1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2011 James Gritton
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29 #include <sys/types.h>
30 #include <sys/cpuset.h>
31 #include <sys/event.h>
32 #include <sys/mount.h>
33 #include <sys/stat.h>
34 #include <sys/sysctl.h>
35 #include <sys/user.h>
36 #include <sys/wait.h>
37
38 #include <err.h>
39 #include <errno.h>
40 #include <fcntl.h>
41 #include <kvm.h>
42 #include <login_cap.h>
43 #include <paths.h>
44 #include <pwd.h>
45 #include <signal.h>
46 #include <stdio.h>
47 #include <stdlib.h>
48 #include <string.h>
49 #include <unistd.h>
50 #include <vis.h>
51
52 #include "jailp.h"
53
54 #define DEFAULT_STOP_TIMEOUT 10
55 #define PHASH_SIZE 256
56
57 LIST_HEAD(phhead, phash);
58
59 struct phash {
60 LIST_ENTRY(phash) le;
61 struct cfjail *j;
62 pid_t pid;
63 };
64
65 int paralimit = -1;
66
67 extern char **environ;
68
69 static int run_command(struct cfjail *j);
70 static int add_proc(struct cfjail *j, pid_t pid);
71 static void clear_procs(struct cfjail *j);
72 static struct cfjail *find_proc(pid_t pid);
73 static int term_procs(struct cfjail *j);
74 static int get_user_info(struct cfjail *j, const char *username,
75 const struct passwd **pwdp, login_cap_t **lcapp);
76 static int check_path(struct cfjail *j, const char *pname, const char *path,
77 int isfile, const char *umount_type);
78
79 static struct cfjails sleeping = TAILQ_HEAD_INITIALIZER(sleeping);
80 static struct cfjails runnable = TAILQ_HEAD_INITIALIZER(runnable);
81 static struct cfstring dummystring = { .len = 1 };
82 static struct phhead phash[PHASH_SIZE];
83 static int kq;
84
85 static cpusetid_t
root_cpuset_id(void)86 root_cpuset_id(void)
87 {
88 static cpusetid_t setid = CPUSET_INVALID;
89 static int error;
90
91 /* Only try to get the cpuset once. */
92 if (error == 0 && setid == CPUSET_INVALID)
93 error = cpuset_getid(CPU_LEVEL_ROOT, CPU_WHICH_PID, -1, &setid);
94 if (error != 0)
95 return (CPUSET_INVALID);
96 return (setid);
97 }
98
99 /*
100 * Run the next command associated with a jail.
101 */
102 int
next_command(struct cfjail * j)103 next_command(struct cfjail *j)
104 {
105 enum intparam comparam;
106 int create_failed, stopping;
107
108 if (paralimit == 0) {
109 if (j->flags & JF_FROM_RUNQ)
110 requeue_head(j, &runnable);
111 else
112 requeue(j, &runnable);
113 return 1;
114 }
115 j->flags &= ~JF_FROM_RUNQ;
116 create_failed = (j->flags & (JF_STOP | JF_FAILED)) == JF_FAILED;
117 stopping = (j->flags & JF_STOP) != 0;
118 comparam = *j->comparam;
119 for (;;) {
120 if (j->comstring == NULL) {
121 j->comparam += create_failed ? -1 : 1;
122 switch ((comparam = *j->comparam)) {
123 case IP__NULL:
124 return 0;
125 case IP_MOUNT_DEVFS:
126 if (!bool_param(j->intparams[IP_MOUNT_DEVFS]))
127 continue;
128 j->comstring = &dummystring;
129 break;
130 case IP_MOUNT_FDESCFS:
131 if (!bool_param(j->intparams[IP_MOUNT_FDESCFS]))
132 continue;
133 j->comstring = &dummystring;
134 break;
135 case IP_MOUNT_PROCFS:
136 if (!bool_param(j->intparams[IP_MOUNT_PROCFS]))
137 continue;
138 j->comstring = &dummystring;
139 break;
140 case IP__OP:
141 case IP_STOP_TIMEOUT:
142 j->comstring = &dummystring;
143 break;
144 default:
145 if (j->intparams[comparam] == NULL)
146 continue;
147 j->comstring = create_failed || (stopping &&
148 (j->intparams[comparam]->flags & PF_REV))
149 ? TAILQ_LAST(&j->intparams[comparam]->val,
150 cfstrings)
151 : TAILQ_FIRST(&j->intparams[comparam]->val);
152 }
153 } else {
154 j->comstring = j->comstring == &dummystring ? NULL :
155 create_failed || (stopping &&
156 (j->intparams[comparam]->flags & PF_REV))
157 ? TAILQ_PREV(j->comstring, cfstrings, tq)
158 : TAILQ_NEXT(j->comstring, tq);
159 }
160 if (j->comstring == NULL || j->comstring->len == 0 ||
161 (create_failed && (comparam == IP_EXEC_PRESTART ||
162 comparam == IP_EXEC_CREATED || comparam == IP_EXEC_START ||
163 comparam == IP_COMMAND || comparam == IP_EXEC_POSTSTART ||
164 comparam == IP_EXEC_PREPARE)))
165 continue;
166 switch (run_command(j)) {
167 case -1:
168 failed(j);
169 /* FALLTHROUGH */
170 case 1:
171 return 1;
172 }
173 }
174 }
175
176 /*
177 * Check command exit status
178 */
179 int
finish_command(struct cfjail * j)180 finish_command(struct cfjail *j)
181 {
182 struct cfjail *rj;
183 int error;
184
185 if (!(j->flags & JF_SLEEPQ))
186 return 0;
187 j->flags &= ~JF_SLEEPQ;
188 if (*j->comparam == IP_STOP_TIMEOUT) {
189 j->flags &= ~JF_TIMEOUT;
190 j->pstatus = 0;
191 return 0;
192 }
193 paralimit++;
194 if (!TAILQ_EMPTY(&runnable)) {
195 rj = TAILQ_FIRST(&runnable);
196 rj->flags |= JF_FROM_RUNQ;
197 requeue(rj, &ready);
198 }
199 error = 0;
200 if (j->flags & JF_TIMEOUT) {
201 j->flags &= ~JF_TIMEOUT;
202 if (*j->comparam != IP_STOP_TIMEOUT) {
203 jail_warnx(j, "%s: timed out", j->comline);
204 failed(j);
205 error = -1;
206 } else if (verbose > 0)
207 jail_note(j, "timed out\n");
208 } else if (j->pstatus != 0) {
209 if (WIFSIGNALED(j->pstatus))
210 jail_warnx(j, "%s: exited on signal %d",
211 j->comline, WTERMSIG(j->pstatus));
212 else
213 jail_warnx(j, "%s: failed", j->comline);
214 j->pstatus = 0;
215 failed(j);
216 error = -1;
217 }
218 free(j->comline);
219 j->comline = NULL;
220 return error;
221 }
222
223 /*
224 * Check for finished processes or timeouts.
225 */
226 struct cfjail *
next_proc(int nonblock)227 next_proc(int nonblock)
228 {
229 struct kevent ke;
230 struct timespec ts;
231 struct timespec *tsp;
232 struct cfjail *j;
233
234 if (!TAILQ_EMPTY(&sleeping)) {
235 again:
236 tsp = NULL;
237 if ((j = TAILQ_FIRST(&sleeping)) && j->timeout.tv_sec) {
238 clock_gettime(CLOCK_REALTIME, &ts);
239 ts.tv_sec = j->timeout.tv_sec - ts.tv_sec;
240 ts.tv_nsec = j->timeout.tv_nsec - ts.tv_nsec;
241 if (ts.tv_nsec < 0) {
242 ts.tv_sec--;
243 ts.tv_nsec += 1000000000;
244 }
245 if (ts.tv_sec < 0 ||
246 (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
247 j->flags |= JF_TIMEOUT;
248 clear_procs(j);
249 return j;
250 }
251 tsp = &ts;
252 }
253 if (nonblock) {
254 ts.tv_sec = 0;
255 ts.tv_nsec = 0;
256 tsp = &ts;
257 }
258 switch (kevent(kq, NULL, 0, &ke, 1, tsp)) {
259 case -1:
260 if (errno != EINTR)
261 err(1, "kevent");
262 goto again;
263 case 0:
264 if (!nonblock) {
265 j = TAILQ_FIRST(&sleeping);
266 j->flags |= JF_TIMEOUT;
267 clear_procs(j);
268 return j;
269 }
270 break;
271 case 1:
272 (void)waitpid(ke.ident, NULL, WNOHANG);
273 if ((j = find_proc(ke.ident))) {
274 j->pstatus = ke.data;
275 return j;
276 }
277 goto again;
278 }
279 }
280 return NULL;
281 }
282
283 /*
284 * Run a single command for a jail, possibly inside the jail.
285 */
286 static int
run_command(struct cfjail * j)287 run_command(struct cfjail *j)
288 {
289 const struct passwd *pwd;
290 const struct cfstring *comstring, *s;
291 login_cap_t *lcap;
292 const char **argv;
293 char *acs, *cs, *comcs, *devpath;
294 const char *jidstr, *conslog, *fmt, *path, *ruleset, *term, *username;
295 enum intparam comparam;
296 size_t comlen, ret;
297 pid_t pid;
298 cpusetid_t setid;
299 int argc, bg, clean, consfd, down, fib, i, injail, sjuser, timeout;
300 #if defined(INET) || defined(INET6)
301 char *addr, *extrap, *p, *val;
302 #endif
303
304 static char *cleanenv;
305
306 /* Perform some operations that aren't actually commands */
307 comparam = *j->comparam;
308 down = j->flags & (JF_STOP | JF_FAILED);
309 switch (comparam) {
310 case IP_STOP_TIMEOUT:
311 return term_procs(j);
312
313 case IP__OP:
314 if (down) {
315 if (jail_remove(j->jid) < 0 && errno == EPERM) {
316 jail_warnx(j, "jail_remove: %s",
317 strerror(errno));
318 return -1;
319 }
320 if (verbose > 0 || (verbose == 0 && (j->flags & JF_STOP
321 ? note_remove : j->name != NULL)))
322 jail_note(j, "removed\n");
323 j->jid = -1;
324 if (j->flags & JF_STOP)
325 dep_done(j, DF_LIGHT);
326 else
327 j->flags &= ~JF_PERSIST;
328 } else {
329 if (create_jail(j) < 0)
330 return -1;
331 if (iflag)
332 printf("%d\n", j->jid);
333 if (verbose >= 0 && (j->name || verbose > 0))
334 jail_note(j, "created\n");
335
336 /*
337 * Populate our jid and name parameters if they were not
338 * provided. This simplifies later logic that wants to
339 * use the jid or name to be able to do so reliably.
340 */
341 if (j->intparams[KP_JID] == NULL) {
342 char ljidstr[16];
343
344 (void)snprintf(ljidstr, sizeof(ljidstr), "%d",
345 j->jid);
346 add_param(j, NULL, KP_JID, ljidstr);
347 }
348
349 /* This matches the kernel behavior. */
350 if (j->intparams[KP_NAME] == NULL)
351 add_param(j, j->intparams[KP_JID], KP_NAME,
352 NULL);
353
354 dep_done(j, DF_LIGHT);
355 }
356 return 0;
357
358 default: ;
359 }
360 /*
361 * Collect exec arguments. Internal commands for network and
362 * mounting build their own argument lists.
363 */
364 comstring = j->comstring;
365 bg = 0;
366 switch (comparam) {
367 #ifdef INET
368 case IP__IP4_IFADDR:
369 argc = 0;
370 val = alloca(strlen(comstring->s) + 1);
371 strcpy(val, comstring->s);
372 cs = val;
373 extrap = NULL;
374 while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
375 if (extrap == NULL) {
376 *p = '\0';
377 extrap = p + 1;
378 }
379 cs = p + 1;
380 argc++;
381 }
382
383 argv = alloca((8 + argc) * sizeof(char *));
384 argv[0] = _PATH_IFCONFIG;
385 if ((cs = strchr(val, '|'))) {
386 argv[1] = acs = alloca(cs - val + 1);
387 strlcpy(acs, val, cs - val + 1);
388 addr = cs + 1;
389 } else {
390 argv[1] = string_param(j->intparams[IP_INTERFACE]);
391 addr = val;
392 }
393 argv[2] = "inet";
394 if (!(cs = strchr(addr, '/'))) {
395 argv[3] = addr;
396 argv[4] = "netmask";
397 argv[5] = "255.255.255.255";
398 argc = 6;
399 } else if (strchr(cs + 1, '.')) {
400 argv[3] = acs = alloca(cs - addr + 1);
401 strlcpy(acs, addr, cs - addr + 1);
402 argv[4] = "netmask";
403 argv[5] = cs + 1;
404 argc = 6;
405 } else {
406 argv[3] = addr;
407 argc = 4;
408 }
409
410 if (!down && extrap != NULL) {
411 for (cs = strtok(extrap, " "); cs;
412 cs = strtok(NULL, " ")) {
413 size_t len = strlen(cs) + 1;
414 argv[argc++] = acs = alloca(len);
415 strlcpy(acs, cs, len);
416 }
417 }
418
419 argv[argc] = down ? "-alias" : "alias";
420 argv[argc + 1] = NULL;
421 break;
422 #endif
423
424 #ifdef INET6
425 case IP__IP6_IFADDR:
426 argc = 0;
427 val = alloca(strlen(comstring->s) + 1);
428 strcpy(val, comstring->s);
429 cs = val;
430 extrap = NULL;
431 while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
432 if (extrap == NULL) {
433 *p = '\0';
434 extrap = p + 1;
435 }
436 cs = p + 1;
437 argc++;
438 }
439
440 argv = alloca((8 + argc) * sizeof(char *));
441 argv[0] = _PATH_IFCONFIG;
442 if ((cs = strchr(val, '|'))) {
443 argv[1] = acs = alloca(cs - val + 1);
444 strlcpy(acs, val, cs - val + 1);
445 addr = cs + 1;
446 } else {
447 argv[1] = string_param(j->intparams[IP_INTERFACE]);
448 addr = val;
449 }
450 argv[2] = "inet6";
451 argv[3] = addr;
452 if (!(cs = strchr(addr, '/'))) {
453 argv[4] = "prefixlen";
454 argv[5] = "128";
455 argc = 6;
456 } else
457 argc = 4;
458
459 if (!down && extrap != NULL) {
460 for (cs = strtok(extrap, " "); cs;
461 cs = strtok(NULL, " ")) {
462 size_t len = strlen(cs) + 1;
463 argv[argc++] = acs = alloca(len);
464 strlcpy(acs, cs, len);
465 }
466 }
467
468 argv[argc] = down ? "-alias" : "alias";
469 argv[argc + 1] = NULL;
470 break;
471 #endif
472
473 case IP_VNET_INTERFACE:
474 argv = alloca(5 * sizeof(char *));
475 argv[0] = _PATH_IFCONFIG;
476 argv[1] = comstring->s;
477 argv[2] = down ? "-vnet" : "vnet";
478 argv[3] = string_param(j->intparams[KP_JID]);
479 argv[4] = NULL;
480 break;
481
482 case IP_MOUNT:
483 case IP__MOUNT_FROM_FSTAB:
484 argv = alloca(8 * sizeof(char *));
485 comcs = alloca(comstring->len + 1);
486 strcpy(comcs, comstring->s);
487 argc = 0;
488 for (cs = strtok(comcs, " \t\f\v\r\n"); cs && argc < 4;
489 cs = strtok(NULL, " \t\f\v\r\n")) {
490 if (argc <= 1 && strunvis(cs, cs) < 0) {
491 jail_warnx(j, "%s: %s: fstab parse error",
492 j->intparams[comparam]->name, comstring->s);
493 return -1;
494 }
495 argv[argc++] = cs;
496 }
497 if (argc == 0)
498 return 0;
499 if (argc < 3) {
500 jail_warnx(j, "%s: %s: missing information",
501 j->intparams[comparam]->name, comstring->s);
502 return -1;
503 }
504 if (check_path(j, j->intparams[comparam]->name, argv[1], 0,
505 down ? argv[2] : NULL) < 0)
506 return -1;
507 if (down) {
508 argv[4] = NULL;
509 argv[3] = argv[1];
510 argv[0] = "/sbin/umount";
511 } else {
512 if (argc == 4) {
513 argv[7] = NULL;
514 argv[6] = argv[1];
515 argv[5] = argv[0];
516 argv[4] = argv[3];
517 argv[3] = "-o";
518 } else {
519 argv[5] = NULL;
520 argv[4] = argv[1];
521 argv[3] = argv[0];
522 }
523 argv[0] = _PATH_MOUNT;
524 }
525 argv[1] = "-t";
526 break;
527
528 case IP_MOUNT_DEVFS:
529 argv = alloca(7 * sizeof(char *));
530 path = string_param(j->intparams[KP_PATH]);
531 if (path == NULL) {
532 jail_warnx(j, "mount.devfs: no jail root path defined");
533 return -1;
534 }
535 devpath = alloca(strlen(path) + 5);
536 sprintf(devpath, "%s/dev", path);
537 if (check_path(j, "mount.devfs", devpath, 0,
538 down ? "devfs" : NULL) < 0)
539 return -1;
540 if (down) {
541 argv[0] = "/sbin/umount";
542 argv[1] = devpath;
543 argv[2] = NULL;
544 } else {
545 argv[0] = _PATH_MOUNT;
546 argv[1] = "-t";
547 argv[2] = "devfs";
548 ruleset = string_param(j->intparams[KP_DEVFS_RULESET]);
549 if (!ruleset)
550 ruleset = "4"; /* devfsrules_jail */
551 argv[3] = acs = alloca(11 + strlen(ruleset));
552 sprintf(acs, "-oruleset=%s", ruleset);
553 argv[4] = ".";
554 argv[5] = devpath;
555 argv[6] = NULL;
556 }
557 break;
558
559 case IP_MOUNT_FDESCFS:
560 argv = alloca(7 * sizeof(char *));
561 path = string_param(j->intparams[KP_PATH]);
562 if (path == NULL) {
563 jail_warnx(j, "mount.fdescfs: no jail root path defined");
564 return -1;
565 }
566 devpath = alloca(strlen(path) + 8);
567 sprintf(devpath, "%s/dev/fd", path);
568 if (check_path(j, "mount.fdescfs", devpath, 0,
569 down ? "fdescfs" : NULL) < 0)
570 return -1;
571 if (down) {
572 argv[0] = "/sbin/umount";
573 argv[1] = devpath;
574 argv[2] = NULL;
575 } else {
576 argv[0] = _PATH_MOUNT;
577 argv[1] = "-t";
578 argv[2] = "fdescfs";
579 argv[3] = ".";
580 argv[4] = devpath;
581 argv[5] = NULL;
582 }
583 break;
584
585 case IP_MOUNT_PROCFS:
586 argv = alloca(7 * sizeof(char *));
587 path = string_param(j->intparams[KP_PATH]);
588 if (path == NULL) {
589 jail_warnx(j, "mount.procfs: no jail root path defined");
590 return -1;
591 }
592 devpath = alloca(strlen(path) + 6);
593 sprintf(devpath, "%s/proc", path);
594 if (check_path(j, "mount.procfs", devpath, 0,
595 down ? "procfs" : NULL) < 0)
596 return -1;
597 if (down) {
598 argv[0] = "/sbin/umount";
599 argv[1] = devpath;
600 argv[2] = NULL;
601 } else {
602 argv[0] = _PATH_MOUNT;
603 argv[1] = "-t";
604 argv[2] = "procfs";
605 argv[3] = ".";
606 argv[4] = devpath;
607 argv[5] = NULL;
608 }
609 break;
610
611 case IP_ZFS_DATASET:
612 argv = alloca(4 * sizeof(char *));
613 jidstr = string_param(j->intparams[KP_JID]);
614 fmt = "if [ $(/sbin/zfs get -H -o value jailed %s) = on ]; then /sbin/zfs jail %s %s || echo error, attaching %s to jail %s failed; else echo error, you need to set jailed=on for dataset %s; fi";
615 comlen = strlen(fmt)
616 + 2 * strlen(jidstr)
617 + 4 * comstring->len
618 - 6 * 2 /* 6 * "%s" */
619 + 1;
620 comcs = alloca(comlen);
621 ret = snprintf(comcs, comlen, fmt, comstring->s,
622 jidstr, comstring->s, comstring->s, jidstr,
623 comstring->s);
624 if (ret >= comlen) {
625 jail_warnx(j, "internal error in ZFS dataset handling");
626 exit(1);
627 }
628 argv[0] = _PATH_BSHELL;
629 argv[1] = "-c";
630 argv[2] = comcs;
631 argv[3] = NULL;
632 break;
633
634 case IP_COMMAND:
635 if (j->name != NULL)
636 goto default_command;
637 argc = 0;
638 TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
639 argc++;
640 argv = alloca((argc + 1) * sizeof(char *));
641 argc = 0;
642 TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
643 argv[argc++] = s->s;
644 argv[argc] = NULL;
645 j->comstring = &dummystring;
646 break;
647
648 default:
649 default_command:
650 if ((cs = strpbrk(comstring->s, "!\"$&'()*;<>?[\\]`{|}~")) &&
651 !(cs[0] == '&' && cs[1] == '\0')) {
652 argv = alloca(4 * sizeof(char *));
653 argv[0] = _PATH_BSHELL;
654 argv[1] = "-c";
655 argv[2] = comstring->s;
656 argv[3] = NULL;
657 } else {
658 if (cs) {
659 *cs = 0;
660 bg = 1;
661 }
662 comcs = alloca(comstring->len + 1);
663 strcpy(comcs, comstring->s);
664 argc = 0;
665 for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
666 cs = strtok(NULL, " \t\f\v\r\n"))
667 argc++;
668 argv = alloca((argc + 1) * sizeof(char *));
669 strcpy(comcs, comstring->s);
670 argc = 0;
671 for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
672 cs = strtok(NULL, " \t\f\v\r\n"))
673 argv[argc++] = cs;
674 argv[argc] = NULL;
675 }
676 }
677 if (argv[0] == NULL)
678 return 0;
679
680 if (int_param(j->intparams[IP_EXEC_TIMEOUT], &timeout) &&
681 timeout != 0) {
682 clock_gettime(CLOCK_REALTIME, &j->timeout);
683 j->timeout.tv_sec += timeout;
684 } else
685 j->timeout.tv_sec = 0;
686
687 injail = comparam == IP_EXEC_START || comparam == IP_COMMAND ||
688 comparam == IP_EXEC_STOP;
689 if (injail)
690 setid = root_cpuset_id();
691 else
692 setid = CPUSET_INVALID;
693 clean = bool_param(j->intparams[IP_EXEC_CLEAN]);
694 username = string_param(j->intparams[injail
695 ? IP_EXEC_JAIL_USER : IP_EXEC_SYSTEM_USER]);
696 sjuser = bool_param(j->intparams[IP_EXEC_SYSTEM_JAIL_USER]);
697
698 consfd = 0;
699 if (injail &&
700 (conslog = string_param(j->intparams[IP_EXEC_CONSOLELOG]))) {
701 if (check_path(j, "exec.consolelog", conslog, 1, NULL) < 0)
702 return -1;
703 consfd =
704 open(conslog, O_WRONLY | O_CREAT | O_APPEND, DEFFILEMODE);
705 if (consfd < 0) {
706 jail_warnx(j, "open %s: %s", conslog, strerror(errno));
707 return -1;
708 }
709 }
710
711 comlen = 0;
712 for (i = 0; argv[i]; i++)
713 comlen += strlen(argv[i]) + 1;
714 j->comline = cs = emalloc(comlen);
715 for (i = 0; argv[i]; i++) {
716 strcpy(cs, argv[i]);
717 if (argv[i + 1]) {
718 cs += strlen(argv[i]) + 1;
719 cs[-1] = ' ';
720 }
721 }
722 if (verbose > 0)
723 jail_note(j, "run command%s%s%s: %s\n",
724 injail ? " in jail" : "", username ? " as " : "",
725 username ? username : "", j->comline);
726
727 pid = fork();
728 if (pid < 0)
729 err(1, "fork");
730 if (pid > 0) {
731 if (bg || !add_proc(j, pid)) {
732 free(j->comline);
733 j->comline = NULL;
734 return 0;
735 } else {
736 paralimit--;
737 return 1;
738 }
739 }
740 if (bg)
741 setsid();
742
743 /* Set up the environment and run the command */
744 pwd = NULL;
745 lcap = NULL;
746 if ((clean || username) && injail && sjuser &&
747 get_user_info(j, username, &pwd, &lcap) < 0)
748 exit(1);
749 if (injail) {
750 /* jail_attach won't chdir along with its chroot. */
751 path = string_param(j->intparams[KP_PATH]);
752 if (path && chdir(path) < 0) {
753 jail_warnx(j, "chdir %s: %s", path, strerror(errno));
754 exit(1);
755 }
756 if (int_param(j->intparams[IP_EXEC_FIB], &fib) &&
757 setfib(fib) < 0) {
758 jail_warnx(j, "setfib: %s", strerror(errno));
759 exit(1);
760 }
761
762 /*
763 * We wouldn't have specialized our affinity, so just setid to
764 * root. We do this prior to attaching to avoid the kernel
765 * having to create a transient cpuset that we'll promptly
766 * free up with a reset to the jail's cpuset.
767 *
768 * This is just a best-effort to use as wide of mask as
769 * possible.
770 */
771 if (setid != CPUSET_INVALID)
772 (void)cpuset_setid(CPU_WHICH_PID, -1, setid);
773
774 if (jail_attach(j->jid) < 0) {
775 jail_warnx(j, "jail_attach: %s", strerror(errno));
776 exit(1);
777 }
778 }
779 if (clean || username) {
780 if (!(injail && sjuser) &&
781 get_user_info(j, username, &pwd, &lcap) < 0)
782 exit(1);
783 if (clean) {
784 term = getenv("TERM");
785 environ = &cleanenv;
786 setenv("PATH", "/bin:/usr/bin", 0);
787 if (term != NULL)
788 setenv("TERM", term, 1);
789 }
790 if (setgid(pwd->pw_gid) < 0) {
791 jail_warnx(j, "setgid %d: %s", pwd->pw_gid,
792 strerror(errno));
793 exit(1);
794 }
795 if (setusercontext(lcap, pwd, pwd->pw_uid, username
796 ? LOGIN_SETALL & ~LOGIN_SETGROUP & ~LOGIN_SETLOGIN
797 : LOGIN_SETPATH | LOGIN_SETENV) < 0) {
798 jail_warnx(j, "setusercontext %s: %s", pwd->pw_name,
799 strerror(errno));
800 exit(1);
801 }
802 login_close(lcap);
803 setenv("USER", pwd->pw_name, 1);
804 setenv("HOME", pwd->pw_dir, 1);
805 setenv("SHELL",
806 *pwd->pw_shell ? pwd->pw_shell : _PATH_BSHELL, 1);
807 if (clean && username && chdir(pwd->pw_dir) < 0) {
808 jail_warnx(j, "chdir %s: %s",
809 pwd->pw_dir, strerror(errno));
810 exit(1);
811 }
812 endpwent();
813 }
814 if (!injail) {
815 if (string_param(j->intparams[KP_JID]))
816 setenv("JID", string_param(j->intparams[KP_JID]), 1);
817 setenv("JNAME", string_param(j->intparams[KP_NAME]), 1);
818
819 path = string_param(j->intparams[KP_PATH]);
820 setenv("JPATH", path ? path : "", 1);
821 }
822
823 if (consfd != 0 && (dup2(consfd, 1) < 0 || dup2(consfd, 2) < 0)) {
824 jail_warnx(j, "exec.consolelog: %s", strerror(errno));
825 exit(1);
826 }
827 closefrom(3);
828 execvp(argv[0], __DECONST(char *const*, argv));
829 jail_warnx(j, "exec %s: %s", argv[0], strerror(errno));
830 exit(1);
831 }
832
833 /*
834 * Add a process to the hash, tied to a jail.
835 */
836 static int
add_proc(struct cfjail * j,pid_t pid)837 add_proc(struct cfjail *j, pid_t pid)
838 {
839 struct kevent ke;
840 struct cfjail *tj;
841 struct phash *ph;
842
843 if (!kq && (kq = kqueue()) < 0)
844 err(1, "kqueue");
845 EV_SET(&ke, pid, EVFILT_PROC, EV_ADD, NOTE_EXIT, 0, NULL);
846 if (kevent(kq, &ke, 1, NULL, 0, NULL) < 0) {
847 if (errno == ESRCH)
848 return 0;
849 err(1, "kevent");
850 }
851 ph = emalloc(sizeof(struct phash));
852 ph->j = j;
853 ph->pid = pid;
854 LIST_INSERT_HEAD(&phash[pid % PHASH_SIZE], ph, le);
855 j->nprocs++;
856 j->flags |= JF_SLEEPQ;
857 if (j->timeout.tv_sec == 0)
858 requeue(j, &sleeping);
859 else {
860 /* File the jail in the sleep queue according to its timeout. */
861 TAILQ_REMOVE(j->queue, j, tq);
862 TAILQ_FOREACH(tj, &sleeping, tq) {
863 if (!tj->timeout.tv_sec ||
864 j->timeout.tv_sec < tj->timeout.tv_sec ||
865 (j->timeout.tv_sec == tj->timeout.tv_sec &&
866 j->timeout.tv_nsec <= tj->timeout.tv_nsec)) {
867 TAILQ_INSERT_BEFORE(tj, j, tq);
868 break;
869 }
870 }
871 if (tj == NULL)
872 TAILQ_INSERT_TAIL(&sleeping, j, tq);
873 j->queue = &sleeping;
874 }
875 return 1;
876 }
877
878 /*
879 * Remove any processes from the hash that correspond to a jail.
880 */
881 static void
clear_procs(struct cfjail * j)882 clear_procs(struct cfjail *j)
883 {
884 struct kevent ke;
885 struct phash *ph, *tph;
886 int i;
887
888 j->nprocs = 0;
889 for (i = 0; i < PHASH_SIZE; i++)
890 LIST_FOREACH_SAFE(ph, &phash[i], le, tph)
891 if (ph->j == j) {
892 EV_SET(&ke, ph->pid, EVFILT_PROC, EV_DELETE,
893 NOTE_EXIT, 0, NULL);
894 (void)kevent(kq, &ke, 1, NULL, 0, NULL);
895 LIST_REMOVE(ph, le);
896 free(ph);
897 }
898 }
899
900 /*
901 * Find the jail that corresponds to an exited process.
902 */
903 static struct cfjail *
find_proc(pid_t pid)904 find_proc(pid_t pid)
905 {
906 struct cfjail *j;
907 struct phash *ph;
908
909 LIST_FOREACH(ph, &phash[pid % PHASH_SIZE], le)
910 if (ph->pid == pid) {
911 j = ph->j;
912 LIST_REMOVE(ph, le);
913 free(ph);
914 return --j->nprocs ? NULL : j;
915 }
916 return NULL;
917 }
918
919 /*
920 * Send SIGTERM to all processes in a jail and wait for them to die.
921 */
922 static int
term_procs(struct cfjail * j)923 term_procs(struct cfjail *j)
924 {
925 struct kinfo_proc *ki;
926 int i, noted, pcnt, timeout;
927
928 static kvm_t *kd;
929
930 if (!int_param(j->intparams[IP_STOP_TIMEOUT], &timeout))
931 timeout = DEFAULT_STOP_TIMEOUT;
932 else if (timeout == 0)
933 return 0;
934
935 if (kd == NULL) {
936 kd = kvm_open(NULL, NULL, NULL, O_RDONLY, NULL);
937 if (kd == NULL)
938 return 0;
939 }
940
941 ki = kvm_getprocs(kd, KERN_PROC_PROC, 0, &pcnt);
942 if (ki == NULL)
943 return 0;
944 noted = 0;
945 for (i = 0; i < pcnt; i++)
946 if (ki[i].ki_jid == j->jid &&
947 kill(ki[i].ki_pid, SIGTERM) == 0) {
948 (void)add_proc(j, ki[i].ki_pid);
949 if (verbose > 0) {
950 if (!noted) {
951 noted = 1;
952 jail_note(j, "sent SIGTERM to:");
953 }
954 printf(" %d", ki[i].ki_pid);
955 }
956 }
957 if (noted)
958 printf("\n");
959 if (j->nprocs > 0) {
960 clock_gettime(CLOCK_REALTIME, &j->timeout);
961 j->timeout.tv_sec += timeout;
962 return 1;
963 }
964 return 0;
965 }
966
967 /*
968 * Look up a user in the passwd and login.conf files.
969 */
970 static int
get_user_info(struct cfjail * j,const char * username,const struct passwd ** pwdp,login_cap_t ** lcapp)971 get_user_info(struct cfjail *j, const char *username,
972 const struct passwd **pwdp, login_cap_t **lcapp)
973 {
974 const struct passwd *pwd;
975
976 errno = 0;
977 *pwdp = pwd = username ? getpwnam(username) : getpwuid(getuid());
978 if (pwd == NULL) {
979 if (errno)
980 jail_warnx(j, "getpwnam%s%s: %s", username ? " " : "",
981 username ? username : "", strerror(errno));
982 else if (username)
983 jail_warnx(j, "%s: no such user", username);
984 else
985 jail_warnx(j, "unknown uid %d", getuid());
986 return -1;
987 }
988 *lcapp = login_getpwclass(pwd);
989 if (*lcapp == NULL) {
990 jail_warnx(j, "getpwclass %s: %s", pwd->pw_name,
991 strerror(errno));
992 return -1;
993 }
994 /* Set the groups while the group file is still available */
995 if (initgroups(pwd->pw_name, pwd->pw_gid) < 0) {
996 jail_warnx(j, "initgroups %s: %s", pwd->pw_name,
997 strerror(errno));
998 return -1;
999 }
1000 return 0;
1001 }
1002
1003 /*
1004 * Make sure a mount or consolelog path is a valid absolute pathname
1005 * with no symlinks.
1006 */
1007 static int
check_path(struct cfjail * j,const char * pname,const char * path,int isfile,const char * umount_type)1008 check_path(struct cfjail *j, const char *pname, const char *path, int isfile,
1009 const char *umount_type)
1010 {
1011 struct stat st, mpst;
1012 struct statfs stfs;
1013 char *tpath, *p;
1014 const char *jailpath;
1015 size_t jplen;
1016
1017 if (path[0] != '/') {
1018 jail_warnx(j, "%s: %s: not an absolute pathname",
1019 pname, path);
1020 return -1;
1021 }
1022 /*
1023 * Only check for symlinks in components below the jail's path,
1024 * since that's where the security risk lies.
1025 */
1026 jailpath = string_param(j->intparams[KP_PATH]);
1027 if (jailpath == NULL)
1028 jailpath = "";
1029 jplen = strlen(jailpath);
1030 if (!strncmp(path, jailpath, jplen) && path[jplen] == '/') {
1031 tpath = alloca(strlen(path) + 1);
1032 strcpy(tpath, path);
1033 for (p = tpath + jplen; p != NULL; ) {
1034 p = strchr(p + 1, '/');
1035 if (p)
1036 *p = '\0';
1037 if (lstat(tpath, &st) < 0) {
1038 if (errno == ENOENT && isfile && !p)
1039 break;
1040 jail_warnx(j, "%s: %s: %s", pname, tpath,
1041 strerror(errno));
1042 return -1;
1043 }
1044 if (S_ISLNK(st.st_mode)) {
1045 jail_warnx(j, "%s: %s is a symbolic link",
1046 pname, tpath);
1047 return -1;
1048 }
1049 if (p)
1050 *p = '/';
1051 }
1052 }
1053 if (umount_type != NULL) {
1054 if (stat(path, &st) < 0 || statfs(path, &stfs) < 0) {
1055 jail_warnx(j, "%s: %s: %s", pname, path,
1056 strerror(errno));
1057 return -1;
1058 }
1059 if (stat(stfs.f_mntonname, &mpst) < 0) {
1060 jail_warnx(j, "%s: %s: %s", pname, stfs.f_mntonname,
1061 strerror(errno));
1062 return -1;
1063 }
1064 if (st.st_ino != mpst.st_ino) {
1065 jail_warnx(j, "%s: %s: not a mount point",
1066 pname, path);
1067 return -1;
1068 }
1069 if (strcmp(stfs.f_fstypename, umount_type)) {
1070 jail_warnx(j, "%s: %s: not a %s mount",
1071 pname, path, umount_type);
1072 return -1;
1073 }
1074 }
1075 return 0;
1076 }
1077