xref: /freebsd/usr.sbin/jail/command.c (revision 18849b5da0c5eaa88500b457be05b038813b51b1)
1 /*-
2  * Copyright (c) 2011 James Gritton
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26 
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
29 
30 #include <sys/types.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 /*
86  * Run the next command associated with a jail.
87  */
88 int
89 next_command(struct cfjail *j)
90 {
91 	enum intparam comparam;
92 	int create_failed, stopping;
93 
94 	if (paralimit == 0) {
95 		requeue(j, &runnable);
96 		return 1;
97 	}
98 	create_failed = (j->flags & (JF_STOP | JF_FAILED)) == JF_FAILED;
99 	stopping = (j->flags & JF_STOP) != 0;
100 	comparam = *j->comparam;
101 	for (;;) {
102 		if (j->comstring == NULL) {
103 			j->comparam += create_failed ? -1 : 1;
104 			switch ((comparam = *j->comparam)) {
105 			case IP__NULL:
106 				return 0;
107 			case IP_MOUNT_DEVFS:
108 				if (!bool_param(j->intparams[IP_MOUNT_DEVFS]))
109 					continue;
110 				j->comstring = &dummystring;
111 				break;
112 			case IP_MOUNT_FDESCFS:
113 				if (!bool_param(j->intparams[IP_MOUNT_FDESCFS]))
114 					continue;
115 				j->comstring = &dummystring;
116 				break;
117 			case IP_MOUNT_PROCFS:
118 				if (!bool_param(j->intparams[IP_MOUNT_PROCFS]))
119 					continue;
120 				j->comstring = &dummystring;
121 				break;
122 			case IP__OP:
123 			case IP_STOP_TIMEOUT:
124 				j->comstring = &dummystring;
125 				break;
126 			default:
127 				if (j->intparams[comparam] == NULL)
128 					continue;
129 				j->comstring = create_failed || (stopping &&
130 				    (j->intparams[comparam]->flags & PF_REV))
131 				    ? TAILQ_LAST(&j->intparams[comparam]->val,
132 					cfstrings)
133 				    : TAILQ_FIRST(&j->intparams[comparam]->val);
134 			}
135 		} else {
136 			j->comstring = j->comstring == &dummystring ? NULL :
137 			    create_failed || (stopping &&
138 			    (j->intparams[comparam]->flags & PF_REV))
139 			    ? TAILQ_PREV(j->comstring, cfstrings, tq)
140 			    : TAILQ_NEXT(j->comstring, tq);
141 		}
142 		if (j->comstring == NULL || j->comstring->len == 0 ||
143 		    (create_failed && (comparam == IP_EXEC_PRESTART ||
144 		    comparam == IP_EXEC_START || comparam == IP_COMMAND ||
145 		    comparam == IP_EXEC_POSTSTART)))
146 			continue;
147 		switch (run_command(j)) {
148 		case -1:
149 			failed(j);
150 			/* FALLTHROUGH */
151 		case 1:
152 			return 1;
153 		}
154 	}
155 }
156 
157 /*
158  * Check command exit status
159  */
160 int
161 finish_command(struct cfjail *j)
162 {
163 	int error;
164 
165 	if (!(j->flags & JF_SLEEPQ))
166 		return 0;
167 	j->flags &= ~JF_SLEEPQ;
168 	if (*j->comparam == IP_STOP_TIMEOUT)
169 	{
170 		j->flags &= ~JF_TIMEOUT;
171 		j->pstatus = 0;
172 		return 0;
173 	}
174 	paralimit++;
175 	if (!TAILQ_EMPTY(&runnable))
176 		requeue(TAILQ_FIRST(&runnable), &ready);
177 	error = 0;
178 	if (j->flags & JF_TIMEOUT) {
179 		j->flags &= ~JF_TIMEOUT;
180 		if (*j->comparam != IP_STOP_TIMEOUT) {
181 			jail_warnx(j, "%s: timed out", j->comline);
182 			failed(j);
183 			error = -1;
184 		} else if (verbose > 0)
185 			jail_note(j, "timed out\n");
186 	} else if (j->pstatus != 0) {
187 		if (WIFSIGNALED(j->pstatus))
188 			jail_warnx(j, "%s: exited on signal %d",
189 			    j->comline, WTERMSIG(j->pstatus));
190 		else
191 			jail_warnx(j, "%s: failed", j->comline);
192 		j->pstatus = 0;
193 		failed(j);
194 		error = -1;
195 	}
196 	free(j->comline);
197 	j->comline = NULL;
198 	return error;
199 }
200 
201 /*
202  * Check for finished processes or timeouts.
203  */
204 struct cfjail *
205 next_proc(int nonblock)
206 {
207 	struct kevent ke;
208 	struct timespec ts;
209 	struct timespec *tsp;
210 	struct cfjail *j;
211 
212 	if (!TAILQ_EMPTY(&sleeping)) {
213 	again:
214 		tsp = NULL;
215 		if ((j = TAILQ_FIRST(&sleeping)) && j->timeout.tv_sec) {
216 			clock_gettime(CLOCK_REALTIME, &ts);
217 			ts.tv_sec = j->timeout.tv_sec - ts.tv_sec;
218 			ts.tv_nsec = j->timeout.tv_nsec - ts.tv_nsec;
219 			if (ts.tv_nsec < 0) {
220 				ts.tv_sec--;
221 				ts.tv_nsec += 1000000000;
222 			}
223 			if (ts.tv_sec < 0 ||
224 			    (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
225 				j->flags |= JF_TIMEOUT;
226 				clear_procs(j);
227 				return j;
228 			}
229 			tsp = &ts;
230 		}
231 		if (nonblock) {
232 			ts.tv_sec = 0;
233 			ts.tv_nsec = 0;
234 			tsp = &ts;
235 		}
236 		switch (kevent(kq, NULL, 0, &ke, 1, tsp)) {
237 		case -1:
238 			if (errno != EINTR)
239 				err(1, "kevent");
240 			goto again;
241 		case 0:
242 			if (!nonblock) {
243 				j = TAILQ_FIRST(&sleeping);
244 				j->flags |= JF_TIMEOUT;
245 				clear_procs(j);
246 				return j;
247 			}
248 			break;
249 		case 1:
250 			(void)waitpid(ke.ident, NULL, WNOHANG);
251 			if ((j = find_proc(ke.ident))) {
252 				j->pstatus = ke.data;
253 				return j;
254 			}
255 			goto again;
256 		}
257 	}
258 	return NULL;
259 }
260 
261 /*
262  * Run a single command for a jail, possible inside the jail.
263  */
264 static int
265 run_command(struct cfjail *j)
266 {
267 	const struct passwd *pwd;
268 	const struct cfstring *comstring, *s;
269 	login_cap_t *lcap;
270 	const char **argv;
271 	char *acs, *cs, *comcs, *devpath;
272 	const char *jidstr, *conslog, *path, *ruleset, *term, *username;
273 	enum intparam comparam;
274 	size_t comlen;
275 	pid_t pid;
276 	int argc, bg, clean, consfd, down, fib, i, injail, sjuser, timeout;
277 #if defined(INET) || defined(INET6)
278 	char *addr, *extrap, *p, *val;
279 #endif
280 
281 	static char *cleanenv;
282 
283 	/* Perform some operations that aren't actually commands */
284 	comparam = *j->comparam;
285 	down = j->flags & (JF_STOP | JF_FAILED);
286 	switch (comparam) {
287 	case IP_STOP_TIMEOUT:
288 		return term_procs(j);
289 
290 	case IP__OP:
291 		if (down) {
292 			if (jail_remove(j->jid) < 0 && errno == EPERM) {
293 				jail_warnx(j, "jail_remove: %s",
294 					   strerror(errno));
295 				return -1;
296 			}
297 			if (verbose > 0 || (verbose == 0 && (j->flags & JF_STOP
298 			    ? note_remove : j->name != NULL)))
299 			    jail_note(j, "removed\n");
300 			j->jid = -1;
301 			if (j->flags & JF_STOP)
302 				dep_done(j, DF_LIGHT);
303 			else
304 				j->flags &= ~JF_PERSIST;
305 		} else {
306 			if (create_jail(j) < 0)
307 				return -1;
308 			if (iflag)
309 				printf("%d\n", j->jid);
310 			if (verbose >= 0 && (j->name || verbose > 0))
311 				jail_note(j, "created\n");
312 			dep_done(j, DF_LIGHT);
313 		}
314 		return 0;
315 
316 	default: ;
317 	}
318 	/*
319 	 * Collect exec arguments.  Internal commands for network and
320 	 * mounting build their own argument lists.
321 	 */
322 	comstring = j->comstring;
323 	bg = 0;
324 	switch (comparam) {
325 #ifdef INET
326 	case IP__IP4_IFADDR:
327 		argc = 0;
328 		val = alloca(strlen(comstring->s) + 1);
329 		strcpy(val, comstring->s);
330 		cs = val;
331 		extrap = NULL;
332 		while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
333 			if (extrap == NULL) {
334 				*p = '\0';
335 				extrap = p + 1;
336 			}
337 			cs = p + 1;
338 			argc++;
339 		}
340 
341 		argv = alloca((8 + argc) * sizeof(char *));
342 		argv[0] = _PATH_IFCONFIG;
343 		if ((cs = strchr(val, '|'))) {
344 			argv[1] = acs = alloca(cs - val + 1);
345 			strlcpy(acs, val, cs - val + 1);
346 			addr = cs + 1;
347 		} else {
348 			argv[1] = string_param(j->intparams[IP_INTERFACE]);
349 			addr = val;
350 		}
351 		argv[2] = "inet";
352 		if (!(cs = strchr(addr, '/'))) {
353 			argv[3] = addr;
354 			argv[4] = "netmask";
355 			argv[5] = "255.255.255.255";
356 			argc = 6;
357 		} else if (strchr(cs + 1, '.')) {
358 			argv[3] = acs = alloca(cs - addr + 1);
359 			strlcpy(acs, addr, cs - addr + 1);
360 			argv[4] = "netmask";
361 			argv[5] = cs + 1;
362 			argc = 6;
363 		} else {
364 			argv[3] = addr;
365 			argc = 4;
366 		}
367 
368 		if (!down) {
369 			for (cs = strtok(extrap, " "); cs;
370 			     cs = strtok(NULL, " ")) {
371 				size_t len = strlen(cs) + 1;
372 				argv[argc++] = acs = alloca(len);
373 				strlcpy(acs, cs, len);
374 			}
375 		}
376 
377 		argv[argc] = down ? "-alias" : "alias";
378 		argv[argc + 1] = NULL;
379 		break;
380 #endif
381 
382 #ifdef INET6
383 	case IP__IP6_IFADDR:
384 		argc = 0;
385 		val = alloca(strlen(comstring->s) + 1);
386 		strcpy(val, comstring->s);
387 		cs = val;
388 		extrap = NULL;
389 		while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
390 			if (extrap == NULL) {
391 				*p = '\0';
392 				extrap = p + 1;
393 			}
394 			cs = p + 1;
395 			argc++;
396 		}
397 
398 		argv = alloca((8 + argc) * sizeof(char *));
399 		argv[0] = _PATH_IFCONFIG;
400 		if ((cs = strchr(val, '|'))) {
401 			argv[1] = acs = alloca(cs - val + 1);
402 			strlcpy(acs, val, cs - val + 1);
403 			addr = cs + 1;
404 		} else {
405 			argv[1] = string_param(j->intparams[IP_INTERFACE]);
406 			addr = val;
407 		}
408 		argv[2] = "inet6";
409 		argv[3] = addr;
410 		if (!(cs = strchr(addr, '/'))) {
411 			argv[4] = "prefixlen";
412 			argv[5] = "128";
413 			argc = 6;
414 		} else
415 			argc = 4;
416 
417 		if (!down) {
418 			for (cs = strtok(extrap, " "); cs;
419 			     cs = strtok(NULL, " ")) {
420 				size_t len = strlen(cs) + 1;
421 				argv[argc++] = acs = alloca(len);
422 				strlcpy(acs, cs, len);
423 			}
424 		}
425 
426 		argv[argc] = down ? "-alias" : "alias";
427 		argv[argc + 1] = NULL;
428 		break;
429 #endif
430 
431 	case IP_VNET_INTERFACE:
432 		argv = alloca(5 * sizeof(char *));
433 		argv[0] = _PATH_IFCONFIG;
434 		argv[1] = comstring->s;
435 		argv[2] = down ? "-vnet" : "vnet";
436 		jidstr = string_param(j->intparams[KP_JID]);
437 		argv[3] = jidstr ? jidstr : string_param(j->intparams[KP_NAME]);
438 		argv[4] = NULL;
439 		break;
440 
441 	case IP_MOUNT:
442 	case IP__MOUNT_FROM_FSTAB:
443 		argv = alloca(8 * sizeof(char *));
444 		comcs = alloca(comstring->len + 1);
445 		strcpy(comcs, comstring->s);
446 		argc = 0;
447 		for (cs = strtok(comcs, " \t\f\v\r\n"); cs && argc < 4;
448 		     cs = strtok(NULL, " \t\f\v\r\n")) {
449 			if (argc <= 1 && strunvis(cs, cs) < 0) {
450 				jail_warnx(j, "%s: %s: fstab parse error",
451 				    j->intparams[comparam]->name, comstring->s);
452 				return -1;
453 			}
454 			argv[argc++] = cs;
455 		}
456 		if (argc == 0)
457 			return 0;
458 		if (argc < 3) {
459 			jail_warnx(j, "%s: %s: missing information",
460 			    j->intparams[comparam]->name, comstring->s);
461 			return -1;
462 		}
463 		if (check_path(j, j->intparams[comparam]->name, argv[1], 0,
464 		    down ? argv[2] : NULL) < 0)
465 			return -1;
466 		if (down) {
467 			argv[4] = NULL;
468 			argv[3] = argv[1];
469 			argv[0] = "/sbin/umount";
470 		} else {
471 			if (argc == 4) {
472 				argv[7] = NULL;
473 				argv[6] = argv[1];
474 				argv[5] = argv[0];
475 				argv[4] = argv[3];
476 				argv[3] = "-o";
477 			} else {
478 				argv[5] = NULL;
479 				argv[4] = argv[1];
480 				argv[3] = argv[0];
481 			}
482 			argv[0] = _PATH_MOUNT;
483 		}
484 		argv[1] = "-t";
485 		break;
486 
487 	case IP_MOUNT_DEVFS:
488 		argv = alloca(7 * sizeof(char *));
489 		path = string_param(j->intparams[KP_PATH]);
490 		if (path == NULL) {
491 			jail_warnx(j, "mount.devfs: no path");
492 			return -1;
493 		}
494 		devpath = alloca(strlen(path) + 5);
495 		sprintf(devpath, "%s/dev", path);
496 		if (check_path(j, "mount.devfs", devpath, 0,
497 		    down ? "devfs" : NULL) < 0)
498 			return -1;
499 		if (down) {
500 			argv[0] = "/sbin/umount";
501 			argv[1] = devpath;
502 			argv[2] = NULL;
503 		} else {
504 			argv[0] = _PATH_MOUNT;
505 			argv[1] = "-t";
506 			argv[2] = "devfs";
507 			ruleset = string_param(j->intparams[KP_DEVFS_RULESET]);
508 			if (!ruleset)
509 			    ruleset = "4";	/* devfsrules_jail */
510 			argv[3] = acs = alloca(11 + strlen(ruleset));
511 			sprintf(acs, "-oruleset=%s", ruleset);
512 			argv[4] = ".";
513 			argv[5] = devpath;
514 			argv[6] = NULL;
515 		}
516 		break;
517 
518 	case IP_MOUNT_FDESCFS:
519 		argv = alloca(7 * sizeof(char *));
520 		path = string_param(j->intparams[KP_PATH]);
521 		if (path == NULL) {
522 			jail_warnx(j, "mount.fdescfs: no path");
523 			return -1;
524 		}
525 		devpath = alloca(strlen(path) + 8);
526 		sprintf(devpath, "%s/dev/fd", path);
527 		if (check_path(j, "mount.fdescfs", devpath, 0,
528 		    down ? "fdescfs" : NULL) < 0)
529 			return -1;
530 		if (down) {
531 			argv[0] = "/sbin/umount";
532 			argv[1] = devpath;
533 			argv[2] = NULL;
534 		} else {
535 			argv[0] = _PATH_MOUNT;
536 			argv[1] = "-t";
537 			argv[2] = "fdescfs";
538 			argv[3] = ".";
539 			argv[4] = devpath;
540 			argv[5] = NULL;
541 		}
542 		break;
543 
544 	case IP_MOUNT_PROCFS:
545 		argv = alloca(7 * sizeof(char *));
546 		path = string_param(j->intparams[KP_PATH]);
547 		if (path == NULL) {
548 			jail_warnx(j, "mount.procfs: no path");
549 			return -1;
550 		}
551 		devpath = alloca(strlen(path) + 6);
552 		sprintf(devpath, "%s/proc", path);
553 		if (check_path(j, "mount.procfs", devpath, 0,
554 		    down ? "procfs" : NULL) < 0)
555 			return -1;
556 		if (down) {
557 			argv[0] = "/sbin/umount";
558 			argv[1] = devpath;
559 			argv[2] = NULL;
560 		} else {
561 			argv[0] = _PATH_MOUNT;
562 			argv[1] = "-t";
563 			argv[2] = "procfs";
564 			argv[3] = ".";
565 			argv[4] = devpath;
566 			argv[5] = NULL;
567 		}
568 		break;
569 
570 	case IP_COMMAND:
571 		if (j->name != NULL)
572 			goto default_command;
573 		argc = 0;
574 		TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
575 			argc++;
576 		argv = alloca((argc + 1) * sizeof(char *));
577 		argc = 0;
578 		TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
579 			argv[argc++] = s->s;
580 		argv[argc] = NULL;
581 		j->comstring = &dummystring;
582 		break;
583 
584 	default:
585 	default_command:
586 		if ((cs = strpbrk(comstring->s, "!\"$&'()*;<>?[\\]`{|}~")) &&
587 		    !(cs[0] == '&' && cs[1] == '\0')) {
588 			argv = alloca(4 * sizeof(char *));
589 			argv[0] = _PATH_BSHELL;
590 			argv[1] = "-c";
591 			argv[2] = comstring->s;
592 			argv[3] = NULL;
593 		} else {
594 			if (cs) {
595 				*cs = 0;
596 				bg = 1;
597 			}
598 			comcs = alloca(comstring->len + 1);
599 			strcpy(comcs, comstring->s);
600 			argc = 0;
601 			for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
602 			     cs = strtok(NULL, " \t\f\v\r\n"))
603 				argc++;
604 			argv = alloca((argc + 1) * sizeof(char *));
605 			strcpy(comcs, comstring->s);
606 			argc = 0;
607 			for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
608 			     cs = strtok(NULL, " \t\f\v\r\n"))
609 				argv[argc++] = cs;
610 			argv[argc] = NULL;
611 		}
612 	}
613 	if (argv[0] == NULL)
614 		return 0;
615 
616 	if (int_param(j->intparams[IP_EXEC_TIMEOUT], &timeout) &&
617 	    timeout != 0) {
618 		clock_gettime(CLOCK_REALTIME, &j->timeout);
619 		j->timeout.tv_sec += timeout;
620 	} else
621 		j->timeout.tv_sec = 0;
622 
623 	injail = comparam == IP_EXEC_START || comparam == IP_COMMAND ||
624 	    comparam == IP_EXEC_STOP;
625 	clean = bool_param(j->intparams[IP_EXEC_CLEAN]);
626 	username = string_param(j->intparams[injail
627 	    ? IP_EXEC_JAIL_USER : IP_EXEC_SYSTEM_USER]);
628 	sjuser = bool_param(j->intparams[IP_EXEC_SYSTEM_JAIL_USER]);
629 
630 	consfd = 0;
631 	if (injail &&
632 	    (conslog = string_param(j->intparams[IP_EXEC_CONSOLELOG]))) {
633 		if (check_path(j, "exec.consolelog", conslog, 1, NULL) < 0)
634 			return -1;
635 		consfd =
636 		    open(conslog, O_WRONLY | O_CREAT | O_APPEND, DEFFILEMODE);
637 		if (consfd < 0) {
638 			jail_warnx(j, "open %s: %s", conslog, strerror(errno));
639 			return -1;
640 		}
641 	}
642 
643 	comlen = 0;
644 	for (i = 0; argv[i]; i++)
645 		comlen += strlen(argv[i]) + 1;
646 	j->comline = cs = emalloc(comlen);
647 	for (i = 0; argv[i]; i++) {
648 		strcpy(cs, argv[i]);
649 		if (argv[i + 1]) {
650 			cs += strlen(argv[i]) + 1;
651 			cs[-1] = ' ';
652 		}
653 	}
654 	if (verbose > 0)
655 		jail_note(j, "run command%s%s%s: %s\n",
656 		    injail ? " in jail" : "", username ? " as " : "",
657 		    username ? username : "", j->comline);
658 
659 	pid = fork();
660 	if (pid < 0)
661 		err(1, "fork");
662 	if (pid > 0) {
663 		if (bg || !add_proc(j, pid)) {
664 			free(j->comline);
665 			j->comline = NULL;
666 			return 0;
667 		} else {
668 			paralimit--;
669 			return 1;
670 		}
671 	}
672 	if (bg)
673 		setsid();
674 
675 	/* Set up the environment and run the command */
676 	pwd = NULL;
677 	lcap = NULL;
678 	if ((clean || username) && injail && sjuser &&
679 	    get_user_info(j, username, &pwd, &lcap) < 0)
680 		exit(1);
681 	if (injail) {
682 		/* jail_attach won't chdir along with its chroot. */
683 		path = string_param(j->intparams[KP_PATH]);
684 		if (path && chdir(path) < 0) {
685 			jail_warnx(j, "chdir %s: %s", path, strerror(errno));
686 			exit(1);
687 		}
688 		if (int_param(j->intparams[IP_EXEC_FIB], &fib) &&
689 		    setfib(fib) < 0) {
690 			jail_warnx(j, "setfib: %s", strerror(errno));
691 			exit(1);
692 		}
693 		if (jail_attach(j->jid) < 0) {
694 			jail_warnx(j, "jail_attach: %s", strerror(errno));
695 			exit(1);
696 		}
697 	}
698 	if (clean || username) {
699 		if (!(injail && sjuser) &&
700 		    get_user_info(j, username, &pwd, &lcap) < 0)
701 			exit(1);
702 		if (clean) {
703 			term = getenv("TERM");
704 			environ = &cleanenv;
705 			setenv("PATH", "/bin:/usr/bin", 0);
706 			if (term != NULL)
707 				setenv("TERM", term, 1);
708 		}
709 		if (setgid(pwd->pw_gid) < 0) {
710 			jail_warnx(j, "setgid %d: %s", pwd->pw_gid,
711 			    strerror(errno));
712 			exit(1);
713 		}
714 		if (setusercontext(lcap, pwd, pwd->pw_uid, username
715 		    ? LOGIN_SETALL & ~LOGIN_SETGROUP & ~LOGIN_SETLOGIN
716 		    : LOGIN_SETPATH | LOGIN_SETENV) < 0) {
717 			jail_warnx(j, "setusercontext %s: %s", pwd->pw_name,
718 			    strerror(errno));
719 			exit(1);
720 		}
721 		login_close(lcap);
722 		setenv("USER", pwd->pw_name, 1);
723 		setenv("HOME", pwd->pw_dir, 1);
724 		setenv("SHELL",
725 		    *pwd->pw_shell ? pwd->pw_shell : _PATH_BSHELL, 1);
726 		if (clean && chdir(pwd->pw_dir) < 0) {
727 			jail_warnx(j, "chdir %s: %s",
728 			    pwd->pw_dir, strerror(errno));
729 			exit(1);
730 		}
731 		endpwent();
732 	}
733 
734 	if (consfd != 0 && (dup2(consfd, 1) < 0 || dup2(consfd, 2) < 0)) {
735 		jail_warnx(j, "exec.consolelog: %s", strerror(errno));
736 		exit(1);
737 	}
738 	closefrom(3);
739 	execvp(argv[0], __DECONST(char *const*, argv));
740 	jail_warnx(j, "exec %s: %s", argv[0], strerror(errno));
741 	exit(1);
742 }
743 
744 /*
745  * Add a process to the hash, tied to a jail.
746  */
747 static int
748 add_proc(struct cfjail *j, pid_t pid)
749 {
750 	struct kevent ke;
751 	struct cfjail *tj;
752 	struct phash *ph;
753 
754 	if (!kq && (kq = kqueue()) < 0)
755 		err(1, "kqueue");
756 	EV_SET(&ke, pid, EVFILT_PROC, EV_ADD, NOTE_EXIT, 0, NULL);
757 	if (kevent(kq, &ke, 1, NULL, 0, NULL) < 0) {
758 		if (errno == ESRCH)
759 			return 0;
760 		err(1, "kevent");
761 	}
762 	ph = emalloc(sizeof(struct phash));
763 	ph->j = j;
764 	ph->pid = pid;
765 	LIST_INSERT_HEAD(&phash[pid % PHASH_SIZE], ph, le);
766 	j->nprocs++;
767 	j->flags |= JF_SLEEPQ;
768 	if (j->timeout.tv_sec == 0)
769 		requeue(j, &sleeping);
770 	else {
771 		/* File the jail in the sleep queue according to its timeout. */
772 		TAILQ_REMOVE(j->queue, j, tq);
773 		TAILQ_FOREACH(tj, &sleeping, tq) {
774 			if (!tj->timeout.tv_sec ||
775 			    j->timeout.tv_sec < tj->timeout.tv_sec ||
776 			    (j->timeout.tv_sec == tj->timeout.tv_sec &&
777 			    j->timeout.tv_nsec <= tj->timeout.tv_nsec)) {
778 				TAILQ_INSERT_BEFORE(tj, j, tq);
779 				break;
780 			}
781 		}
782 		if (tj == NULL)
783 			TAILQ_INSERT_TAIL(&sleeping, j, tq);
784 		j->queue = &sleeping;
785 	}
786 	return 1;
787 }
788 
789 /*
790  * Remove any processes from the hash that correspond to a jail.
791  */
792 static void
793 clear_procs(struct cfjail *j)
794 {
795 	struct kevent ke;
796 	struct phash *ph, *tph;
797 	int i;
798 
799 	j->nprocs = 0;
800 	for (i = 0; i < PHASH_SIZE; i++)
801 		LIST_FOREACH_SAFE(ph, &phash[i], le, tph)
802 			if (ph->j == j) {
803 				EV_SET(&ke, ph->pid, EVFILT_PROC, EV_DELETE,
804 				    NOTE_EXIT, 0, NULL);
805 				(void)kevent(kq, &ke, 1, NULL, 0, NULL);
806 				LIST_REMOVE(ph, le);
807 				free(ph);
808 			}
809 }
810 
811 /*
812  * Find the jail that corresponds to an exited process.
813  */
814 static struct cfjail *
815 find_proc(pid_t pid)
816 {
817 	struct cfjail *j;
818 	struct phash *ph;
819 
820 	LIST_FOREACH(ph, &phash[pid % PHASH_SIZE], le)
821 		if (ph->pid == pid) {
822 			j = ph->j;
823 			LIST_REMOVE(ph, le);
824 			free(ph);
825 			return --j->nprocs ? NULL : j;
826 		}
827 	return NULL;
828 }
829 
830 /*
831  * Send SIGTERM to all processes in a jail and wait for them to die.
832  */
833 static int
834 term_procs(struct cfjail *j)
835 {
836 	struct kinfo_proc *ki;
837 	int i, noted, pcnt, timeout;
838 
839 	static kvm_t *kd;
840 
841 	if (!int_param(j->intparams[IP_STOP_TIMEOUT], &timeout))
842 		timeout = DEFAULT_STOP_TIMEOUT;
843 	else if (timeout == 0)
844 		return 0;
845 
846 	if (kd == NULL) {
847 		kd = kvm_open(NULL, NULL, NULL, O_RDONLY, NULL);
848 		if (kd == NULL)
849 			return 0;
850 	}
851 
852 	ki = kvm_getprocs(kd, KERN_PROC_PROC, 0, &pcnt);
853 	if (ki == NULL)
854 		return 0;
855 	noted = 0;
856 	for (i = 0; i < pcnt; i++)
857 		if (ki[i].ki_jid == j->jid &&
858 		    kill(ki[i].ki_pid, SIGTERM) == 0) {
859 			(void)add_proc(j, ki[i].ki_pid);
860 			if (verbose > 0) {
861 				if (!noted) {
862 					noted = 1;
863 					jail_note(j, "sent SIGTERM to:");
864 				}
865 				printf(" %d", ki[i].ki_pid);
866 			}
867 		}
868 	if (noted)
869 		printf("\n");
870 	if (j->nprocs > 0) {
871 		clock_gettime(CLOCK_REALTIME, &j->timeout);
872 		j->timeout.tv_sec += timeout;
873 		return 1;
874 	}
875 	return 0;
876 }
877 
878 /*
879  * Look up a user in the passwd and login.conf files.
880  */
881 static int
882 get_user_info(struct cfjail *j, const char *username,
883     const struct passwd **pwdp, login_cap_t **lcapp)
884 {
885 	const struct passwd *pwd;
886 
887 	errno = 0;
888 	*pwdp = pwd = username ? getpwnam(username) : getpwuid(getuid());
889 	if (pwd == NULL) {
890 		if (errno)
891 			jail_warnx(j, "getpwnam%s%s: %s", username ? " " : "",
892 			    username ? username : "", strerror(errno));
893 		else if (username)
894 			jail_warnx(j, "%s: no such user", username);
895 		else
896 			jail_warnx(j, "unknown uid %d", getuid());
897 		return -1;
898 	}
899 	*lcapp = login_getpwclass(pwd);
900 	if (*lcapp == NULL) {
901 		jail_warnx(j, "getpwclass %s: %s", pwd->pw_name,
902 		    strerror(errno));
903 		return -1;
904 	}
905 	/* Set the groups while the group file is still available */
906 	if (initgroups(pwd->pw_name, pwd->pw_gid) < 0) {
907 		jail_warnx(j, "initgroups %s: %s", pwd->pw_name,
908 		    strerror(errno));
909 		return -1;
910 	}
911 	return 0;
912 }
913 
914 /*
915  * Make sure a mount or consolelog path is a valid absolute pathname
916  * with no symlinks.
917  */
918 static int
919 check_path(struct cfjail *j, const char *pname, const char *path, int isfile,
920     const char *umount_type)
921 {
922 	struct stat st, mpst;
923 	struct statfs stfs;
924 	char *tpath, *p;
925 	const char *jailpath;
926 	size_t jplen;
927 
928 	if (path[0] != '/') {
929 		jail_warnx(j, "%s: %s: not an absolute pathname",
930 		    pname, path);
931 		return -1;
932 	}
933 	/*
934 	 * Only check for symlinks in components below the jail's path,
935 	 * since that's where the security risk lies.
936 	 */
937 	jailpath = string_param(j->intparams[KP_PATH]);
938 	if (jailpath == NULL)
939 		jailpath = "";
940 	jplen = strlen(jailpath);
941 	if (!strncmp(path, jailpath, jplen) && path[jplen] == '/') {
942 		tpath = alloca(strlen(path) + 1);
943 		strcpy(tpath, path);
944 		for (p = tpath + jplen; p != NULL; ) {
945 			p = strchr(p + 1, '/');
946 			if (p)
947 				*p = '\0';
948 			if (lstat(tpath, &st) < 0) {
949 				if (errno == ENOENT && isfile && !p)
950 					break;
951 				jail_warnx(j, "%s: %s: %s", pname, tpath,
952 				    strerror(errno));
953 				return -1;
954 			}
955 			if (S_ISLNK(st.st_mode)) {
956 				jail_warnx(j, "%s: %s is a symbolic link",
957 				    pname, tpath);
958 				return -1;
959 			}
960 			if (p)
961 				*p = '/';
962 		}
963 	}
964 	if (umount_type != NULL) {
965 		if (stat(path, &st) < 0 || statfs(path, &stfs) < 0) {
966 			jail_warnx(j, "%s: %s: %s", pname, path,
967 			    strerror(errno));
968 			return -1;
969 		}
970 		if (stat(stfs.f_mntonname, &mpst) < 0) {
971 			jail_warnx(j, "%s: %s: %s", pname, stfs.f_mntonname,
972 			    strerror(errno));
973 			return -1;
974 		}
975 		if (st.st_ino != mpst.st_ino) {
976 			jail_warnx(j, "%s: %s: not a mount point",
977 			    pname, path);
978 			return -1;
979 		}
980 		if (strcmp(stfs.f_fstypename, umount_type)) {
981 			jail_warnx(j, "%s: %s: not a %s mount",
982 			    pname, path, umount_type);
983 			return -1;
984 		}
985 	}
986 	return 0;
987 }
988