2 * Copyright (C) 2015 John Crispin <blogic@openwrt.org>
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU Lesser General Public License version 2.1
6 * as published by the Free Software Foundation
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
15 #include <sys/mount.h>
16 #include <sys/prctl.h>
27 #include <linux/limits.h>
30 #include "capabilities.h"
36 #include <libubox/uloop.h>
38 #define STACK_SIZE (1024 * 1024)
39 #define OPT_ARGS "S:C:n:h:r:w:d:psuloc"
54 extern int pivot_root(const char *new_root, const char *put_old);
58 static char child_stack[STACK_SIZE];
60 static int mkdir_p(char *dir, mode_t mask)
62 char *l = strrchr(dir, '/');
70 if (mkdir_p(dir, mask))
75 ret = mkdir(dir, mask);
76 if (ret && errno == EEXIST)
80 ERROR("mkdir(%s, %d) failed: %s\n", dir, mask, strerror(errno));
85 int mount_bind(const char *root, const char *path, int readonly, int error)
92 ERROR("stat(%s) failed: %s\n", path, strerror(errno));
96 snprintf(new, sizeof(new), "%s%s", root, path);
97 if (S_ISDIR(s.st_mode)) {
100 mkdir_p(dirname(new), 0755);
101 snprintf(new, sizeof(new), "%s%s", root, path);
102 fd = creat(new, 0644);
104 ERROR("creat(%s) failed: %s\n", new, strerror(errno));
110 if (mount(path, new, NULL, MS_BIND, NULL)) {
111 ERROR("failed to mount -B %s %s: %s\n", path, new, strerror(errno));
115 if (readonly && mount(NULL, new, NULL, MS_BIND | MS_REMOUNT | MS_RDONLY, NULL)) {
116 ERROR("failed to remount ro %s: %s\n", new, strerror(errno));
120 DEBUG("mount -B %s %s (%s)\n", path, new, readonly?"ro":"rw");
125 static int build_jail_fs(void)
127 char jail_root[] = "/tmp/ujail-XXXXXX";
128 if (mkdtemp(jail_root) == NULL) {
129 ERROR("mkdtemp(%s) failed: %s\n", jail_root, strerror(errno));
133 /* oldroot can't be MS_SHARED else pivot_root() fails */
134 if (mount("none", "/", NULL, MS_REC|MS_PRIVATE, NULL)) {
135 ERROR("private mount failed %s\n", strerror(errno));
139 if (mount("tmpfs", jail_root, "tmpfs", MS_NOATIME, "mode=0755")) {
140 ERROR("tmpfs mount failed %s\n", strerror(errno));
144 if (chdir(jail_root)) {
145 ERROR("chdir(%s) (jail_root) failed: %s\n", jail_root, strerror(errno));
149 if (mount_all(jail_root)) {
150 ERROR("mount_all() failed\n");
154 char dirbuf[sizeof(jail_root) + 4];
155 snprintf(dirbuf, sizeof(dirbuf), "%s/old", jail_root);
158 if (pivot_root(jail_root, dirbuf) == -1) {
159 ERROR("pivot_root(%s, %s) failed: %s\n", jail_root, dirbuf, strerror(errno));
163 ERROR("chdir(/) (after pivot_root) failed: %s\n", strerror(errno));
167 snprintf(dirbuf, sizeof(dirbuf), "/old%s", jail_root);
169 umount2("/old", MNT_DETACH);
173 mkdir("/proc", 0755);
174 mount("proc", "/proc", "proc", MS_NOATIME | MS_NODEV | MS_NOEXEC | MS_NOSUID, 0);
178 mount("sysfs", "/sys", "sysfs", MS_NOATIME | MS_NODEV | MS_NOEXEC | MS_NOSUID, 0);
181 mount(NULL, "/", NULL, MS_RDONLY | MS_REMOUNT, 0);
187 static char** build_envp(const char *seccomp)
189 static char *envp[MAX_ENVP];
190 static char preload_var[PATH_MAX];
191 static char seccomp_var[PATH_MAX];
192 static char debug_var[] = "LD_DEBUG=all";
193 const char *preload_lib = find_lib("libpreload-seccomp.so");
196 if (seccomp && !preload_lib) {
197 ERROR("failed to add preload-lib to env\n");
201 snprintf(seccomp_var, sizeof(seccomp_var), "SECCOMP_FILE=%s", seccomp);
202 envp[count++] = seccomp_var;
203 snprintf(preload_var, sizeof(preload_var), "LD_PRELOAD=%s", preload_lib);
204 envp[count++] = preload_var;
207 envp[count++] = debug_var;
212 static void usage(void)
214 fprintf(stderr, "ujail <options> -- <binary> <params ...>\n");
215 fprintf(stderr, " -d <num>\tshow debug log (increase num to increase verbosity)\n");
216 fprintf(stderr, " -S <file>\tseccomp filter config\n");
217 fprintf(stderr, " -C <file>\tcapabilities drop config\n");
218 fprintf(stderr, " -c\t\tset PR_SET_NO_NEW_PRIVS\n");
219 fprintf(stderr, " -n <name>\tthe name of the jail\n");
220 fprintf(stderr, "namespace jail options:\n");
221 fprintf(stderr, " -h <hostname>\tchange the hostname of the jail\n");
222 fprintf(stderr, " -r <file>\treadonly files that should be staged\n");
223 fprintf(stderr, " -w <file>\twriteable files that should be staged\n");
224 fprintf(stderr, " -p\t\tjail has /proc\n");
225 fprintf(stderr, " -s\t\tjail has /sys\n");
226 fprintf(stderr, " -l\t\tjail has /dev/log\n");
227 fprintf(stderr, " -u\t\tjail has a ubus socket\n");
228 fprintf(stderr, " -o\t\tremont jail root (/) read only\n");
229 fprintf(stderr, "\nWarning: by default root inside the jail is the same\n\
230 and he has the same powers as root outside the jail,\n\
231 thus he can escape the jail and/or break stuff.\n\
232 Please use seccomp/capabilities (-S/-C) to restrict his powers\n\n\
233 If you use none of the namespace jail options,\n\
234 ujail will not use namespace/build a jail,\n\
235 and will only drop capabilities/apply seccomp filter.\n\n");
238 static int exec_jail(void *_notused)
240 if (opts.capabilities && drop_capabilities(opts.capabilities))
243 if (opts.no_new_privs && prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0)) {
244 ERROR("prctl(PR_SET_NO_NEW_PRIVS) failed: %s\n", strerror(errno));
248 if (opts.namespace && opts.hostname && strlen(opts.hostname) > 0
249 && sethostname(opts.hostname, strlen(opts.hostname))) {
250 ERROR("sethostname(%s) failed: %s\n", opts.hostname, strerror(errno));
254 if (opts.namespace && build_jail_fs()) {
255 ERROR("failed to build jail fs\n");
259 char **envp = build_envp(opts.seccomp);
263 INFO("exec-ing %s\n", *opts.jail_argv);
264 execve(*opts.jail_argv, opts.jail_argv, envp);
265 /* we get there only if execve fails */
266 ERROR("failed to execve %s: %s\n", *opts.jail_argv, strerror(errno));
270 static int jail_running = 1;
271 static int jail_return_code = 0;
273 static void jail_process_timeout_cb(struct uloop_timeout *t);
274 static struct uloop_timeout jail_process_timeout = {
275 .cb = jail_process_timeout_cb,
278 static void jail_process_handler(struct uloop_process *c, int ret)
280 uloop_timeout_cancel(&jail_process_timeout);
281 if (WIFEXITED(ret)) {
282 jail_return_code = WEXITSTATUS(ret);
283 INFO("jail (%d) exited with exit: %d\n", c->pid, jail_return_code);
285 jail_return_code = WTERMSIG(ret);
286 INFO("jail (%d) exited with signal: %d\n", c->pid, jail_return_code);
292 static struct uloop_process jail_process = {
293 .cb = jail_process_handler,
296 static void jail_process_timeout_cb(struct uloop_timeout *t)
298 DEBUG("jail process failed to stop, sending SIGKILL\n");
299 kill(jail_process.pid, SIGKILL);
302 static void jail_handle_signal(int signo)
304 DEBUG("forwarding signal %d to the jailed process\n", signo);
305 kill(jail_process.pid, signo);
308 int main(int argc, char **argv)
311 uid_t uid = getuid();
312 char log[] = "/dev/log";
313 char ubus[] = "/var/run/ubus.sock";
317 ERROR("not root, aborting: %s\n", strerror(errno));
323 init_library_search();
325 while ((ch = getopt(argc, argv, OPT_ARGS)) != -1) {
328 debug = atoi(optarg);
343 opts.seccomp = optarg;
344 add_mount(optarg, 1, -1);
347 opts.capabilities = optarg;
350 opts.no_new_privs = 1;
356 opts.hostname = optarg;
360 add_path_and_deps(optarg, 1, 0, 0);
364 add_path_and_deps(optarg, 0, 0, 0);
368 add_mount(ubus, 0, -1);
372 add_mount(log, 0, -1);
377 /* no <binary> param found */
378 if (argc - optind < 1) {
382 if (!(opts.namespace||opts.capabilities||opts.seccomp)) {
383 ERROR("Not using namespaces, capabilities or seccomp !!!\n\n");
387 DEBUG("Using namespaces(%d), capabilities(%d), seccomp(%d)\n",
389 opts.capabilities != 0,
392 opts.jail_argv = &argv[optind];
394 if (opts.namespace && add_path_and_deps(*opts.jail_argv, 1, -1, 0)) {
395 ERROR("failed to load dependencies\n");
399 if (opts.namespace && opts.seccomp && add_path_and_deps("libpreload-seccomp.so", 1, -1, 1)) {
400 ERROR("failed to load libpreload-seccomp.so\n");
405 prctl(PR_SET_NAME, opts.name, NULL, NULL, NULL);
409 sigfillset(&sigmask);
410 for (i = 0; i < _NSIG; i++) {
411 struct sigaction s = { 0 };
413 if (!sigismember(&sigmask, i))
415 if ((i == SIGCHLD) || (i == SIGPIPE))
418 s.sa_handler = jail_handle_signal;
419 sigaction(i, &s, NULL);
422 if (opts.namespace) {
423 add_mount("/dev/full", 0, -1);
424 add_mount("/dev/null", 0, -1);
425 add_mount("/dev/urandom", 0, -1);
426 add_mount("/dev/zero", 0, -1);
428 int flags = CLONE_NEWPID | CLONE_NEWNS | CLONE_NEWIPC | SIGCHLD;
430 flags |= CLONE_NEWUTS;
431 jail_process.pid = clone(exec_jail, child_stack + STACK_SIZE, flags, NULL);
433 jail_process.pid = fork();
436 if (jail_process.pid > 0) {
438 uloop_process_add(&jail_process);
441 DEBUG("uloop interrupted, killing jail process\n");
442 kill(jail_process.pid, SIGTERM);
443 uloop_timeout_set(&jail_process_timeout, 1000);
447 return jail_return_code;
448 } else if (jail_process.pid == 0) {
449 /* fork child process */
450 return exec_jail(NULL);
452 ERROR("failed to clone/fork: %s\n", strerror(errno));