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)))
153 switch (run_command(j)) {
164 * Check command exit status
167 finish_command(struct cfjail *j)
172 if (!(j->flags & JF_SLEEPQ))
174 j->flags &= ~JF_SLEEPQ;
175 if (*j->comparam == IP_STOP_TIMEOUT) {
176 j->flags &= ~JF_TIMEOUT;
181 if (!TAILQ_EMPTY(&runnable)) {
182 rj = TAILQ_FIRST(&runnable);
183 rj->flags |= JF_FROM_RUNQ;
187 if (j->flags & JF_TIMEOUT) {
188 j->flags &= ~JF_TIMEOUT;
189 if (*j->comparam != IP_STOP_TIMEOUT) {
190 jail_warnx(j, "%s: timed out", j->comline);
193 } else if (verbose > 0)
194 jail_note(j, "timed out\n");
195 } else if (j->pstatus != 0) {
196 if (WIFSIGNALED(j->pstatus))
197 jail_warnx(j, "%s: exited on signal %d",
198 j->comline, WTERMSIG(j->pstatus));
200 jail_warnx(j, "%s: failed", j->comline);
211 * Check for finished processes or timeouts.
214 next_proc(int nonblock)
218 struct timespec *tsp;
221 if (!TAILQ_EMPTY(&sleeping)) {
224 if ((j = TAILQ_FIRST(&sleeping)) && j->timeout.tv_sec) {
225 clock_gettime(CLOCK_REALTIME, &ts);
226 ts.tv_sec = j->timeout.tv_sec - ts.tv_sec;
227 ts.tv_nsec = j->timeout.tv_nsec - ts.tv_nsec;
228 if (ts.tv_nsec < 0) {
230 ts.tv_nsec += 1000000000;
233 (ts.tv_sec == 0 && ts.tv_nsec == 0)) {
234 j->flags |= JF_TIMEOUT;
245 switch (kevent(kq, NULL, 0, &ke, 1, tsp)) {
252 j = TAILQ_FIRST(&sleeping);
253 j->flags |= JF_TIMEOUT;
259 (void)waitpid(ke.ident, NULL, WNOHANG);
260 if ((j = find_proc(ke.ident))) {
261 j->pstatus = ke.data;
271 * Run a single command for a jail, possibly inside the jail.
274 run_command(struct cfjail *j)
276 const struct passwd *pwd;
277 const struct cfstring *comstring, *s;
280 char *acs, *cs, *comcs, *devpath;
281 const char *jidstr, *conslog, *path, *ruleset, *term, *username;
282 enum intparam comparam;
285 int argc, bg, clean, consfd, down, fib, i, injail, sjuser, timeout;
286 #if defined(INET) || defined(INET6)
287 char *addr, *extrap, *p, *val;
290 static char *cleanenv;
292 /* Perform some operations that aren't actually commands */
293 comparam = *j->comparam;
294 down = j->flags & (JF_STOP | JF_FAILED);
296 case IP_STOP_TIMEOUT:
297 return term_procs(j);
301 if (jail_remove(j->jid) < 0 && errno == EPERM) {
302 jail_warnx(j, "jail_remove: %s",
306 if (verbose > 0 || (verbose == 0 && (j->flags & JF_STOP
307 ? note_remove : j->name != NULL)))
308 jail_note(j, "removed\n");
310 if (j->flags & JF_STOP)
311 dep_done(j, DF_LIGHT);
313 j->flags &= ~JF_PERSIST;
315 if (create_jail(j) < 0)
318 printf("%d\n", j->jid);
319 if (verbose >= 0 && (j->name || verbose > 0))
320 jail_note(j, "created\n");
321 dep_done(j, DF_LIGHT);
328 * Collect exec arguments. Internal commands for network and
329 * mounting build their own argument lists.
331 comstring = j->comstring;
337 val = alloca(strlen(comstring->s) + 1);
338 strcpy(val, comstring->s);
341 while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
342 if (extrap == NULL) {
350 argv = alloca((8 + argc) * sizeof(char *));
351 argv[0] = _PATH_IFCONFIG;
352 if ((cs = strchr(val, '|'))) {
353 argv[1] = acs = alloca(cs - val + 1);
354 strlcpy(acs, val, cs - val + 1);
357 argv[1] = string_param(j->intparams[IP_INTERFACE]);
361 if (!(cs = strchr(addr, '/'))) {
364 argv[5] = "255.255.255.255";
366 } else if (strchr(cs + 1, '.')) {
367 argv[3] = acs = alloca(cs - addr + 1);
368 strlcpy(acs, addr, cs - addr + 1);
377 if (!down && extrap != NULL) {
378 for (cs = strtok(extrap, " "); cs;
379 cs = strtok(NULL, " ")) {
380 size_t len = strlen(cs) + 1;
381 argv[argc++] = acs = alloca(len);
382 strlcpy(acs, cs, len);
386 argv[argc] = down ? "-alias" : "alias";
387 argv[argc + 1] = NULL;
394 val = alloca(strlen(comstring->s) + 1);
395 strcpy(val, comstring->s);
398 while ((p = strchr(cs, ' ')) != NULL && strlen(p) > 1) {
399 if (extrap == NULL) {
407 argv = alloca((8 + argc) * sizeof(char *));
408 argv[0] = _PATH_IFCONFIG;
409 if ((cs = strchr(val, '|'))) {
410 argv[1] = acs = alloca(cs - val + 1);
411 strlcpy(acs, val, cs - val + 1);
414 argv[1] = string_param(j->intparams[IP_INTERFACE]);
419 if (!(cs = strchr(addr, '/'))) {
420 argv[4] = "prefixlen";
427 for (cs = strtok(extrap, " "); cs;
428 cs = strtok(NULL, " ")) {
429 size_t len = strlen(cs) + 1;
430 argv[argc++] = acs = alloca(len);
431 strlcpy(acs, cs, len);
435 argv[argc] = down ? "-alias" : "alias";
436 argv[argc + 1] = NULL;
440 case IP_VNET_INTERFACE:
441 argv = alloca(5 * sizeof(char *));
442 argv[0] = _PATH_IFCONFIG;
443 argv[1] = comstring->s;
444 argv[2] = down ? "-vnet" : "vnet";
445 jidstr = string_param(j->intparams[KP_JID]);
446 argv[3] = jidstr ? jidstr : string_param(j->intparams[KP_NAME]);
451 case IP__MOUNT_FROM_FSTAB:
452 argv = alloca(8 * sizeof(char *));
453 comcs = alloca(comstring->len + 1);
454 strcpy(comcs, comstring->s);
456 for (cs = strtok(comcs, " \t\f\v\r\n"); cs && argc < 4;
457 cs = strtok(NULL, " \t\f\v\r\n")) {
458 if (argc <= 1 && strunvis(cs, cs) < 0) {
459 jail_warnx(j, "%s: %s: fstab parse error",
460 j->intparams[comparam]->name, comstring->s);
468 jail_warnx(j, "%s: %s: missing information",
469 j->intparams[comparam]->name, comstring->s);
472 if (check_path(j, j->intparams[comparam]->name, argv[1], 0,
473 down ? argv[2] : NULL) < 0)
478 argv[0] = "/sbin/umount";
491 argv[0] = _PATH_MOUNT;
497 argv = alloca(7 * sizeof(char *));
498 path = string_param(j->intparams[KP_PATH]);
500 jail_warnx(j, "mount.devfs: no jail root path defined");
503 devpath = alloca(strlen(path) + 5);
504 sprintf(devpath, "%s/dev", path);
505 if (check_path(j, "mount.devfs", devpath, 0,
506 down ? "devfs" : NULL) < 0)
509 argv[0] = "/sbin/umount";
513 argv[0] = _PATH_MOUNT;
516 ruleset = string_param(j->intparams[KP_DEVFS_RULESET]);
518 ruleset = "4"; /* devfsrules_jail */
519 argv[3] = acs = alloca(11 + strlen(ruleset));
520 sprintf(acs, "-oruleset=%s", ruleset);
527 case IP_MOUNT_FDESCFS:
528 argv = alloca(7 * sizeof(char *));
529 path = string_param(j->intparams[KP_PATH]);
531 jail_warnx(j, "mount.fdescfs: no jail root path defined");
534 devpath = alloca(strlen(path) + 8);
535 sprintf(devpath, "%s/dev/fd", path);
536 if (check_path(j, "mount.fdescfs", devpath, 0,
537 down ? "fdescfs" : NULL) < 0)
540 argv[0] = "/sbin/umount";
544 argv[0] = _PATH_MOUNT;
553 case IP_MOUNT_PROCFS:
554 argv = alloca(7 * sizeof(char *));
555 path = string_param(j->intparams[KP_PATH]);
557 jail_warnx(j, "mount.procfs: no jail root path defined");
560 devpath = alloca(strlen(path) + 6);
561 sprintf(devpath, "%s/proc", path);
562 if (check_path(j, "mount.procfs", devpath, 0,
563 down ? "procfs" : NULL) < 0)
566 argv[0] = "/sbin/umount";
570 argv[0] = _PATH_MOUNT;
581 goto default_command;
583 TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
585 argv = alloca((argc + 1) * sizeof(char *));
587 TAILQ_FOREACH(s, &j->intparams[IP_COMMAND]->val, tq)
590 j->comstring = &dummystring;
595 if ((cs = strpbrk(comstring->s, "!\"$&'()*;<>?[\\]`{|}~")) &&
596 !(cs[0] == '&' && cs[1] == '\0')) {
597 argv = alloca(4 * sizeof(char *));
598 argv[0] = _PATH_BSHELL;
600 argv[2] = comstring->s;
607 comcs = alloca(comstring->len + 1);
608 strcpy(comcs, comstring->s);
610 for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
611 cs = strtok(NULL, " \t\f\v\r\n"))
613 argv = alloca((argc + 1) * sizeof(char *));
614 strcpy(comcs, comstring->s);
616 for (cs = strtok(comcs, " \t\f\v\r\n"); cs;
617 cs = strtok(NULL, " \t\f\v\r\n"))
625 if (int_param(j->intparams[IP_EXEC_TIMEOUT], &timeout) &&
627 clock_gettime(CLOCK_REALTIME, &j->timeout);
628 j->timeout.tv_sec += timeout;
630 j->timeout.tv_sec = 0;
632 injail = comparam == IP_EXEC_START || comparam == IP_COMMAND ||
633 comparam == IP_EXEC_STOP;
634 clean = bool_param(j->intparams[IP_EXEC_CLEAN]);
635 username = string_param(j->intparams[injail
636 ? IP_EXEC_JAIL_USER : IP_EXEC_SYSTEM_USER]);
637 sjuser = bool_param(j->intparams[IP_EXEC_SYSTEM_JAIL_USER]);
641 (conslog = string_param(j->intparams[IP_EXEC_CONSOLELOG]))) {
642 if (check_path(j, "exec.consolelog", conslog, 1, NULL) < 0)
645 open(conslog, O_WRONLY | O_CREAT | O_APPEND, DEFFILEMODE);
647 jail_warnx(j, "open %s: %s", conslog, strerror(errno));
653 for (i = 0; argv[i]; i++)
654 comlen += strlen(argv[i]) + 1;
655 j->comline = cs = emalloc(comlen);
656 for (i = 0; argv[i]; i++) {
659 cs += strlen(argv[i]) + 1;
664 jail_note(j, "run command%s%s%s: %s\n",
665 injail ? " in jail" : "", username ? " as " : "",
666 username ? username : "", j->comline);
672 if (bg || !add_proc(j, pid)) {
684 /* Set up the environment and run the command */
687 if ((clean || username) && injail && sjuser &&
688 get_user_info(j, username, &pwd, &lcap) < 0)
691 /* jail_attach won't chdir along with its chroot. */
692 path = string_param(j->intparams[KP_PATH]);
693 if (path && chdir(path) < 0) {
694 jail_warnx(j, "chdir %s: %s", path, strerror(errno));
697 if (int_param(j->intparams[IP_EXEC_FIB], &fib) &&
699 jail_warnx(j, "setfib: %s", strerror(errno));
702 if (jail_attach(j->jid) < 0) {
703 jail_warnx(j, "jail_attach: %s", strerror(errno));
707 if (clean || username) {
708 if (!(injail && sjuser) &&
709 get_user_info(j, username, &pwd, &lcap) < 0)
712 term = getenv("TERM");
714 setenv("PATH", "/bin:/usr/bin", 0);
716 setenv("TERM", term, 1);
718 if (setgid(pwd->pw_gid) < 0) {
719 jail_warnx(j, "setgid %d: %s", pwd->pw_gid,
723 if (setusercontext(lcap, pwd, pwd->pw_uid, username
724 ? LOGIN_SETALL & ~LOGIN_SETGROUP & ~LOGIN_SETLOGIN
725 : LOGIN_SETPATH | LOGIN_SETENV) < 0) {
726 jail_warnx(j, "setusercontext %s: %s", pwd->pw_name,
731 setenv("USER", pwd->pw_name, 1);
732 setenv("HOME", pwd->pw_dir, 1);
734 *pwd->pw_shell ? pwd->pw_shell : _PATH_BSHELL, 1);
735 if (clean && chdir(pwd->pw_dir) < 0) {
736 jail_warnx(j, "chdir %s: %s",
737 pwd->pw_dir, strerror(errno));
743 if (consfd != 0 && (dup2(consfd, 1) < 0 || dup2(consfd, 2) < 0)) {
744 jail_warnx(j, "exec.consolelog: %s", strerror(errno));
748 execvp(argv[0], __DECONST(char *const*, argv));
749 jail_warnx(j, "exec %s: %s", argv[0], strerror(errno));
754 * Add a process to the hash, tied to a jail.
757 add_proc(struct cfjail *j, pid_t pid)
763 if (!kq && (kq = kqueue()) < 0)
765 EV_SET(&ke, pid, EVFILT_PROC, EV_ADD, NOTE_EXIT, 0, NULL);
766 if (kevent(kq, &ke, 1, NULL, 0, NULL) < 0) {
771 ph = emalloc(sizeof(struct phash));
774 LIST_INSERT_HEAD(&phash[pid % PHASH_SIZE], ph, le);
776 j->flags |= JF_SLEEPQ;
777 if (j->timeout.tv_sec == 0)
778 requeue(j, &sleeping);
780 /* File the jail in the sleep queue according to its timeout. */
781 TAILQ_REMOVE(j->queue, j, tq);
782 TAILQ_FOREACH(tj, &sleeping, tq) {
783 if (!tj->timeout.tv_sec ||
784 j->timeout.tv_sec < tj->timeout.tv_sec ||
785 (j->timeout.tv_sec == tj->timeout.tv_sec &&
786 j->timeout.tv_nsec <= tj->timeout.tv_nsec)) {
787 TAILQ_INSERT_BEFORE(tj, j, tq);
792 TAILQ_INSERT_TAIL(&sleeping, j, tq);
793 j->queue = &sleeping;
799 * Remove any processes from the hash that correspond to a jail.
802 clear_procs(struct cfjail *j)
805 struct phash *ph, *tph;
809 for (i = 0; i < PHASH_SIZE; i++)
810 LIST_FOREACH_SAFE(ph, &phash[i], le, tph)
812 EV_SET(&ke, ph->pid, EVFILT_PROC, EV_DELETE,
814 (void)kevent(kq, &ke, 1, NULL, 0, NULL);
821 * Find the jail that corresponds to an exited process.
823 static struct cfjail *
829 LIST_FOREACH(ph, &phash[pid % PHASH_SIZE], le)
830 if (ph->pid == pid) {
834 return --j->nprocs ? NULL : j;
840 * Send SIGTERM to all processes in a jail and wait for them to die.
843 term_procs(struct cfjail *j)
845 struct kinfo_proc *ki;
846 int i, noted, pcnt, timeout;
850 if (!int_param(j->intparams[IP_STOP_TIMEOUT], &timeout))
851 timeout = DEFAULT_STOP_TIMEOUT;
852 else if (timeout == 0)
856 kd = kvm_open(NULL, NULL, NULL, O_RDONLY, NULL);
861 ki = kvm_getprocs(kd, KERN_PROC_PROC, 0, &pcnt);
865 for (i = 0; i < pcnt; i++)
866 if (ki[i].ki_jid == j->jid &&
867 kill(ki[i].ki_pid, SIGTERM) == 0) {
868 (void)add_proc(j, ki[i].ki_pid);
872 jail_note(j, "sent SIGTERM to:");
874 printf(" %d", ki[i].ki_pid);
880 clock_gettime(CLOCK_REALTIME, &j->timeout);
881 j->timeout.tv_sec += timeout;
888 * Look up a user in the passwd and login.conf files.
891 get_user_info(struct cfjail *j, const char *username,
892 const struct passwd **pwdp, login_cap_t **lcapp)
894 const struct passwd *pwd;
897 *pwdp = pwd = username ? getpwnam(username) : getpwuid(getuid());
900 jail_warnx(j, "getpwnam%s%s: %s", username ? " " : "",
901 username ? username : "", strerror(errno));
903 jail_warnx(j, "%s: no such user", username);
905 jail_warnx(j, "unknown uid %d", getuid());
908 *lcapp = login_getpwclass(pwd);
909 if (*lcapp == NULL) {
910 jail_warnx(j, "getpwclass %s: %s", pwd->pw_name,
914 /* Set the groups while the group file is still available */
915 if (initgroups(pwd->pw_name, pwd->pw_gid) < 0) {
916 jail_warnx(j, "initgroups %s: %s", pwd->pw_name,
924 * Make sure a mount or consolelog path is a valid absolute pathname
928 check_path(struct cfjail *j, const char *pname, const char *path, int isfile,
929 const char *umount_type)
931 struct stat st, mpst;
934 const char *jailpath;
937 if (path[0] != '/') {
938 jail_warnx(j, "%s: %s: not an absolute pathname",
943 * Only check for symlinks in components below the jail's path,
944 * since that's where the security risk lies.
946 jailpath = string_param(j->intparams[KP_PATH]);
947 if (jailpath == NULL)
949 jplen = strlen(jailpath);
950 if (!strncmp(path, jailpath, jplen) && path[jplen] == '/') {
951 tpath = alloca(strlen(path) + 1);
953 for (p = tpath + jplen; p != NULL; ) {
954 p = strchr(p + 1, '/');
957 if (lstat(tpath, &st) < 0) {
958 if (errno == ENOENT && isfile && !p)
960 jail_warnx(j, "%s: %s: %s", pname, tpath,
964 if (S_ISLNK(st.st_mode)) {
965 jail_warnx(j, "%s: %s is a symbolic link",
973 if (umount_type != NULL) {
974 if (stat(path, &st) < 0 || statfs(path, &stfs) < 0) {
975 jail_warnx(j, "%s: %s: %s", pname, path,
979 if (stat(stfs.f_mntonname, &mpst) < 0) {
980 jail_warnx(j, "%s: %s: %s", pname, stfs.f_mntonname,
984 if (st.st_ino != mpst.st_ino) {
985 jail_warnx(j, "%s: %s: not a mount point",
989 if (strcmp(stfs.f_fstypename, umount_type)) {
990 jail_warnx(j, "%s: %s: not a %s mount",
991 pname, path, umount_type);