)) < 0)
{
if (errno == EWOULDBLOCK)
die("This box is currently in use by another process");
die("Cannot lock %s: %m", lock_name);
}
int n = read(lock_fd, &lock, sizeof(lock));
if (n < 0)
die("Cannot read %s: %m");
if (n > 0)
{
if (n != sizeof(lock) || lock.magic != LOCK_MAGIC)
die("Lock file %s has incompatible format", lock_name);
if (lock.is_initialized && lock.owner_uid != orig_uid && !invoked_by_root)
die("This box belongs to a different user (uid %d)", lock.owner_uid);
if (lock.cg_enabled != cg_enable)
die("This box was initialized with an incompatible control group mode");
}
if (is_init)
{
lock.magic = LOCK_MAGIC;
lock.owner_uid = orig_uid;
lock.cg_enabled = cg_enable;
lock.is_initialized = 0;
lock_write();
return true;
}
else
{
if (n > 0)
{
if (!lock.is_initialized)
die("This box was not initialized properly");
return true;
}
else
{
// This means that somebody else is just creating the sandbox and we locked it
// between his creation of the lock file and locking it.
return false;
}
}
// The acquired lock will be automatically released on process exit.
}
static void
lock_close(void)
{
if (lock_fd >= 0)
{
close(lock_fd);
lock_fd = -1;
}
}
static void
lock_remove(void)
{
// To avoid race conditions, we must never unlink lock files.
// We just truncate them to zero length.
assert(lock_fd >= 0);
if (ftruncate(lock_fd, 0) < 0)
die("Cannot truncate lock file: %m");
close(lock_fd);
lock_fd = -1;
}
/*** Messages and exits ***/
static void
final_stats(struct rusage *rus)
{
total_ms = get_run_time_ms(rus);
wall_ms = get_wall_time_ms();
meta_printf("time:%d.%03d\n", total_ms/1000, total_ms%1000);
meta_printf("time-wall:%d.%03d\n", wall_ms/1000, wall_ms%1000);
meta_printf("max-rss:%ld\n", rus->ru_maxrss);
meta_printf("csw-voluntary:%ld\n", rus->ru_nvcsw);
meta_printf("csw-forced:%ld\n", rus->ru_nivcsw);
cg_stats();
}
static void NONRET
box_exit(int rc)
{
if (proxy_pid > 0)
{
if (box_pid > 0)
{
kill(-box_pid, SIGKILL);
kill(box_pid, SIGKILL);
}
if (cg_enable)
{
/*
* In non-CG mode, we must not kill the proxy explicitly.
* This is important, because the proxy could exit before the box
* completes its exit, causing rusage of the box to be lost.
*
* In CG mode, we must kill the proxy, because it is the init
* process of the CG and killing it causes all other processes
* inside the CG to be killed. However, we do not care about
* rusage.
*/
kill(-proxy_pid, SIGKILL);
kill(proxy_pid, SIGKILL);
}
meta_printf("killed:1\n");
/*
* The rusage will contain time spent by the proxy and its children (i.e., the box).
* (See comments on killing of the proxy above, though.)
*/
struct rusage rus;
int p, stat;
do
p = wait4(proxy_pid, &stat, 0, &rus);
while (p < 0 && errno == EINTR);
if (p < 0)
fprintf(stderr, "UGH: Lost track of the process (%m)\n");
else
final_stats(&rus);
}
if (tty_hack && isatty(1))
{
/*
* If stdout is a tty, make us the foreground process group again.
* We do not need it (we ignore SIGTTOU anyway), but programs executed
* after our exit will.
*/
tcsetpgrp(1, getpgrp());
}
if (rc < 2 && cleanup_ownership)
chowntree("box", orig_uid, orig_gid, special_files);
meta_close();
exit(rc);
}
static void
flush_line(void)
{
if (partial_line)
fputc('\n', stderr);
partial_line = 0;
}
/* Report an error of the sandbox itself */
void NONRET __attribute__((format(printf,1,2)))
die(char *msg, ...)
{
va_list args;
va_start(args, msg);
char buf[1024];
int n = vsnprintf(buf, sizeof(buf), msg, args);
// If the child processes are still running, show no mercy.
if (box_pid > 0)
{
kill(-box_pid, SIGKILL);
kill(box_pid, SIGKILL);
}
if (proxy_pid > 0)
{
kill(-proxy_pid, SIGKILL);
kill(proxy_pid, SIGKILL);
}
if (write_errors_to_fd)
{
// We are inside the box, have to use error pipe for error reporting.
// We hope that the whole error message fits in PIPE_BUF bytes.
write(write_errors_to_fd, buf, n);
exit(2);
}
// Otherwise, we in the box keeper process, so we report errors normally
flush_line();
meta_printf("status:XX\nmessage:%s\n", buf);
fputs(buf, stderr);
fputc('\n', stderr);
box_exit(2);
}
/* Report an error of the program inside the sandbox */
void NONRET __attribute__((format(printf,1,2)))
err(char *msg, ...)
{
va_list args;
va_start(args, msg);
flush_line();
if (msg[0] && msg[1] && msg[2] == ':' && msg[3] == ' ')
{
meta_printf("status:%c%c\n", msg[0], msg[1]);
msg += 4;
}
char buf[1024];
vsnprintf(buf, sizeof(buf), msg, args);
meta_printf("message:%s\n", buf);
if (!silent)
{
fputs(buf, stderr);
fputc('\n', stderr);
}
box_exit(1);
}
/* Write a message, but only if in verbose mode */
void __attribute__((format(printf,1,2)))
msg(char *msg, ...)
{
va_list args;
va_start(args, msg);
if (verbose)
{
int len = strlen(msg);
if (len > 0)
partial_line = (msg[len-1] != '\n');
vfprintf(stderr, msg, args);
fflush(stderr);
}
va_end(args);
}
/*** Signal handling in keeper process ***/
/*
* Signal handling is tricky. We must set up signal handlers before
* we start the child process (and reset them in the child process).
* Otherwise, there is a short time window where a SIGINT can kill
* us and leave the child process running.
*/
struct signal_rule {
int signum;
enum { SIGNAL_IGNORE, SIGNAL_INTERRUPT, SIGNAL_FATAL } action;
};
static const struct signal_rule signal_rules[] = {
{ SIGHUP, SIGNAL_INTERRUPT },
{ SIGINT, SIGNAL_INTERRUPT },
{ SIGQUIT, SIGNAL_INTERRUPT },
{ SIGILL, SIGNAL_FATAL },
{ SIGABRT, SIGNAL_FATAL },
{ SIGFPE, SIGNAL_FATAL },
{ SIGSEGV, SIGNAL_FATAL },
{ SIGPIPE, SIGNAL_IGNORE },
{ SIGTERM, SIGNAL_INTERRUPT },
{ SIGUSR1, SIGNAL_IGNORE },
{ SIGUSR2, SIGNAL_IGNORE },
{ SIGBUS, SIGNAL_FATAL },
{ SIGTTOU, SIGNAL_IGNORE },
};
static void
signal_alarm(int unused UNUSED)
{
/* Time limit checks are synchronous, so we only schedule them there. */
timer_tick = 1;
msg("[timer]");
}
static void
signal_int(int signum)
{
/* Interrupts (e.g., SIGINT) are synchronous, too. */
interrupt = signum;
}
static void
signal_fatal(int signum)
{
/* If we receive SIGSEGV or a similar signal, we try to die gracefully. */
die("Sandbox keeper received fatal signal %d", signum);
}
static void
setup_signals(void)
{
struct sigaction sa_int, sa_fatal;
bzero(&sa_int, sizeof(sa_int));
sa_int.sa_handler = signal_int;
bzero(&sa_fatal, sizeof(sa_fatal));
sa_fatal.sa_handler = signal_fatal;
for (int i=0; i < ARRAY_SIZE(signal_rules); i++)
{
const struct signal_rule *sr = &signal_rules[i];
switch (sr->action)
{
case SIGNAL_IGNORE:
signal(sr->signum, SIG_IGN);
break;
case SIGNAL_INTERRUPT:
sigaction(sr->signum, &sa_int, NULL);
break;
case SIGNAL_FATAL:
sigaction(sr->signum, &sa_fatal, NULL);
break;
default:
die("Invalid signal rule");
}
}
}
static void
reset_signals(void)
{
for (int i=0; i < ARRAY_SIZE(signal_rules); i++)
signal(signal_rules[i].signum, SIG_DFL);
}
/*** The keeper process ***/
#define PROC_BUF_SIZE 4096
static int
read_proc_file(char *buf, char *name, int *fdp)
{
int c;
if (*fdp < 0)
{
snprintf(buf, PROC_BUF_SIZE, "/proc/%d/%s", (int) box_pid, name);
*fdp = open(buf, O_RDONLY);
if (*fdp < 0)
return 0; // This is OK, the process could have finished
}
lseek(*fdp, 0, SEEK_SET);
if ((c = read(*fdp, buf, PROC_BUF_SIZE-1)) < 0)
{
// Even this could fail if the process disappeared since open()
return 0;
}
if (c >= PROC_BUF_SIZE-1)
die("/proc/$pid/%s too long", name);
buf[c] = 0;
return 1;
}
static int
get_wall_time_ms(void)
{
struct timeval now, wall;
gettimeofday(&now, NULL);
timersub(&now, &start_time, &wall);
return wall.tv_sec*1000 + wall.tv_usec/1000;
}
static int
get_run_time_ms(struct rusage *rus)
{
if (cg_enable)
return cg_get_run_time_ms();
if (rus)
{
struct timeval total;
timeradd(&rus->ru_utime, &rus->ru_stime, &total);
return total.tv_sec*1000 + total.tv_usec/1000;
}
// It might happen that we do not know the box_pid (see comments in find_box_pid())
if (!box_pid)
return 0;
char buf[PROC_BUF_SIZE], *x;
int utime, stime;
static int proc_stat_fd = -1;
if (!read_proc_file(buf, "stat", &proc_stat_fd))
return 0;
x = buf;
while (*x && *x != ' ')
x++;
while (*x == ' ')
x++;
if (*x++ != '(')
die("proc stat syntax error 1");
while (*x && (*x != ')' || x[1] != ' '))
x++;
while (*x == ')' || *x == ' ')
x++;
if (sscanf(x, "%*c %*d %*d %*d %*d %*d %*d %*d %*d %*d %*d %d %d", &utime, &stime) != 2)
die("proc stat syntax error 2");
return (utime + stime) * 1000 / ticks_per_sec;
}
static void
check_timeout(void)
{
if (wall_timeout)
{
int wall_ms = get_wall_time_ms();
if (wall_ms > wall_timeout)
err("TO: Time limit exceeded (wall clock)");
if (verbose > 1)
fprintf(stderr, "[wall time check: %d msec]\n", wall_ms);
}
if (timeout)
{
int ms = get_run_time_ms(NULL);
if (verbose > 1)
fprintf(stderr, "[time check: %d msec]\n", ms);
if (ms > timeout && ms > extra_timeout)
err("TO: Time limit exceeded");
}
}
static void
box_keeper(void)
{
read_errors_from_fd = error_pipes[0];
close(error_pipes[1]);
close(status_pipes[1]);
gettimeofday(&start_time, NULL);
ticks_per_sec = sysconf(_SC_CLK_TCK);
if (ticks_per_sec <= 0)
die("Invalid ticks_per_sec!");
if (timeout || wall_timeout)
{
struct sigaction sa;
bzero(&sa, sizeof(sa));
sa.sa_handler = signal_alarm;
sigaction(SIGALRM, &sa, NULL);
struct itimerval timer = {
.it_interval = { .tv_usec = TIMER_INTERVAL_US },
.it_value = { .tv_usec = TIMER_INTERVAL_US },
};
setitimer(ITIMER_REAL, &timer, NULL);
}
for(;;)
{
struct rusage rus;
int stat;
pid_t p;
if (interrupt)
{
meta_printf("exitsig:%d\n", interrupt);
err("SG: Interrupted");
}
if (timer_tick)
{
check_timeout();
timer_tick = 0;
}
p = wait4(proxy_pid, &stat, 0, &rus);
if (p < 0)
{
if (errno == EINTR)
continue;
die("wait4: %m");
}
if (p != proxy_pid)
die("wait4: unknown pid %d exited!", p);
proxy_pid = 0;
// Check error pipe if there is an internal error passed from inside the box
char interr[1024];
int n = read(read_errors_from_fd, interr, sizeof(interr) - 1);
if (n > 0)
{
interr[n] = 0;
die("%s", interr);
}
// Check status pipe if there is an exit status reported by the proxy process
n = read(status_pipes[0], &stat, sizeof(stat));
if (n != sizeof(stat))
die("Did not receive exit status from proxy");
// At this point, the rusage includes time spent by the proxy's children.
final_stats(&rus);
if (timeout && total_ms > timeout)
err("TO: Time limit exceeded");
if (wall_timeout && wall_ms > wall_timeout)
err("TO: Time limit exceeded (wall clock)");
if (WIFEXITED(stat))
{
meta_printf("exitcode:%d\n", WEXITSTATUS(stat));
if (WEXITSTATUS(stat))
err("RE: Exited with error status %d", WEXITSTATUS(stat));
flush_line();
if (!silent)
{
fprintf(stderr, "OK (%d.%03d sec real, %d.%03d sec wall)\n",
total_ms/1000, total_ms%1000,
wall_ms/1000, wall_ms%1000);
}
box_exit(0);
}
else if (WIFSIGNALED(stat))
{
meta_printf("exitsig:%d\n", WTERMSIG(stat));
err("SG: Caught fatal signal %d", WTERMSIG(stat));
}
else if (WIFSTOPPED(stat))
{
meta_printf("exitsig:%d\n", WSTOPSIG(stat));
err("SG: Stopped by signal %d", WSTOPSIG(stat));
}
else
die("wait4: unknown status %x, giving up!", stat);
}
}
/*** The process running inside the box ***/
static void
setup_root(void)
{
if (mkdir("root", 0750) < 0 && errno != EEXIST)
die("mkdir('root'): %m");
/*
* Ensure all mounts are private, not shared. We don't want our mounts
* appearing outside of our namespace.
* (systemd since version 188 mounts filesystems shared by default).
*/
if (mount(NULL, "/", NULL, MS_REC|MS_PRIVATE, NULL) < 0)
die("Cannot privatize mounts: %m");
if (mount("none", "root", "tmpfs", 0, "mode=755") < 0)
die("Cannot mount root ramdisk: %m");
apply_dir_rules(default_dirs);
if (chroot("root") < 0)
die("Chroot failed: %m");
if (chdir("root/box") < 0)
die("Cannot change current directory: %m");
}
static void
setup_net(void)
{
if (share_net)
return;
int fd = socket(PF_INET, SOCK_DGRAM, 0);
if (fd < 0)
die("Cannot create PF_INET socket: %m");
struct ifreq ifr = { .ifr_name = "lo" };
if (ioctl(fd, SIOCGIFFLAGS, &ifr) < 0)
die("SIOCGIFFLAGS on 'lo' failed: %m");
ifr.ifr_flags |= IFF_UP;
if (ioctl(fd, SIOCSIFFLAGS, &ifr) < 0)
die("SIOCSIFFLAGS on 'lo' failed: %m");
close(fd);
}
static void
setup_credentials(void)
{
if (setresgid(box_gid, box_gid, box_gid) < 0)
die("setresgid: %m");
if (setgroups(0, NULL) < 0)
die("setgroups: %m");
if (setresuid(box_uid, box_uid, box_uid) < 0)
die("setresuid: %m");
setpgrp();
if (tty_hack && isatty(1))
{
// If stdout is a tty, make us the foreground process group
signal(SIGTTOU, SIG_IGN);
tcsetpgrp(1, getpgrp());
signal(SIGTTOU, SIG_DFL);
}
}
static void
setup_fds(void)
{
if (redir_stdin)
{
close(0);
if (open(redir_stdin, O_RDONLY) != 0)
die("open(\"%s\"): %m", redir_stdin);
}
if (redir_stdout)
{
close(1);
if (open(redir_stdout, O_WRONLY | O_CREAT | O_TRUNC, 0666) != 1)
die("open(\"%s\"): %m", redir_stdout);
}
if (redir_stderr)
{
close(2);
if (open(redir_stderr, O_WRONLY | O_CREAT | O_TRUNC, 0666) != 2)
die("open(\"%s\"): %m", redir_stderr);
}
if (redir_stderr_to_stdout)
{
if (dup2(1, 2) < 0)
die("Cannot dup stdout to stderr: %m");
}
}
static void
setup_rlim(const char *res_name, int res, rlim_t limit)
{
struct rlimit rl = { .rlim_cur = limit, .rlim_max = limit };
if (setrlimit(res, &rl) < 0)
die("setrlimit(%s, %jd)", res_name, (intmax_t) limit);
}
static void
setup_rlimits(void)
{
#define RLIM(res, val) setup_rlim("RLIMIT_" #res, RLIMIT_##res, val)
if (memory_limit)
RLIM(AS, (rlim_t)memory_limit * 1024);
RLIM(FSIZE, (rlim_t)fsize_limit * 1024);
RLIM(NOFILE, 512);
RLIM(STACK, (rlim_t)stack_limit * 1024);
RLIM(MEMLOCK, 0);
RLIM(CORE, (rlim_t)core_limit * 1024);
if (max_processes)
RLIM(NPROC, max_processes);
#undef RLIM
}
static int
box_inside(char **args)
{
cg_enter();
setup_root();
setup_net();
setup_rlimits();
setup_credentials();
setup_fds();
char **env = setup_environment();
if (set_cwd && chdir(set_cwd))
die("chdir: %m");
execve(args[0], args, env);
fprintf(stderr, "execve(\"%s\"): %m\n", args[0]);
exit(127);
}
/*** Proxy ***/
static void
setup_orig_credentials(void)
{
if (setresgid(orig_gid, orig_gid, orig_gid) < 0)
die("setresgid: %m");
if (setgroups(0, NULL) < 0)
die("setgroups: %m");
if (setresuid(orig_uid, orig_uid, orig_uid) < 0)
die("setresuid: %m");
}
static int
box_proxy(void *arg)
{
char **args = arg;
write_errors_to_fd = error_pipes[1];
close(error_pipes[0]);
close(status_pipes[0]);
meta_close();
lock_close();
reset_signals();
pid_t inside_pid = fork();
if (inside_pid < 0)
die("Cannot run process, fork failed: %m");
else if (!inside_pid)
{
close(status_pipes[1]);
box_inside(args);
_exit(42); // We should never get here
}
setup_orig_credentials();
if (write(status_pipes[1], &inside_pid, sizeof(inside_pid)) != sizeof(inside_pid))
die("Proxy write to pipe failed: %m");
int stat;
pid_t p = waitpid(inside_pid, &stat, 0);
if (p < 0)
die("Proxy waitpid() failed: %m");
if (write(status_pipes[1], &stat, sizeof(stat)) != sizeof(stat))
die("Proxy write to pipe failed: %m");
_exit(0);
}
static void
box_init(void)
{
if (box_id < 0 || box_id >= cf_num_boxes)
die("Sandbox ID out of range (allowed: 0-%d)", cf_num_boxes-1);
box_uid = cf_first_uid + box_id;
box_gid = cf_first_gid + box_id;
snprintf(box_dir, sizeof(box_dir), "%s/%d", cf_box_root, box_id);
}
/*** Commands ***/
static const char *
self_name(void)
{
return cg_enable ? "isolate --cg" : "isolate";
}
static void
get_credentials(void)
{
if (geteuid())
die("Must be started as root");
if (getegid() && setegid(0) < 0)
die("Cannot switch to root group: %m");
orig_uid = getuid();
orig_gid = getgid();
invoked_by_root = !orig_uid;
if (as_uid >= 0 || as_gid >= 0)
{
if (!invoked_by_root)
die("You must be root to use --as-uid or --as-gid");
if (as_uid < 0 || as_gid < 0)
die("--as-uid and --as-gid must be used either both or none");
orig_uid = as_uid;
orig_gid = as_gid;
}
}
static void
do_cleanup(void)
{
if (dir_exists(box_dir))
{
msg("Removing box directory\n");
rmtree(box_dir);
}
cg_remove();
}
static void
init(void)
{
if (cf_restricted_init && !invoked_by_root)
die("New sandboxes can be created only by root");
lock_box(true);
do_cleanup();
msg("Preparing sandbox\n");
make_dir(box_dir);
if (chdir(box_dir) < 0)
die("chdir(%s): %m", box_dir);
if (mkdir("box", 0700) < 0)
die("Cannot create box: %m");
if (chown("box", orig_uid, orig_gid) < 0)
die("Cannot chown box: %m");
cg_create();
set_quota();
lock.is_initialized = 1;
lock_write();
puts(box_dir);
}
static void
cleanup(void)
{
if (!lock_box(false))
msg("Nothing to do -- box did not exist\n");
else
{
msg("Deleting sandbox\n");
do_cleanup();
lock_remove();
}
}
static void
setup_pipe(int *fds, int nonblocking)
{
if (pipe(fds) < 0)
die("pipe: %m");
for (int i=0; i<2; i++)
if (fcntl(fds[i], F_SETFD, fcntl(fds[i], F_GETFD) | FD_CLOEXEC) < 0 ||
nonblocking && fcntl(fds[i], F_SETFL, fcntl(fds[i], F_GETFL) | O_NONBLOCK) < 0)
die("fcntl on pipe: %m");
}
static void
find_box_pid(void)
{
/*
* The box keeper process wants to poll status of the inside process,
* so it needs to know the box_pid. However, it is not easy to obtain:
* we got the PID from the proxy, but it is local to the PID namespace.
* Instead, we ask /proc to enumerate the children of the proxy.