xref: /freebsd/usr.sbin/jail/command.c (revision 273c26a3c3bea87a241d6879abd4f991db180bf0)
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 		if (j->flags & JF_FROM_RUNQ)
96 			requeue_head(j, &runnable);
97 		else
98 			requeue(j, &runnable);
99 		return 1;
100 	}
101 	j->flags &= ~JF_FROM_RUNQ;
102 	create_failed = (j->flags & (JF_STOP | JF_FAILED)) == JF_FAILED;
103 	stopping = (j->flags & JF_STOP) != 0;
104 	comparam = *j->comparam;
105 	for (;;) {
106 		if (j->comstring == NULL) {
107 			j->comparam += create_failed ? -1 : 1;
108 			switch ((comparam = *j->comparam)) {
109 			case IP__NULL:
110 				return 0;
111 			case IP_MOUNT_DEVFS:
112 				if (!bool_param(j->intparams[IP_MOUNT_DEVFS]))
113 					continue;
114 				j->comstring = &dummystring;
115 				break;
116 			case IP_MOUNT_FDESCFS:
117 				if (!bool_param(j->intparams[IP_MOUNT_FDESCFS]))
118 					continue;
119 				j->comstring = &dummystring;
120 				break;
121 			case IP_MOUNT_PROCFS:
122 				if (!bool_param(j->intparams[IP_MOUNT_PROCFS]))
123 					continue;
124 				j->comstring = &dummystring;
125 				break;
126 			case IP__OP:
127 			case IP_STOP_TIMEOUT:
128 				j->comstring = &dummystring;
129 				break;
130 			default:
131 				if (j->intparams[comparam] == NULL)
132 					continue;
133 				j->comstring = create_failed || (stopping &&
134 				    (j->intparams[comparam]->flags & PF_REV))
135 				    ? TAILQ_LAST(&j->intparams[comparam]->val,
136 					cfstrings)
137 				    : TAILQ_FIRST(&j->intparams[comparam]->val);
138 			}
139 		} else {
140 			j->comstring = j->comstring == &dummystring ? NULL :
141 			    create_failed || (stopping &&
142 			    (j->intparams[comparam]->flags & PF_REV))
143 			    ? TAILQ_PREV(j->comstring, cfstrings, tq)
144 			    : TAILQ_NEXT(j->comstring, tq);
145 		}
146 		if (j->comstring == NULL || j->comstring->len == 0 ||
147 		    (create_failed && (comparam == IP_EXEC_PRESTART ||
148 		    comparam == IP_EXEC_START || comparam == IP_COMMAND ||
149 		    comparam == IP_EXEC_POSTSTART)))
150 			continue;
151 		switch (run_command(j)) {
152 		case -1:
153 			failed(j);
154 			/* FALLTHROUGH */
155 		case 1:
156 			return 1;
157 		}
158 	}
159 }
160 
161 /*
162  * Check command exit status
163  */
164 int
165 finish_command(struct cfjail *j)
166 {
167 	struct cfjail *rj;
168 	int error;
169 
170 	if (!(j->flags & JF_SLEEPQ))
171 		return 0;
172 	j->flags &= ~JF_SLEEPQ;
173 	if (*j->comparam == IP_STOP_TIMEOUT) {
174 		j->flags &= ~JF_TIMEOUT;
175 		j->pstatus = 0;
176 		return 0;
177 	}
178 	paralimit++;
179 	if (!TAILQ_EMPTY(&runnable)) {
180 		rj = TAILQ_FIRST(&runnable);
181 		rj->flags |= JF_FROM_RUNQ;
182 		requeue(rj, &ready);
183 	}
184 	error = 0;
185 	if (j->flags & JF_TIMEOUT) {
186 		j->flags &= ~JF_TIMEOUT;
187 		if (*j->comparam != IP_STOP_TIMEOUT) {
188 			jail_warnx(j, "%s: timed out", j->comline);
189 			failed(j);
190 			error = -1;
191 		} else if (verbose > 0)
192 			jail_note(j, "timed out\n");
193 	} else if (j->pstatus != 0) {
194 		if (WIFSIGNALED(j->pstatus))
195 			jail_warnx(j, "%s: exited on signal %d",
196 			    j->comline, WTERMSIG(j->pstatus));
197 		else
198 			jail_warnx(j, "%s: failed", j->comline);
199 		j->pstatus = 0;
200 		failed(j);
201 		error = -1;
202 	}
203 	free(j->comline);
204 	j->comline = NULL;
205 	return error;
206 }
207 
208 /*
209  * Check for finished processes or timeouts.
210  */
211 struct cfjail *
212 next_proc(int nonblock)
213 {
214 	struct kevent ke;
215 	struct timespec ts;
216 	struct timespec *tsp;
217 	struct cfjail *j;
218 
219 	if (!TAILQ_EMPTY(&sleeping)) {
220 	again:
221 		tsp = NULL;
222 		if ((j = TAILQ_FIRST(&sleeping)) && j->timeout.tv_sec) {
223 			clock_gettime(CLOCK_REALTIME, &ts);
224 			ts.tv_sec = j->timeout.tv_sec - ts.tv_sec;
225 			ts.tv_nsec = j->timeout.tv_nsec - ts.tv_nsec;
226 			if (ts.tv_nsec < 0) {
227 				ts.tv_sec--;
228 				ts.tv_nsec += 1000000000;
229 			}
230 			if (ts.tv_sec < 0 ||
231 			    (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
232 				j->flags |= JF_TIMEOUT;
233 				clear_procs(j);
234 				return j;
235 			}
236 			tsp = &ts;
237 		}
238 		if (nonblock) {
239 			ts.tv_sec = 0;
240 			ts.tv_nsec = 0;
241 			tsp = &ts;
242 		}
243 		switch (kevent(kq, NULL, 0, &ke, 1, tsp)) {
244 		case -1:
245 			if (errno != EINTR)
246 				err(1, "kevent");
247 			goto again;
248 		case 0:
249 			if (!nonblock) {
250 				j = TAILQ_FIRST(&sleeping);
251 				j->flags |= JF_TIMEOUT;
252 				clear_procs(j);
253 				return j;
254 			}
255 			break;
256 		case 1:
257 			(void)waitpid(ke.ident, NULL, WNOHANG);
258 			if ((j = find_proc(ke.ident))) {
259 				j->pstatus = ke.data;
260 				return j;
261 			}
262 			goto again;
263 		}
264 	}
265 	return NULL;
266 }
267 
268 /*
269  * Run a single command for a jail, possibly inside the jail.
270  */
271 static int
272 run_command(struct cfjail *j)
273 {
274 	const struct passwd *pwd;
275 	const struct cfstring *comstring, *s;
276 	login_cap_t *lcap;
277 	const char **argv;
278 	char *acs, *cs, *comcs, *devpath;
279 	const char *jidstr, *conslog, *path, *ruleset, *term, *username;
280 	enum intparam comparam;
281 	size_t comlen;
282 	pid_t pid;
283 	int argc, bg, clean, consfd, down, fib, i, injail, sjuser, timeout;
284 #if defined(INET) || defined(INET6)
285 	char *addr, *extrap, *p, *val;
286 #endif
287 
288 	static char *cleanenv;
289 
290 	/* Perform some operations that aren't actually commands */
291 	comparam = *j->comparam;
292 	down = j->flags & (JF_STOP | JF_FAILED);
293 	switch (comparam) {
294 	case IP_STOP_TIMEOUT:
295 		return term_procs(j);
296 
297 	case IP__OP:
298 		if (down) {
299 			if (jail_remove(j->jid) < 0 && errno == EPERM) {
300 				jail_warnx(j, "jail_remove: %s",
301 					   strerror(errno));
302 				return -1;
303 			}
304 			if (verbose > 0 || (verbose == 0 && (j->flags & JF_STOP
305 			    ? note_remove : j->name != NULL)))
306 			    jail_note(j, "removed\n");
307 			j->jid = -1;
308 			if (j->flags & JF_STOP)
309 				dep_done(j, DF_LIGHT);
310 			else
311 				j->flags &= ~JF_PERSIST;
312 		} else {
313 			if (create_jail(j) < 0)
314 				return -1;
315 			if (iflag)
316 				printf("%d\n", j->jid);
317 			if (verbose >= 0 && (j->name || verbose > 0))
318 				jail_note(j, "created\n");
319 			dep_done(j, DF_LIGHT);
320 		}
321 		return 0;
322 
323 	default: ;
324 	}
325 	/*
326 	 * Collect exec arguments.  Internal commands for network and
327 	 * mounting build their own argument lists.
328 	 */
329 	comstring = j->comstring;
330 	bg = 0;
331 	switch (comparam) {
332 #ifdef INET
333 	case IP__IP4_IFADDR:
334 		argc = 0;
335 		val = alloca(strlen(comstring->s) + 1);
336 		strcpy(val, comstring->s);
337 		cs = val;
338 		extrap = NULL;
339 		while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
340 			if (extrap == NULL) {
341 				*p = '\0';
342 				extrap = p + 1;
343 			}
344 			cs = p + 1;
345 			argc++;
346 		}
347 
348 		argv = alloca((8 + argc) * sizeof(char *));
349 		argv[0] = _PATH_IFCONFIG;
350 		if ((cs = strchr(val, '|'))) {
351 			argv[1] = acs = alloca(cs - val + 1);
352 			strlcpy(acs, val, cs - val + 1);
353 			addr = cs + 1;
354 		} else {
355 			argv[1] = string_param(j->intparams[IP_INTERFACE]);
356 			addr = val;
357 		}
358 		argv[2] = "inet";
359 		if (!(cs = strchr(addr, '/'))) {
360 			argv[3] = addr;
361 			argv[4] = "netmask";
362 			argv[5] = "255.255.255.255";
363 			argc = 6;
364 		} else if (strchr(cs + 1, '.')) {
365 			argv[3] = acs = alloca(cs - addr + 1);
366 			strlcpy(acs, addr, cs - addr + 1);
367 			argv[4] = "netmask";
368 			argv[5] = cs + 1;
369 			argc = 6;
370 		} else {
371 			argv[3] = addr;
372 			argc = 4;
373 		}
374 
375 		if (!down) {
376 			for (cs = strtok(extrap, " "); cs;
377 			     cs = strtok(NULL, " ")) {
378 				size_t len = strlen(cs) + 1;
379 				argv[argc++] = acs = alloca(len);
380 				strlcpy(acs, cs, len);
381 			}
382 		}
383 
384 		argv[argc] = down ? "-alias" : "alias";
385 		argv[argc + 1] = NULL;
386 		break;
387 #endif
388 
389 #ifdef INET6
390 	case IP__IP6_IFADDR:
391 		argc = 0;
392 		val = alloca(strlen(comstring->s) + 1);
393 		strcpy(val, comstring->s);
394 		cs = val;
395 		extrap = NULL;
396 		while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
397 			if (extrap == NULL) {
398 				*p = '\0';
399 				extrap = p + 1;
400 			}
401 			cs = p + 1;
402 			argc++;
403 		}
404 
405 		argv = alloca((8 + argc) * sizeof(char *));
406 		argv[0] = _PATH_IFCONFIG;
407 		if ((cs = strchr(val, '|'))) {
408 			argv[1] = acs = alloca(cs - val + 1);
409 			strlcpy(acs, val, cs - val + 1);
410 			addr = cs + 1;
411 		} else {
412 			argv[1] = string_param(j->intparams[IP_INTERFACE]);
413 			addr = val;
414 		}
415 		argv[2] = "inet6";
416 		argv[3] = addr;
417 		if (!(cs = strchr(addr, '/'))) {
418 			argv[4] = "prefixlen";
419 			argv[5] = "128";
420 			argc = 6;
421 		} else
422 			argc = 4;
423 
424 		if (!down) {
425 			for (cs = strtok(extrap, " "); cs;
426 			     cs = strtok(NULL, " ")) {
427 				size_t len = strlen(cs) + 1;
428 				argv[argc++] = acs = alloca(len);
429 				strlcpy(acs, cs, len);
430 			}
431 		}
432 
433 		argv[argc] = down ? "-alias" : "alias";
434 		argv[argc + 1] = NULL;
435 		break;
436 #endif
437 
438 	case IP_VNET_INTERFACE:
439 		argv = alloca(5 * sizeof(char *));
440 		argv[0] = _PATH_IFCONFIG;
441 		argv[1] = comstring->s;
442 		argv[2] = down ? "-vnet" : "vnet";
443 		jidstr = string_param(j->intparams[KP_JID]);
444 		argv[3] = jidstr ? jidstr : string_param(j->intparams[KP_NAME]);
445 		argv[4] = NULL;
446 		break;
447 
448 	case IP_MOUNT:
449 	case IP__MOUNT_FROM_FSTAB:
450 		argv = alloca(8 * sizeof(char *));
451 		comcs = alloca(comstring->len + 1);
452 		strcpy(comcs, comstring->s);
453 		argc = 0;
454 		for (cs = strtok(comcs, " \t\f\v\r\n"); cs && argc < 4;
455 		     cs = strtok(NULL, " \t\f\v\r\n")) {
456 			if (argc <= 1 && strunvis(cs, cs) < 0) {
457 				jail_warnx(j, "%s: %s: fstab parse error",
458 				    j->intparams[comparam]->name, comstring->s);
459 				return -1;
460 			}
461 			argv[argc++] = cs;
462 		}
463 		if (argc == 0)
464 			return 0;
465 		if (argc < 3) {
466 			jail_warnx(j, "%s: %s: missing information",
467 			    j->intparams[comparam]->name, comstring->s);
468 			return -1;
469 		}
470 		if (check_path(j, j->intparams[comparam]->name, argv[1], 0,
471 		    down ? argv[2] : NULL) < 0)
472 			return -1;
473 		if (down) {
474 			argv[4] = NULL;
475 			argv[3] = argv[1];
476 			argv[0] = "/sbin/umount";
477 		} else {
478 			if (argc == 4) {
479 				argv[7] = NULL;
480 				argv[6] = argv[1];
481 				argv[5] = argv[0];
482 				argv[4] = argv[3];
483 				argv[3] = "-o";
484 			} else {
485 				argv[5] = NULL;
486 				argv[4] = argv[1];
487 				argv[3] = argv[0];
488 			}
489 			argv[0] = _PATH_MOUNT;
490 		}
491 		argv[1] = "-t";
492 		break;
493 
494 	case IP_MOUNT_DEVFS:
495 		argv = alloca(7 * sizeof(char *));
496 		path = string_param(j->intparams[KP_PATH]);
497 		if (path == NULL) {
498 			jail_warnx(j, "mount.devfs: no path");
499 			return -1;
500 		}
501 		devpath = alloca(strlen(path) + 5);
502 		sprintf(devpath, "%s/dev", path);
503 		if (check_path(j, "mount.devfs", devpath, 0,
504 		    down ? "devfs" : NULL) < 0)
505 			return -1;
506 		if (down) {
507 			argv[0] = "/sbin/umount";
508 			argv[1] = devpath;
509 			argv[2] = NULL;
510 		} else {
511 			argv[0] = _PATH_MOUNT;
512 			argv[1] = "-t";
513 			argv[2] = "devfs";
514 			ruleset = string_param(j->intparams[KP_DEVFS_RULESET]);
515 			if (!ruleset)
516 			    ruleset = "4";	/* devfsrules_jail */
517 			argv[3] = acs = alloca(11 + strlen(ruleset));
518 			sprintf(acs, "-oruleset=%s", ruleset);
519 			argv[4] = ".";
520 			argv[5] = devpath;
521 			argv[6] = NULL;
522 		}
523 		break;
524 
525 	case IP_MOUNT_FDESCFS:
526 		argv = alloca(7 * sizeof(char *));
527 		path = string_param(j->intparams[KP_PATH]);
528 		if (path == NULL) {
529 			jail_warnx(j, "mount.fdescfs: no path");
530 			return -1;
531 		}
532 		devpath = alloca(strlen(path) + 8);
533 		sprintf(devpath, "%s/dev/fd", path);
534 		if (check_path(j, "mount.fdescfs", devpath, 0,
535 		    down ? "fdescfs" : NULL) < 0)
536 			return -1;
537 		if (down) {
538 			argv[0] = "/sbin/umount";
539 			argv[1] = devpath;
540 			argv[2] = NULL;
541 		} else {
542 			argv[0] = _PATH_MOUNT;
543 			argv[1] = "-t";
544 			argv[2] = "fdescfs";
545 			argv[3] = ".";
546 			argv[4] = devpath;
547 			argv[5] = NULL;
548 		}
549 		break;
550 
551 	case IP_MOUNT_PROCFS:
552 		argv = alloca(7 * sizeof(char *));
553 		path = string_param(j->intparams[KP_PATH]);
554 		if (path == NULL) {
555 			jail_warnx(j, "mount.procfs: no path");
556 			return -1;
557 		}
558 		devpath = alloca(strlen(path) + 6);
559 		sprintf(devpath, "%s/proc", path);
560 		if (check_path(j, "mount.procfs", devpath, 0,
561 		    down ? "procfs" : NULL) < 0)
562 			return -1;
563 		if (down) {
564 			argv[0] = "/sbin/umount";
565 			argv[1] = devpath;
566 			argv[2] = NULL;
567 		} else {
568 			argv[0] = _PATH_MOUNT;
569 			argv[1] = "-t";
570 			argv[2] = "procfs";
571 			argv[3] = ".";
572 			argv[4] = devpath;
573 			argv[5] = NULL;
574 		}
575 		break;
576 
577 	case IP_COMMAND:
578 		if (j->name != NULL)
579 			goto default_command;
580 		argc = 0;
581 		TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
582 			argc++;
583 		argv = alloca((argc + 1) * sizeof(char *));
584 		argc = 0;
585 		TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
586 			argv[argc++] = s->s;
587 		argv[argc] = NULL;
588 		j->comstring = &dummystring;
589 		break;
590 
591 	default:
592 	default_command:
593 		if ((cs = strpbrk(comstring->s, "!\"$&'()*;<>?[\\]`{|}~")) &&
594 		    !(cs[0] == '&' && cs[1] == '\0')) {
595 			argv = alloca(4 * sizeof(char *));
596 			argv[0] = _PATH_BSHELL;
597 			argv[1] = "-c";
598 			argv[2] = comstring->s;
599 			argv[3] = NULL;
600 		} else {
601 			if (cs) {
602 				*cs = 0;
603 				bg = 1;
604 			}
605 			comcs = alloca(comstring->len + 1);
606 			strcpy(comcs, comstring->s);
607 			argc = 0;
608 			for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
609 			     cs = strtok(NULL, " \t\f\v\r\n"))
610 				argc++;
611 			argv = alloca((argc + 1) * sizeof(char *));
612 			strcpy(comcs, comstring->s);
613 			argc = 0;
614 			for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
615 			     cs = strtok(NULL, " \t\f\v\r\n"))
616 				argv[argc++] = cs;
617 			argv[argc] = NULL;
618 		}
619 	}
620 	if (argv[0] == NULL)
621 		return 0;
622 
623 	if (int_param(j->intparams[IP_EXEC_TIMEOUT], &timeout) &&
624 	    timeout != 0) {
625 		clock_gettime(CLOCK_REALTIME, &j->timeout);
626 		j->timeout.tv_sec += timeout;
627 	} else
628 		j->timeout.tv_sec = 0;
629 
630 	injail = comparam == IP_EXEC_START || comparam == IP_COMMAND ||
631 	    comparam == IP_EXEC_STOP;
632 	clean = bool_param(j->intparams[IP_EXEC_CLEAN]);
633 	username = string_param(j->intparams[injail
634 	    ? IP_EXEC_JAIL_USER : IP_EXEC_SYSTEM_USER]);
635 	sjuser = bool_param(j->intparams[IP_EXEC_SYSTEM_JAIL_USER]);
636 
637 	consfd = 0;
638 	if (injail &&
639 	    (conslog = string_param(j->intparams[IP_EXEC_CONSOLELOG]))) {
640 		if (check_path(j, "exec.consolelog", conslog, 1, NULL) < 0)
641 			return -1;
642 		consfd =
643 		    open(conslog, O_WRONLY | O_CREAT | O_APPEND, DEFFILEMODE);
644 		if (consfd < 0) {
645 			jail_warnx(j, "open %s: %s", conslog, strerror(errno));
646 			return -1;
647 		}
648 	}
649 
650 	comlen = 0;
651 	for (i = 0; argv[i]; i++)
652 		comlen += strlen(argv[i]) + 1;
653 	j->comline = cs = emalloc(comlen);
654 	for (i = 0; argv[i]; i++) {
655 		strcpy(cs, argv[i]);
656 		if (argv[i + 1]) {
657 			cs += strlen(argv[i]) + 1;
658 			cs[-1] = ' ';
659 		}
660 	}
661 	if (verbose > 0)
662 		jail_note(j, "run command%s%s%s: %s\n",
663 		    injail ? " in jail" : "", username ? " as " : "",
664 		    username ? username : "", j->comline);
665 
666 	pid = fork();
667 	if (pid < 0)
668 		err(1, "fork");
669 	if (pid > 0) {
670 		if (bg || !add_proc(j, pid)) {
671 			free(j->comline);
672 			j->comline = NULL;
673 			return 0;
674 		} else {
675 			paralimit--;
676 			return 1;
677 		}
678 	}
679 	if (bg)
680 		setsid();
681 
682 	/* Set up the environment and run the command */
683 	pwd = NULL;
684 	lcap = NULL;
685 	if ((clean || username) && injail && sjuser &&
686 	    get_user_info(j, username, &pwd, &lcap) < 0)
687 		exit(1);
688 	if (injail) {
689 		/* jail_attach won't chdir along with its chroot. */
690 		path = string_param(j->intparams[KP_PATH]);
691 		if (path && chdir(path) < 0) {
692 			jail_warnx(j, "chdir %s: %s", path, strerror(errno));
693 			exit(1);
694 		}
695 		if (int_param(j->intparams[IP_EXEC_FIB], &fib) &&
696 		    setfib(fib) < 0) {
697 			jail_warnx(j, "setfib: %s", strerror(errno));
698 			exit(1);
699 		}
700 		if (jail_attach(j->jid) < 0) {
701 			jail_warnx(j, "jail_attach: %s", strerror(errno));
702 			exit(1);
703 		}
704 	}
705 	if (clean || username) {
706 		if (!(injail && sjuser) &&
707 		    get_user_info(j, username, &pwd, &lcap) < 0)
708 			exit(1);
709 		if (clean) {
710 			term = getenv("TERM");
711 			environ = &cleanenv;
712 			setenv("PATH", "/bin:/usr/bin", 0);
713 			if (term != NULL)
714 				setenv("TERM", term, 1);
715 		}
716 		if (setgid(pwd->pw_gid) < 0) {
717 			jail_warnx(j, "setgid %d: %s", pwd->pw_gid,
718 			    strerror(errno));
719 			exit(1);
720 		}
721 		if (setusercontext(lcap, pwd, pwd->pw_uid, username
722 		    ? LOGIN_SETALL & ~LOGIN_SETGROUP & ~LOGIN_SETLOGIN
723 		    : LOGIN_SETPATH | LOGIN_SETENV) < 0) {
724 			jail_warnx(j, "setusercontext %s: %s", pwd->pw_name,
725 			    strerror(errno));
726 			exit(1);
727 		}
728 		login_close(lcap);
729 		setenv("USER", pwd->pw_name, 1);
730 		setenv("HOME", pwd->pw_dir, 1);
731 		setenv("SHELL",
732 		    *pwd->pw_shell ? pwd->pw_shell : _PATH_BSHELL, 1);
733 		if (clean && chdir(pwd->pw_dir) < 0) {
734 			jail_warnx(j, "chdir %s: %s",
735 			    pwd->pw_dir, strerror(errno));
736 			exit(1);
737 		}
738 		endpwent();
739 	}
740 
741 	if (consfd != 0 && (dup2(consfd, 1) < 0 || dup2(consfd, 2) < 0)) {
742 		jail_warnx(j, "exec.consolelog: %s", strerror(errno));
743 		exit(1);
744 	}
745 	closefrom(3);
746 	execvp(argv[0], __DECONST(char *const*, argv));
747 	jail_warnx(j, "exec %s: %s", argv[0], strerror(errno));
748 	exit(1);
749 }
750 
751 /*
752  * Add a process to the hash, tied to a jail.
753  */
754 static int
755 add_proc(struct cfjail *j, pid_t pid)
756 {
757 	struct kevent ke;
758 	struct cfjail *tj;
759 	struct phash *ph;
760 
761 	if (!kq && (kq = kqueue()) < 0)
762 		err(1, "kqueue");
763 	EV_SET(&ke, pid, EVFILT_PROC, EV_ADD, NOTE_EXIT, 0, NULL);
764 	if (kevent(kq, &ke, 1, NULL, 0, NULL) < 0) {
765 		if (errno == ESRCH)
766 			return 0;
767 		err(1, "kevent");
768 	}
769 	ph = emalloc(sizeof(struct phash));
770 	ph->j = j;
771 	ph->pid = pid;
772 	LIST_INSERT_HEAD(&phash[pid % PHASH_SIZE], ph, le);
773 	j->nprocs++;
774 	j->flags |= JF_SLEEPQ;
775 	if (j->timeout.tv_sec == 0)
776 		requeue(j, &sleeping);
777 	else {
778 		/* File the jail in the sleep queue according to its timeout. */
779 		TAILQ_REMOVE(j->queue, j, tq);
780 		TAILQ_FOREACH(tj, &sleeping, tq) {
781 			if (!tj->timeout.tv_sec ||
782 			    j->timeout.tv_sec < tj->timeout.tv_sec ||
783 			    (j->timeout.tv_sec == tj->timeout.tv_sec &&
784 			    j->timeout.tv_nsec <= tj->timeout.tv_nsec)) {
785 				TAILQ_INSERT_BEFORE(tj, j, tq);
786 				break;
787 			}
788 		}
789 		if (tj == NULL)
790 			TAILQ_INSERT_TAIL(&sleeping, j, tq);
791 		j->queue = &sleeping;
792 	}
793 	return 1;
794 }
795 
796 /*
797  * Remove any processes from the hash that correspond to a jail.
798  */
799 static void
800 clear_procs(struct cfjail *j)
801 {
802 	struct kevent ke;
803 	struct phash *ph, *tph;
804 	int i;
805 
806 	j->nprocs = 0;
807 	for (i = 0; i < PHASH_SIZE; i++)
808 		LIST_FOREACH_SAFE(ph, &phash[i], le, tph)
809 			if (ph->j == j) {
810 				EV_SET(&ke, ph->pid, EVFILT_PROC, EV_DELETE,
811 				    NOTE_EXIT, 0, NULL);
812 				(void)kevent(kq, &ke, 1, NULL, 0, NULL);
813 				LIST_REMOVE(ph, le);
814 				free(ph);
815 			}
816 }
817 
818 /*
819  * Find the jail that corresponds to an exited process.
820  */
821 static struct cfjail *
822 find_proc(pid_t pid)
823 {
824 	struct cfjail *j;
825 	struct phash *ph;
826 
827 	LIST_FOREACH(ph, &phash[pid % PHASH_SIZE], le)
828 		if (ph->pid == pid) {
829 			j = ph->j;
830 			LIST_REMOVE(ph, le);
831 			free(ph);
832 			return --j->nprocs ? NULL : j;
833 		}
834 	return NULL;
835 }
836 
837 /*
838  * Send SIGTERM to all processes in a jail and wait for them to die.
839  */
840 static int
841 term_procs(struct cfjail *j)
842 {
843 	struct kinfo_proc *ki;
844 	int i, noted, pcnt, timeout;
845 
846 	static kvm_t *kd;
847 
848 	if (!int_param(j->intparams[IP_STOP_TIMEOUT], &timeout))
849 		timeout = DEFAULT_STOP_TIMEOUT;
850 	else if (timeout == 0)
851 		return 0;
852 
853 	if (kd == NULL) {
854 		kd = kvm_open(NULL, NULL, NULL, O_RDONLY, NULL);
855 		if (kd == NULL)
856 			return 0;
857 	}
858 
859 	ki = kvm_getprocs(kd, KERN_PROC_PROC, 0, &pcnt);
860 	if (ki == NULL)
861 		return 0;
862 	noted = 0;
863 	for (i = 0; i < pcnt; i++)
864 		if (ki[i].ki_jid == j->jid &&
865 		    kill(ki[i].ki_pid, SIGTERM) == 0) {
866 			(void)add_proc(j, ki[i].ki_pid);
867 			if (verbose > 0) {
868 				if (!noted) {
869 					noted = 1;
870 					jail_note(j, "sent SIGTERM to:");
871 				}
872 				printf(" %d", ki[i].ki_pid);
873 			}
874 		}
875 	if (noted)
876 		printf("\n");
877 	if (j->nprocs > 0) {
878 		clock_gettime(CLOCK_REALTIME, &j->timeout);
879 		j->timeout.tv_sec += timeout;
880 		return 1;
881 	}
882 	return 0;
883 }
884 
885 /*
886  * Look up a user in the passwd and login.conf files.
887  */
888 static int
889 get_user_info(struct cfjail *j, const char *username,
890     const struct passwd **pwdp, login_cap_t **lcapp)
891 {
892 	const struct passwd *pwd;
893 
894 	errno = 0;
895 	*pwdp = pwd = username ? getpwnam(username) : getpwuid(getuid());
896 	if (pwd == NULL) {
897 		if (errno)
898 			jail_warnx(j, "getpwnam%s%s: %s", username ? " " : "",
899 			    username ? username : "", strerror(errno));
900 		else if (username)
901 			jail_warnx(j, "%s: no such user", username);
902 		else
903 			jail_warnx(j, "unknown uid %d", getuid());
904 		return -1;
905 	}
906 	*lcapp = login_getpwclass(pwd);
907 	if (*lcapp == NULL) {
908 		jail_warnx(j, "getpwclass %s: %s", pwd->pw_name,
909 		    strerror(errno));
910 		return -1;
911 	}
912 	/* Set the groups while the group file is still available */
913 	if (initgroups(pwd->pw_name, pwd->pw_gid) < 0) {
914 		jail_warnx(j, "initgroups %s: %s", pwd->pw_name,
915 		    strerror(errno));
916 		return -1;
917 	}
918 	return 0;
919 }
920 
921 /*
922  * Make sure a mount or consolelog path is a valid absolute pathname
923  * with no symlinks.
924  */
925 static int
926 check_path(struct cfjail *j, const char *pname, const char *path, int isfile,
927     const char *umount_type)
928 {
929 	struct stat st, mpst;
930 	struct statfs stfs;
931 	char *tpath, *p;
932 	const char *jailpath;
933 	size_t jplen;
934 
935 	if (path[0] != '/') {
936 		jail_warnx(j, "%s: %s: not an absolute pathname",
937 		    pname, path);
938 		return -1;
939 	}
940 	/*
941 	 * Only check for symlinks in components below the jail's path,
942 	 * since that's where the security risk lies.
943 	 */
944 	jailpath = string_param(j->intparams[KP_PATH]);
945 	if (jailpath == NULL)
946 		jailpath = "";
947 	jplen = strlen(jailpath);
948 	if (!strncmp(path, jailpath, jplen) && path[jplen] == '/') {
949 		tpath = alloca(strlen(path) + 1);
950 		strcpy(tpath, path);
951 		for (p = tpath + jplen; p != NULL; ) {
952 			p = strchr(p + 1, '/');
953 			if (p)
954 				*p = '\0';
955 			if (lstat(tpath, &st) < 0) {
956 				if (errno == ENOENT && isfile && !p)
957 					break;
958 				jail_warnx(j, "%s: %s: %s", pname, tpath,
959 				    strerror(errno));
960 				return -1;
961 			}
962 			if (S_ISLNK(st.st_mode)) {
963 				jail_warnx(j, "%s: %s is a symbolic link",
964 				    pname, tpath);
965 				return -1;
966 			}
967 			if (p)
968 				*p = '/';
969 		}
970 	}
971 	if (umount_type != NULL) {
972 		if (stat(path, &st) < 0 || statfs(path, &stfs) < 0) {
973 			jail_warnx(j, "%s: %s: %s", pname, path,
974 			    strerror(errno));
975 			return -1;
976 		}
977 		if (stat(stfs.f_mntonname, &mpst) < 0) {
978 			jail_warnx(j, "%s: %s: %s", pname, stfs.f_mntonname,
979 			    strerror(errno));
980 			return -1;
981 		}
982 		if (st.st_ino != mpst.st_ino) {
983 			jail_warnx(j, "%s: %s: not a mount point",
984 			    pname, path);
985 			return -1;
986 		}
987 		if (strcmp(stfs.f_fstypename, umount_type)) {
988 			jail_warnx(j, "%s: %s: not a %s mount",
989 			    pname, path, umount_type);
990 			return -1;
991 		}
992 	}
993 	return 0;
994 }
995