2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2011 James Gritton
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/types.h>
33 #include <sys/event.h>
34 #include <sys/mount.h>
36 #include <sys/sysctl.h>
44 #include <login_cap.h>
56 #define DEFAULT_STOP_TIMEOUT 10
57 #define PHASH_SIZE 256
59 LIST_HEAD(phhead, phash);
69 extern char **environ;
71 static int run_command(struct cfjail *j);
72 static int add_proc(struct cfjail *j, pid_t pid);
73 static void clear_procs(struct cfjail *j);
74 static struct cfjail *find_proc(pid_t pid);
75 static int term_procs(struct cfjail *j);
76 static int get_user_info(struct cfjail *j, const char *username,
77 const struct passwd **pwdp, login_cap_t **lcapp);
78 static int check_path(struct cfjail *j, const char *pname, const char *path,
79 int isfile, const char *umount_type);
81 static struct cfjails sleeping = TAILQ_HEAD_INITIALIZER(sleeping);
82 static struct cfjails runnable = TAILQ_HEAD_INITIALIZER(runnable);
83 static struct cfstring dummystring = { .len = 1 };
84 static struct phhead phash[PHASH_SIZE];
88 * Run the next command associated with a jail.
91 next_command(struct cfjail *j)
93 enum intparam comparam;
94 int create_failed, stopping;
97 if (j->flags & JF_FROM_RUNQ)
98 requeue_head(j, &runnable);
100 requeue(j, &runnable);
103 j->flags &= ~JF_FROM_RUNQ;
104 create_failed = (j->flags & (JF_STOP | JF_FAILED)) == JF_FAILED;
105 stopping = (j->flags & JF_STOP) != 0;
106 comparam = *j->comparam;
108 if (j->comstring == NULL) {
109 j->comparam += create_failed ? -1 : 1;
110 switch ((comparam = *j->comparam)) {
114 if (!bool_param(j->intparams[IP_MOUNT_DEVFS]))
116 j->comstring = &dummystring;
118 case IP_MOUNT_FDESCFS:
119 if (!bool_param(j->intparams[IP_MOUNT_FDESCFS]))
121 j->comstring = &dummystring;
123 case IP_MOUNT_PROCFS:
124 if (!bool_param(j->intparams[IP_MOUNT_PROCFS]))
126 j->comstring = &dummystring;
129 case IP_STOP_TIMEOUT:
130 j->comstring = &dummystring;
133 if (j->intparams[comparam] == NULL)
135 j->comstring = create_failed || (stopping &&
136 (j->intparams[comparam]->flags & PF_REV))
137 ? TAILQ_LAST(&j->intparams[comparam]->val,
139 : TAILQ_FIRST(&j->intparams[comparam]->val);
142 j->comstring = j->comstring == &dummystring ? NULL :
143 create_failed || (stopping &&
144 (j->intparams[comparam]->flags & PF_REV))
145 ? TAILQ_PREV(j->comstring, cfstrings, tq)
146 : TAILQ_NEXT(j->comstring, tq);
148 if (j->comstring == NULL || j->comstring->len == 0 ||
149 (create_failed && (comparam == IP_EXEC_PRESTART ||
150 comparam == IP_EXEC_CREATED || comparam == IP_EXEC_START ||
151 comparam == IP_COMMAND || comparam == IP_EXEC_POSTSTART ||
152 comparam == IP_EXEC_PREPARE)))
154 switch (run_command(j)) {
165 * Check command exit status
168 finish_command(struct cfjail *j)
173 if (!(j->flags & JF_SLEEPQ))
175 j->flags &= ~JF_SLEEPQ;
176 if (*j->comparam == IP_STOP_TIMEOUT) {
177 j->flags &= ~JF_TIMEOUT;
182 if (!TAILQ_EMPTY(&runnable)) {
183 rj = TAILQ_FIRST(&runnable);
184 rj->flags |= JF_FROM_RUNQ;
188 if (j->flags & JF_TIMEOUT) {
189 j->flags &= ~JF_TIMEOUT;
190 if (*j->comparam != IP_STOP_TIMEOUT) {
191 jail_warnx(j, "%s: timed out", j->comline);
194 } else if (verbose > 0)
195 jail_note(j, "timed out\n");
196 } else if (j->pstatus != 0) {
197 if (WIFSIGNALED(j->pstatus))
198 jail_warnx(j, "%s: exited on signal %d",
199 j->comline, WTERMSIG(j->pstatus));
201 jail_warnx(j, "%s: failed", j->comline);
212 * Check for finished processes or timeouts.
215 next_proc(int nonblock)
219 struct timespec *tsp;
222 if (!TAILQ_EMPTY(&sleeping)) {
225 if ((j = TAILQ_FIRST(&sleeping)) && j->timeout.tv_sec) {
226 clock_gettime(CLOCK_REALTIME, &ts);
227 ts.tv_sec = j->timeout.tv_sec - ts.tv_sec;
228 ts.tv_nsec = j->timeout.tv_nsec - ts.tv_nsec;
229 if (ts.tv_nsec < 0) {
231 ts.tv_nsec += 1000000000;
234 (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
235 j->flags |= JF_TIMEOUT;
246 switch (kevent(kq, NULL, 0, &ke, 1, tsp)) {
253 j = TAILQ_FIRST(&sleeping);
254 j->flags |= JF_TIMEOUT;
260 (void)waitpid(ke.ident, NULL, WNOHANG);
261 if ((j = find_proc(ke.ident))) {
262 j->pstatus = ke.data;
272 * Run a single command for a jail, possibly inside the jail.
275 run_command(struct cfjail *j)
277 const struct passwd *pwd;
278 const struct cfstring *comstring, *s;
281 char *acs, *cs, *comcs, *devpath;
282 const char *jidstr, *conslog, *path, *ruleset, *term, *username;
283 enum intparam comparam;
286 int argc, bg, clean, consfd, down, fib, i, injail, sjuser, timeout;
287 #if defined(INET) || defined(INET6)
288 char *addr, *extrap, *p, *val;
291 static char *cleanenv;
293 /* Perform some operations that aren't actually commands */
294 comparam = *j->comparam;
295 down = j->flags & (JF_STOP | JF_FAILED);
297 case IP_STOP_TIMEOUT:
298 return term_procs(j);
302 if (jail_remove(j->jid) < 0 && errno == EPERM) {
303 jail_warnx(j, "jail_remove: %s",
307 if (verbose > 0 || (verbose == 0 && (j->flags & JF_STOP
308 ? note_remove : j->name != NULL)))
309 jail_note(j, "removed\n");
311 if (j->flags & JF_STOP)
312 dep_done(j, DF_LIGHT);
314 j->flags &= ~JF_PERSIST;
316 if (create_jail(j) < 0)
319 printf("%d\n", j->jid);
320 if (verbose >= 0 && (j->name || verbose > 0))
321 jail_note(j, "created\n");
322 dep_done(j, DF_LIGHT);
329 * Collect exec arguments. Internal commands for network and
330 * mounting build their own argument lists.
332 comstring = j->comstring;
338 val = alloca(strlen(comstring->s) + 1);
339 strcpy(val, comstring->s);
342 while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
343 if (extrap == NULL) {
351 argv = alloca((8 + argc) * sizeof(char *));
352 argv[0] = _PATH_IFCONFIG;
353 if ((cs = strchr(val, '|'))) {
354 argv[1] = acs = alloca(cs - val + 1);
355 strlcpy(acs, val, cs - val + 1);
358 argv[1] = string_param(j->intparams[IP_INTERFACE]);
362 if (!(cs = strchr(addr, '/'))) {
365 argv[5] = "255.255.255.255";
367 } else if (strchr(cs + 1, '.')) {
368 argv[3] = acs = alloca(cs - addr + 1);
369 strlcpy(acs, addr, cs - addr + 1);
378 if (!down && extrap != NULL) {
379 for (cs = strtok(extrap, " "); cs;
380 cs = strtok(NULL, " ")) {
381 size_t len = strlen(cs) + 1;
382 argv[argc++] = acs = alloca(len);
383 strlcpy(acs, cs, len);
387 argv[argc] = down ? "-alias" : "alias";
388 argv[argc + 1] = NULL;
395 val = alloca(strlen(comstring->s) + 1);
396 strcpy(val, comstring->s);
399 while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
400 if (extrap == NULL) {
408 argv = alloca((8 + argc) * sizeof(char *));
409 argv[0] = _PATH_IFCONFIG;
410 if ((cs = strchr(val, '|'))) {
411 argv[1] = acs = alloca(cs - val + 1);
412 strlcpy(acs, val, cs - val + 1);
415 argv[1] = string_param(j->intparams[IP_INTERFACE]);
420 if (!(cs = strchr(addr, '/'))) {
421 argv[4] = "prefixlen";
428 for (cs = strtok(extrap, " "); cs;
429 cs = strtok(NULL, " ")) {
430 size_t len = strlen(cs) + 1;
431 argv[argc++] = acs = alloca(len);
432 strlcpy(acs, cs, len);
436 argv[argc] = down ? "-alias" : "alias";
437 argv[argc + 1] = NULL;
441 case IP_VNET_INTERFACE:
442 argv = alloca(5 * sizeof(char *));
443 argv[0] = _PATH_IFCONFIG;
444 argv[1] = comstring->s;
445 argv[2] = down ? "-vnet" : "vnet";
446 jidstr = string_param(j->intparams[KP_JID]);
447 argv[3] = jidstr ? jidstr : string_param(j->intparams[KP_NAME]);
452 case IP__MOUNT_FROM_FSTAB:
453 argv = alloca(8 * sizeof(char *));
454 comcs = alloca(comstring->len + 1);
455 strcpy(comcs, comstring->s);
457 for (cs = strtok(comcs, " \t\f\v\r\n"); cs && argc < 4;
458 cs = strtok(NULL, " \t\f\v\r\n")) {
459 if (argc <= 1 && strunvis(cs, cs) < 0) {
460 jail_warnx(j, "%s: %s: fstab parse error",
461 j->intparams[comparam]->name, comstring->s);
469 jail_warnx(j, "%s: %s: missing information",
470 j->intparams[comparam]->name, comstring->s);
473 if (check_path(j, j->intparams[comparam]->name, argv[1], 0,
474 down ? argv[2] : NULL) < 0)
480 argv[0] = "/sbin/umount";
494 argv[0] = _PATH_MOUNT;
499 argv = alloca(7 * sizeof(char *));
500 path = string_param(j->intparams[KP_PATH]);
502 jail_warnx(j, "mount.devfs: no jail root path defined");
505 devpath = alloca(strlen(path) + 5);
506 sprintf(devpath, "%s/dev", path);
507 if (check_path(j, "mount.devfs", devpath, 0,
508 down ? "devfs" : NULL) < 0)
511 argv[0] = "/sbin/umount";
515 argv[0] = _PATH_MOUNT;
518 ruleset = string_param(j->intparams[KP_DEVFS_RULESET]);
520 ruleset = "4"; /* devfsrules_jail */
521 argv[3] = acs = alloca(11 + strlen(ruleset));
522 sprintf(acs, "-oruleset=%s", ruleset);
529 case IP_MOUNT_FDESCFS:
530 argv = alloca(7 * sizeof(char *));
531 path = string_param(j->intparams[KP_PATH]);
533 jail_warnx(j, "mount.fdescfs: no jail root path defined");
536 devpath = alloca(strlen(path) + 8);
537 sprintf(devpath, "%s/dev/fd", path);
538 if (check_path(j, "mount.fdescfs", devpath, 0,
539 down ? "fdescfs" : NULL) < 0)
542 argv[0] = "/sbin/umount";
546 argv[0] = _PATH_MOUNT;
555 case IP_MOUNT_PROCFS:
556 argv = alloca(7 * sizeof(char *));
557 path = string_param(j->intparams[KP_PATH]);
559 jail_warnx(j, "mount.procfs: no jail root path defined");
562 devpath = alloca(strlen(path) + 6);
563 sprintf(devpath, "%s/proc", path);
564 if (check_path(j, "mount.procfs", devpath, 0,
565 down ? "procfs" : NULL) < 0)
568 argv[0] = "/sbin/umount";
572 argv[0] = _PATH_MOUNT;
583 goto default_command;
585 TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
587 argv = alloca((argc + 1) * sizeof(char *));
589 TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
592 j->comstring = &dummystring;
597 if ((cs = strpbrk(comstring->s, "!\"$&'()*;<>?[\\]`{|}~")) &&
598 !(cs[0] == '&' && cs[1] == '\0')) {
599 argv = alloca(4 * sizeof(char *));
600 argv[0] = _PATH_BSHELL;
602 argv[2] = comstring->s;
609 comcs = alloca(comstring->len + 1);
610 strcpy(comcs, comstring->s);
612 for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
613 cs = strtok(NULL, " \t\f\v\r\n"))
615 argv = alloca((argc + 1) * sizeof(char *));
616 strcpy(comcs, comstring->s);
618 for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
619 cs = strtok(NULL, " \t\f\v\r\n"))
627 if (int_param(j->intparams[IP_EXEC_TIMEOUT], &timeout) &&
629 clock_gettime(CLOCK_REALTIME, &j->timeout);
630 j->timeout.tv_sec += timeout;
632 j->timeout.tv_sec = 0;
634 injail = comparam == IP_EXEC_START || comparam == IP_COMMAND ||
635 comparam == IP_EXEC_STOP;
636 clean = bool_param(j->intparams[IP_EXEC_CLEAN]);
637 username = string_param(j->intparams[injail
638 ? IP_EXEC_JAIL_USER : IP_EXEC_SYSTEM_USER]);
639 sjuser = bool_param(j->intparams[IP_EXEC_SYSTEM_JAIL_USER]);
643 (conslog = string_param(j->intparams[IP_EXEC_CONSOLELOG]))) {
644 if (check_path(j, "exec.consolelog", conslog, 1, NULL) < 0)
647 open(conslog, O_WRONLY | O_CREAT | O_APPEND, DEFFILEMODE);
649 jail_warnx(j, "open %s: %s", conslog, strerror(errno));
655 for (i = 0; argv[i]; i++)
656 comlen += strlen(argv[i]) + 1;
657 j->comline = cs = emalloc(comlen);
658 for (i = 0; argv[i]; i++) {
661 cs += strlen(argv[i]) + 1;
666 jail_note(j, "run command%s%s%s: %s\n",
667 injail ? " in jail" : "", username ? " as " : "",
668 username ? username : "", j->comline);
674 if (bg || !add_proc(j, pid)) {
686 /* Set up the environment and run the command */
689 if ((clean || username) && injail && sjuser &&
690 get_user_info(j, username, &pwd, &lcap) < 0)
693 /* jail_attach won't chdir along with its chroot. */
694 path = string_param(j->intparams[KP_PATH]);
695 if (path && chdir(path) < 0) {
696 jail_warnx(j, "chdir %s: %s", path, strerror(errno));
699 if (int_param(j->intparams[IP_EXEC_FIB], &fib) &&
701 jail_warnx(j, "setfib: %s", strerror(errno));
704 if (jail_attach(j->jid) < 0) {
705 jail_warnx(j, "jail_attach: %s", strerror(errno));
709 if (clean || username) {
710 if (!(injail && sjuser) &&
711 get_user_info(j, username, &pwd, &lcap) < 0)
714 term = getenv("TERM");
716 setenv("PATH", "/bin:/usr/bin", 0);
718 setenv("TERM", term, 1);
720 if (setgid(pwd->pw_gid) < 0) {
721 jail_warnx(j, "setgid %d: %s", pwd->pw_gid,
725 if (setusercontext(lcap, pwd, pwd->pw_uid, username
726 ? LOGIN_SETALL & ~LOGIN_SETGROUP & ~LOGIN_SETLOGIN
727 : LOGIN_SETPATH | LOGIN_SETENV) < 0) {
728 jail_warnx(j, "setusercontext %s: %s", pwd->pw_name,
733 setenv("USER", pwd->pw_name, 1);
734 setenv("HOME", pwd->pw_dir, 1);
736 *pwd->pw_shell ? pwd->pw_shell : _PATH_BSHELL, 1);
737 if (clean && chdir(pwd->pw_dir) < 0) {
738 jail_warnx(j, "chdir %s: %s",
739 pwd->pw_dir, strerror(errno));
745 if (consfd != 0 && (dup2(consfd, 1) < 0 || dup2(consfd, 2) < 0)) {
746 jail_warnx(j, "exec.consolelog: %s", strerror(errno));
750 execvp(argv[0], __DECONST(char *const*, argv));
751 jail_warnx(j, "exec %s: %s", argv[0], strerror(errno));
756 * Add a process to the hash, tied to a jail.
759 add_proc(struct cfjail *j, pid_t pid)
765 if (!kq && (kq = kqueue()) < 0)
767 EV_SET(&ke, pid, EVFILT_PROC, EV_ADD, NOTE_EXIT, 0, NULL);
768 if (kevent(kq, &ke, 1, NULL, 0, NULL) < 0) {
773 ph = emalloc(sizeof(struct phash));
776 LIST_INSERT_HEAD(&phash[pid % PHASH_SIZE], ph, le);
778 j->flags |= JF_SLEEPQ;
779 if (j->timeout.tv_sec == 0)
780 requeue(j, &sleeping);
782 /* File the jail in the sleep queue according to its timeout. */
783 TAILQ_REMOVE(j->queue, j, tq);
784 TAILQ_FOREACH(tj, &sleeping, tq) {
785 if (!tj->timeout.tv_sec ||
786 j->timeout.tv_sec < tj->timeout.tv_sec ||
787 (j->timeout.tv_sec == tj->timeout.tv_sec &&
788 j->timeout.tv_nsec <= tj->timeout.tv_nsec)) {
789 TAILQ_INSERT_BEFORE(tj, j, tq);
794 TAILQ_INSERT_TAIL(&sleeping, j, tq);
795 j->queue = &sleeping;
801 * Remove any processes from the hash that correspond to a jail.
804 clear_procs(struct cfjail *j)
807 struct phash *ph, *tph;
811 for (i = 0; i < PHASH_SIZE; i++)
812 LIST_FOREACH_SAFE(ph, &phash[i], le, tph)
814 EV_SET(&ke, ph->pid, EVFILT_PROC, EV_DELETE,
816 (void)kevent(kq, &ke, 1, NULL, 0, NULL);
823 * Find the jail that corresponds to an exited process.
825 static struct cfjail *
831 LIST_FOREACH(ph, &phash[pid % PHASH_SIZE], le)
832 if (ph->pid == pid) {
836 return --j->nprocs ? NULL : j;
842 * Send SIGTERM to all processes in a jail and wait for them to die.
845 term_procs(struct cfjail *j)
847 struct kinfo_proc *ki;
848 int i, noted, pcnt, timeout;
852 if (!int_param(j->intparams[IP_STOP_TIMEOUT], &timeout))
853 timeout = DEFAULT_STOP_TIMEOUT;
854 else if (timeout == 0)
858 kd = kvm_open(NULL, NULL, NULL, O_RDONLY, NULL);
863 ki = kvm_getprocs(kd, KERN_PROC_PROC, 0, &pcnt);
867 for (i = 0; i < pcnt; i++)
868 if (ki[i].ki_jid == j->jid &&
869 kill(ki[i].ki_pid, SIGTERM) == 0) {
870 (void)add_proc(j, ki[i].ki_pid);
874 jail_note(j, "sent SIGTERM to:");
876 printf(" %d", ki[i].ki_pid);
882 clock_gettime(CLOCK_REALTIME, &j->timeout);
883 j->timeout.tv_sec += timeout;
890 * Look up a user in the passwd and login.conf files.
893 get_user_info(struct cfjail *j, const char *username,
894 const struct passwd **pwdp, login_cap_t **lcapp)
896 const struct passwd *pwd;
899 *pwdp = pwd = username ? getpwnam(username) : getpwuid(getuid());
902 jail_warnx(j, "getpwnam%s%s: %s", username ? " " : "",
903 username ? username : "", strerror(errno));
905 jail_warnx(j, "%s: no such user", username);
907 jail_warnx(j, "unknown uid %d", getuid());
910 *lcapp = login_getpwclass(pwd);
911 if (*lcapp == NULL) {
912 jail_warnx(j, "getpwclass %s: %s", pwd->pw_name,
916 /* Set the groups while the group file is still available */
917 if (initgroups(pwd->pw_name, pwd->pw_gid) < 0) {
918 jail_warnx(j, "initgroups %s: %s", pwd->pw_name,
926 * Make sure a mount or consolelog path is a valid absolute pathname
930 check_path(struct cfjail *j, const char *pname, const char *path, int isfile,
931 const char *umount_type)
933 struct stat st, mpst;
936 const char *jailpath;
939 if (path[0] != '/') {
940 jail_warnx(j, "%s: %s: not an absolute pathname",
945 * Only check for symlinks in components below the jail's path,
946 * since that's where the security risk lies.
948 jailpath = string_param(j->intparams[KP_PATH]);
949 if (jailpath == NULL)
951 jplen = strlen(jailpath);
952 if (!strncmp(path, jailpath, jplen) && path[jplen] == '/') {
953 tpath = alloca(strlen(path) + 1);
955 for (p = tpath + jplen; p != NULL; ) {
956 p = strchr(p + 1, '/');
959 if (lstat(tpath, &st) < 0) {
960 if (errno == ENOENT && isfile && !p)
962 jail_warnx(j, "%s: %s: %s", pname, tpath,
966 if (S_ISLNK(st.st_mode)) {
967 jail_warnx(j, "%s: %s is a symbolic link",
975 if (umount_type != NULL) {
976 if (stat(path, &st) < 0 || statfs(path, &stfs) < 0) {
977 jail_warnx(j, "%s: %s: %s", pname, path,
981 if (stat(stfs.f_mntonname, &mpst) < 0) {
982 jail_warnx(j, "%s: %s: %s", pname, stfs.f_mntonname,
986 if (st.st_ino != mpst.st_ino) {
987 jail_warnx(j, "%s: %s: not a mount point",
991 if (strcmp(stfs.f_fstypename, umount_type)) {
992 jail_warnx(j, "%s: %s: not a %s mount",
993 pname, path, umount_type);