jail/jail.c

   1 /*
   2  * Copyright (C) 2015 John Crispin <blogic@openwrt.org>
   3  *
   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
   7  *
   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.
  12  */
  13
  14 #define _GNU_SOURCE
  15 #include <sys/mount.h>
  16 #include <sys/prctl.h>
  17 #include <sys/wait.h>
  18 #include <sys/types.h>
  19
  20 #include <stdlib.h>
  21 #include <unistd.h>
  22 #include <errno.h>
  23 #include <pwd.h>
  24 #include <grp.h>
  25 #include <string.h>
  26 #include <sys/stat.h>
  27 #include <fcntl.h>
  28 #include <libgen.h>
  29 #include <sched.h>
  30 #include <linux/limits.h>
  31 #include <signal.h>
  32
  33 #include "capabilities.h"
  34 #include "elf.h"
  35 #include "fs.h"
  36 #include "jail.h"
  37 #include "log.h"
  38
  39 #include <libubox/uloop.h>
  40 #include <libubus.h>
  41
  42 #define STACK_SIZE      (1024 * 1024)
  43 #define OPT_ARGS        "S:C:n:h:r:w:d:psulocU:G:NR:fFO:T:"
  44
  45 static struct {
  46         char *name;
  47         char *hostname;
  48         char **jail_argv;
  49         char *seccomp;
  50         char *capabilities;
  51         char *user;
  52         char *group;
  53         char *extroot;
  54         char *overlaydir;
  55         char *tmpoverlaysize;
  56         int no_new_privs;
  57         int namespace;
  58         int procfs;
  59         int ronly;
  60         int sysfs;
  61         int pw_uid;
  62         int pw_gid;
  63         int gr_gid;
  64 } opts;
  65
  66
  67 extern int pivot_root(const char *new_root, const char *put_old);
  68
  69 int debug = 0;
  70
  71 static char child_stack[STACK_SIZE];
  72
  73 static int mkdir_p(char *dir, mode_t mask)
  74 {
  75         char *l = strrchr(dir, '/');
  76         int ret;
  77
  78         if (!l)
  79                 return 0;
  80
  81         *l = '\0';
  82
  83         if (mkdir_p(dir, mask))
  84                 return -1;
  85
  86         *l = '/';
  87
  88         ret = mkdir(dir, mask);
  89         if (ret && errno == EEXIST)
  90                 return 0;
  91
  92         if (ret)
  93                 ERROR("mkdir(%s, %d) failed: %m\n", dir, mask);
  94
  95         return ret;
  96 }
  97
  98 static int _mount_bind(const char *root, const char *path, const char *target, int readonly, int strict, int error)
  99 {
 100         struct stat s;
 101         char new[PATH_MAX];
 102         int fd;
 103         int remount_flags = MS_BIND | MS_REMOUNT;
 104
 105         if (stat(path, &s)) {
 106                 ERROR("stat(%s) failed: %m\n", path);
 107                 return error;
 108         }
 109
 110         snprintf(new, sizeof(new), "%s%s", root, target?target:path);
 111
 112         if (S_ISDIR(s.st_mode)) {
 113                 mkdir_p(new, 0755);
 114         } else {
 115                 mkdir_p(dirname(new), 0755);
 116                 snprintf(new, sizeof(new), "%s%s", root, target?target:path);
 117                 fd = creat(new, 0644);
 118                 if (fd == -1) {
 119                         ERROR("creat(%s) failed: %m\n", new);
 120                         return -1;
 121                 }
 122                 close(fd);
 123         }
 124
 125         if (mount(path, new, NULL, MS_BIND, NULL)) {
 126                 ERROR("failed to mount -B %s %s: %m\n", path, new);
 127                 return -1;
 128         }
 129
 130         if (readonly)
 131                 remount_flags |= MS_RDONLY;
 132
 133         if (strict)
 134                 remount_flags |= MS_NOEXEC | MS_NOSUID | MS_NODEV;
 135
 136         if ((strict || readonly) && mount(NULL, new, NULL, remount_flags, NULL)) {
 137                 ERROR("failed to remount (%s%s%s) %s: %m\n", readonly?"ro":"rw",
 138                       (readonly && strict)?", ":"", strict?"strict":"", new);
 139                 return -1;
 140         }
 141
 142         DEBUG("mount -B %s %s (%s%s%s)\n", path, new,
 143               readonly?"ro":"rw", (readonly && strict)?", ":"", strict?"strict":"");
 144
 145         return 0;
 146 }
 147
 148 int mount_bind(const char *root, const char *path, int readonly, int error) {
 149         return _mount_bind(root, path, NULL, readonly, 0, error);
 150 }
 151
 152 static int mount_overlay(char *jail_root, char *overlaydir) {
 153         char *upperdir, *workdir, *optsstr;
 154         const char mountoptsformat[] = "lowerdir=%s,upperdir=%s,workdir=%s";
 155         int ret = -1;
 156
 157         if (asprintf(&upperdir, "%s%s", overlaydir, "/upper") < 0)
 158                 goto out;
 159
 160         if (asprintf(&workdir, "%s%s", overlaydir, "/work") < 0)
 161                 goto upper_printf;
 162
 163         if (asprintf(&optsstr, mountoptsformat, jail_root, upperdir, workdir) < 0)
 164                 goto work_printf;
 165
 166         if (mkdir_p(upperdir, 0755) || mkdir_p(workdir, 0755))
 167                 goto opts_printf;
 168
 169         DEBUG("mount -t overlay %s %s (%s)\n", jail_root, jail_root, optsstr);
 170
 171         if (mount(jail_root, jail_root, "overlay", MS_NOATIME, optsstr))
 172                 goto opts_printf;
 173
 174         ret = 0;
 175
 176 opts_printf:
 177         free(optsstr);
 178 work_printf:
 179         free(workdir);
 180 upper_printf:
 181         free(upperdir);
 182 out:
 183         return ret;
 184 }
 185
 186 static int build_jail_fs(void)
 187 {
 188         char jail_root[] = "/tmp/ujail-XXXXXX";
 189         char tmpovdir[] = "/tmp/ujail-overlay-XXXXXX";
 190         char *overlaydir = NULL;
 191
 192         if (mkdtemp(jail_root) == NULL) {
 193                 ERROR("mkdtemp(%s) failed: %m\n", jail_root);
 194                 return -1;
 195         }
 196
 197         /* oldroot can't be MS_SHARED else pivot_root() fails */
 198         if (mount("none", "/", NULL, MS_REC|MS_PRIVATE, NULL)) {
 199                 ERROR("private mount failed %m\n");
 200                 return -1;
 201         }
 202
 203         if (opts.extroot) {
 204                 if (mount(opts.extroot, jail_root, NULL, MS_BIND, NULL)) {
 205                         ERROR("extroot mount failed %m\n");
 206                         return -1;
 207                 }
 208         } else {
 209                 if (mount("tmpfs", jail_root, "tmpfs", MS_NOATIME, "mode=0755")) {
 210                         ERROR("tmpfs mount failed %m\n");
 211                         return -1;
 212                 }
 213         }
 214
 215         if (opts.tmpoverlaysize) {
 216                 char mountoptsstr[] = "mode=0755,size=XXXXXXXX";
 217
 218                 snprintf(mountoptsstr, sizeof(mountoptsstr),
 219                          "mode=0755,size=%s", opts.tmpoverlaysize);
 220                 if (mkdtemp(tmpovdir) == NULL) {
 221                         ERROR("mkdtemp(%s) failed: %m\n", jail_root);
 222                         return -1;
 223                 }
 224                 if (mount("tmpfs", tmpovdir, "tmpfs", MS_NOATIME,
 225                           mountoptsstr)) {
 226                         ERROR("failed to mount tmpfs for overlay (size=%s)\n", opts.tmpoverlaysize);
 227                         return -1;
 228                 }
 229                 overlaydir = tmpovdir;
 230         }
 231
 232         if (opts.overlaydir)
 233                 overlaydir = opts.overlaydir;
 234
 235         if (overlaydir)
 236                 mount_overlay(jail_root, overlaydir);
 237
 238         if (chdir(jail_root)) {
 239                 ERROR("chdir(%s) (jail_root) failed: %m\n", jail_root);
 240                 return -1;
 241         }
 242
 243         if (mount_all(jail_root)) {
 244                 ERROR("mount_all() failed\n");
 245                 return -1;
 246         }
 247
 248         if (opts.namespace & CLONE_NEWNET) {
 249                 char hostdir[PATH_MAX], jailetc[PATH_MAX], jaillink[PATH_MAX];
 250
 251                 snprintf(hostdir, PATH_MAX, "/tmp/resolv.conf-%s.d", opts.name);
 252                 mkdir_p(hostdir, 0755);
 253                 _mount_bind(jail_root, hostdir, "/tmp/resolv.conf.d", 1, 1, -1);
 254                 snprintf(jailetc, PATH_MAX, "%s/etc", jail_root);
 255                 mkdir_p(jailetc, 0755);
 256                 snprintf(jaillink, PATH_MAX, "%s/etc/resolv.conf", jail_root);
 257                 symlink("../tmp/resolv.conf.d/resolv.conf.auto", jaillink);
 258         }
 259
 260         char dirbuf[sizeof(jail_root) + 4];
 261         snprintf(dirbuf, sizeof(dirbuf), "%s/old", jail_root);
 262         mkdir(dirbuf, 0755);
 263
 264         if (pivot_root(jail_root, dirbuf) == -1) {
 265                 ERROR("pivot_root(%s, %s) failed: %m\n", jail_root, dirbuf);
 266                 return -1;
 267         }
 268         if (chdir("/")) {
 269                 ERROR("chdir(/) (after pivot_root) failed: %m\n");
 270                 return -1;
 271         }
 272
 273         snprintf(dirbuf, sizeof(dirbuf), "/old%s", jail_root);
 274         umount2(dirbuf, MNT_DETACH);
 275         rmdir(dirbuf);
 276         if (opts.tmpoverlaysize) {
 277                 char tmpdirbuf[sizeof(tmpovdir) + 4];
 278                 snprintf(tmpdirbuf, sizeof(tmpdirbuf), "/old%s", tmpovdir);
 279                 umount2(tmpdirbuf, MNT_DETACH);
 280                 rmdir(tmpdirbuf);
 281         }
 282
 283         umount2("/old", MNT_DETACH);
 284         rmdir("/old");
 285
 286         if (opts.procfs) {
 287                 mkdir("/proc", 0755);
 288                 mount("proc", "/proc", "proc", MS_NOATIME | MS_NODEV | MS_NOEXEC | MS_NOSUID, 0);
 289         }
 290         if (opts.sysfs) {
 291                 mkdir("/sys", 0755);
 292                 mount("sysfs", "/sys", "sysfs", MS_NOATIME | MS_NODEV | MS_NOEXEC | MS_NOSUID, 0);
 293         }
 294         if (opts.ronly)
 295                 mount(NULL, "/", NULL, MS_RDONLY | MS_REMOUNT, 0);
 296
 297         return 0;
 298 }
 299
 300 static int write_uid_gid_map(pid_t child_pid, bool gidmap, int id)
 301 {
 302         int map_file;
 303         char map_path[64];
 304         const char *map_format = "%d %d %d\n";
 305         if (snprintf(map_path, sizeof(map_path), "/proc/%d/%s",
 306                 child_pid, gidmap?"gid_map":"uid_map") < 0)
 307                 return -1;
 308
 309         if ((map_file = open(map_path, O_WRONLY)) == -1)
 310                 return -1;
 311
 312         if (dprintf(map_file, map_format, 0, id, 1) == -1) {
 313                 close(map_file);
 314                 return -1;
 315         }
 316
 317         close(map_file);
 318         return 0;
 319 }
 320
 321 static int write_setgroups(pid_t child_pid, bool allow)
 322 {
 323         int setgroups_file;
 324         char setgroups_path[64];
 325
 326         if (snprintf(setgroups_path, sizeof(setgroups_path), "/proc/%d/setgroups",
 327                 child_pid) < 0) {
 328                 return -1;
 329         }
 330
 331         if ((setgroups_file = open(setgroups_path, O_WRONLY)) == -1) {
 332                 return -1;
 333         }
 334
 335         if (dprintf(setgroups_file, allow?"allow":"deny") == -1) {
 336                 close(setgroups_file);
 337                 return -1;
 338         }
 339
 340         close(setgroups_file);
 341         return 0;
 342 }
 343
 344 static void get_jail_user(int *user, int *user_gid, int *gr_gid)
 345 {
 346         struct passwd *p = NULL;
 347         struct group *g = NULL;
 348
 349         if (opts.user) {
 350                 p = getpwnam(opts.user);
 351                 if (!p) {
 352                         ERROR("failed to get uid/gid for user %s: %d (%s)\n",
 353                               opts.user, errno, strerror(errno));
 354                         exit(EXIT_FAILURE);
 355                 }
 356                 *user = p->pw_uid;
 357                 *user_gid = p->pw_gid;
 358         } else {
 359                 *user = -1;
 360                 *user_gid = -1;
 361         }
 362
 363         if (opts.group) {
 364                 g = getgrnam(opts.group);
 365                 if (!g) {
 366                         ERROR("failed to get gid for group %s: %m\n", opts.group);
 367                         exit(EXIT_FAILURE);
 368                 }
 369                 *gr_gid = g->gr_gid;
 370         } else {
 371                 *gr_gid = -1;
 372         }
 373 };
 374
 375 static void set_jail_user(int pw_uid, int user_gid, int gr_gid)
 376 {
 377         if ((user_gid != -1) && initgroups(opts.user, user_gid)) {
 378                 ERROR("failed to initgroups() for user %s: %m\n", opts.user);
 379                 exit(EXIT_FAILURE);
 380         }
 381
 382         if ((gr_gid != -1) && setregid(gr_gid, gr_gid)) {
 383                 ERROR("failed to set group id %d: %m\n", gr_gid);
 384                 exit(EXIT_FAILURE);
 385         }
 386
 387         if ((pw_uid != -1) && setreuid(pw_uid, pw_uid)) {
 388                 ERROR("failed to set user id %d: %m\n", pw_uid);
 389                 exit(EXIT_FAILURE);
 390         }
 391 }
 392
 393 #define MAX_ENVP        8
 394 static char** build_envp(const char *seccomp)
 395 {
 396         static char *envp[MAX_ENVP];
 397         static char preload_var[PATH_MAX];
 398         static char seccomp_var[PATH_MAX];
 399         static char debug_var[] = "LD_DEBUG=all";
 400         static char container_var[] = "container=ujail";
 401         const char *preload_lib = find_lib("libpreload-seccomp.so");
 402         int count = 0;
 403
 404         if (seccomp && !preload_lib) {
 405                 ERROR("failed to add preload-lib to env\n");
 406                 return NULL;
 407         }
 408         if (seccomp) {
 409                 snprintf(seccomp_var, sizeof(seccomp_var), "SECCOMP_FILE=%s", seccomp);
 410                 envp[count++] = seccomp_var;
 411                 snprintf(preload_var, sizeof(preload_var), "LD_PRELOAD=%s", preload_lib);
 412                 envp[count++] = preload_var;
 413         }
 414
 415         envp[count++] = container_var;
 416
 417         if (debug > 1)
 418                 envp[count++] = debug_var;
 419
 420         return envp;
 421 }
 422
 423 static void usage(void)
 424 {
 425         fprintf(stderr, "ujail <options> -- <binary> <params ...>\n");
 426         fprintf(stderr, "  -d <num>\tshow debug log (increase num to increase verbosity)\n");
 427         fprintf(stderr, "  -S <file>\tseccomp filter config\n");
 428         fprintf(stderr, "  -C <file>\tcapabilities drop config\n");
 429         fprintf(stderr, "  -c\t\tset PR_SET_NO_NEW_PRIVS\n");
 430         fprintf(stderr, "  -n <name>\tthe name of the jail\n");
 431         fprintf(stderr, "namespace jail options:\n");
 432         fprintf(stderr, "  -h <hostname>\tchange the hostname of the jail\n");
 433         fprintf(stderr, "  -N\t\tjail has network namespace\n");
 434         fprintf(stderr, "  -f\t\tjail has user namespace\n");
 435         fprintf(stderr, "  -F\t\tjail has cgroups namespace\n");
 436         fprintf(stderr, "  -r <file>\treadonly files that should be staged\n");
 437         fprintf(stderr, "  -w <file>\twriteable files that should be staged\n");
 438         fprintf(stderr, "  -p\t\tjail has /proc\n");
 439         fprintf(stderr, "  -s\t\tjail has /sys\n");
 440         fprintf(stderr, "  -l\t\tjail has /dev/log\n");
 441         fprintf(stderr, "  -u\t\tjail has a ubus socket\n");
 442         fprintf(stderr, "  -U <name>\tuser to run jailed process\n");
 443         fprintf(stderr, "  -G <name>\tgroup to run jailed process\n");
 444         fprintf(stderr, "  -o\t\tremont jail root (/) read only\n");
 445         fprintf(stderr, "  -R <dir>\texternal jail rootfs (system container)\n");
 446         fprintf(stderr, "  -O <dir>\tdirectory for r/w overlayfs\n");
 447         fprintf(stderr, "  -T <size>\tuse tmpfs r/w overlayfs with <size>\n");
 448         fprintf(stderr, "\nWarning: by default root inside the jail is the same\n\
 449 and he has the same powers as root outside the jail,\n\
 450 thus he can escape the jail and/or break stuff.\n\
 451 Please use seccomp/capabilities (-S/-C) to restrict his powers\n\n\
 452 If you use none of the namespace jail options,\n\
 453 ujail will not use namespace/build a jail,\n\
 454 and will only drop capabilities/apply seccomp filter.\n\n");
 455 }
 456
 457 static int exec_jail(void *pipes_ptr)
 458 {
 459         int *pipes = (int*)pipes_ptr;
 460         char buf[1];
 461         int pw_uid, pw_gid, gr_gid;
 462
 463         close(pipes[0]);
 464         close(pipes[3]);
 465
 466
 467         buf[0] = 'i';
 468         if (write(pipes[1], buf, 1) < 1) {
 469                 ERROR("can't write to parent\n");
 470                 exit(EXIT_FAILURE);
 471         }
 472         if (read(pipes[2], buf, 1) < 1) {
 473                 ERROR("can't read from parent\n");
 474                 exit(EXIT_FAILURE);
 475         }
 476         if (buf[0] != 'O') {
 477                 ERROR("parent had an error, child exiting\n");
 478                 exit(EXIT_FAILURE);
 479         }
 480
 481         close(pipes[1]);
 482         close(pipes[2]);
 483
 484         if (opts.namespace & CLONE_NEWUSER) {
 485                 if (setgid(0) < 0) {
 486                         ERROR("setgid\n");
 487                         exit(EXIT_FAILURE);
 488                 }
 489                 if (setuid(0) < 0) {
 490                         ERROR("setuid\n");
 491                         exit(EXIT_FAILURE);
 492                 }
 493 //              if (setgroups(0, NULL) < 0) {
 494 //                      ERROR("setgroups\n");
 495 //                      exit(EXIT_FAILURE);
 496 //              }
 497         }
 498
 499         if (opts.namespace && opts.hostname && strlen(opts.hostname) > 0
 500                         && sethostname(opts.hostname, strlen(opts.hostname))) {
 501                 ERROR("sethostname(%s) failed: %m\n", opts.hostname);
 502                 exit(EXIT_FAILURE);
 503         }
 504
 505         if (opts.namespace && build_jail_fs()) {
 506                 ERROR("failed to build jail fs\n");
 507                 exit(EXIT_FAILURE);
 508         }
 509
 510         if (opts.capabilities && drop_capabilities(opts.capabilities))
 511                 exit(EXIT_FAILURE);
 512
 513         if (opts.no_new_privs && prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0)) {
 514                 ERROR("prctl(PR_SET_NO_NEW_PRIVS) failed: %m\n");
 515                 exit(EXIT_FAILURE);
 516         }
 517
 518         if (!(opts.namespace & CLONE_NEWUSER)) {
 519                 get_jail_user(&pw_uid, &pw_gid, &gr_gid);
 520                 set_jail_user(pw_uid, pw_gid, gr_gid);
 521         }
 522
 523         char **envp = build_envp(opts.seccomp);
 524         if (!envp)
 525                 exit(EXIT_FAILURE);
 526
 527         INFO("exec-ing %s\n", *opts.jail_argv);
 528         execve(*opts.jail_argv, opts.jail_argv, envp);
 529         /* we get there only if execve fails */
 530         ERROR("failed to execve %s: %m\n", *opts.jail_argv);
 531         exit(EXIT_FAILURE);
 532 }
 533
 534 static int jail_running = 1;
 535 static int jail_return_code = 0;
 536
 537 static void jail_process_timeout_cb(struct uloop_timeout *t);
 538 static struct uloop_timeout jail_process_timeout = {
 539         .cb = jail_process_timeout_cb,
 540 };
 541
 542 static void jail_process_handler(struct uloop_process *c, int ret)
 543 {
 544         uloop_timeout_cancel(&jail_process_timeout);
 545         if (WIFEXITED(ret)) {
 546                 jail_return_code = WEXITSTATUS(ret);
 547                 INFO("jail (%d) exited with exit: %d\n", c->pid, jail_return_code);
 548         } else {
 549                 jail_return_code = WTERMSIG(ret);
 550                 INFO("jail (%d) exited with signal: %d\n", c->pid, jail_return_code);
 551         }
 552         jail_running = 0;
 553         uloop_end();
 554 }
 555
 556 static struct uloop_process jail_process = {
 557         .cb = jail_process_handler,
 558 };
 559
 560 static void jail_process_timeout_cb(struct uloop_timeout *t)
 561 {
 562         DEBUG("jail process failed to stop, sending SIGKILL\n");
 563         kill(jail_process.pid, SIGKILL);
 564 }
 565
 566 static void jail_handle_signal(int signo)
 567 {
 568         DEBUG("forwarding signal %d to the jailed process\n", signo);
 569         kill(jail_process.pid, signo);
 570 }
 571
 572 static int netns_open_pid(const pid_t target_ns)
 573 {
 574         char pid_net_path[PATH_MAX];
 575
 576         snprintf(pid_net_path, sizeof(pid_net_path), "/proc/%u/ns/net", target_ns);
 577
 578         return open(pid_net_path, O_RDONLY);
 579 }
 580
 581 static void netns_updown(pid_t pid, bool start)
 582 {
 583         struct ubus_context *ctx = ubus_connect(NULL);
 584         static struct blob_buf req;
 585         uint32_t id;
 586
 587         if (!ctx)
 588                 return;
 589
 590         blob_buf_init(&req, 0);
 591         blobmsg_add_string(&req, "jail", opts.name);
 592         blobmsg_add_u32(&req, "pid", pid);
 593         blobmsg_add_u8(&req, "start", start);
 594
 595         if (ubus_lookup_id(ctx, "network", &id) ||
 596             ubus_invoke(ctx, id, "netns_updown", req.head, NULL, NULL, 3000))
 597                 INFO("ubus request failed\n");
 598
 599         blob_buf_free(&req);
 600         ubus_free(ctx);
 601 }
 602
 603 int main(int argc, char **argv)
 604 {
 605         sigset_t sigmask;
 606         uid_t uid = getuid();
 607         char log[] = "/dev/log";
 608         char ubus[] = "/var/run/ubus.sock";
 609         int ch, i;
 610         int pipes[4];
 611         char sig_buf[1];
 612         int netns_fd;
 613
 614         if (uid) {
 615                 ERROR("not root, aborting: %m\n");
 616                 return EXIT_FAILURE;
 617         }
 618
 619         umask(022);
 620         mount_list_init();
 621         init_library_search();
 622
 623         while ((ch = getopt(argc, argv, OPT_ARGS)) != -1) {
 624                 switch (ch) {
 625                 case 'd':
 626                         debug = atoi(optarg);
 627                         break;
 628                 case 'p':
 629                         opts.namespace |= CLONE_NEWNS;
 630                         opts.procfs = 1;
 631                         break;
 632                 case 'o':
 633                         opts.namespace |= CLONE_NEWNS;
 634                         opts.ronly = 1;
 635                         break;
 636                 case 'f':
 637                         opts.namespace |= CLONE_NEWUSER;
 638                         break;
 639                 case 'F':
 640                         opts.namespace |= CLONE_NEWCGROUP;
 641                         break;
 642                 case 'R':
 643                         opts.extroot = optarg;
 644                         break;
 645                 case 's':
 646                         opts.namespace |= CLONE_NEWNS;
 647                         opts.sysfs = 1;
 648                         break;
 649                 case 'S':
 650                         opts.seccomp = optarg;
 651                         add_mount(optarg, 1, -1);
 652                         break;
 653                 case 'C':
 654                         opts.capabilities = optarg;
 655                         break;
 656                 case 'c':
 657                         opts.no_new_privs = 1;
 658                         break;
 659                 case 'n':
 660                         opts.name = optarg;
 661                         break;
 662                 case 'N':
 663                         opts.namespace |= CLONE_NEWNET;
 664                         break;
 665                 case 'h':
 666                         opts.namespace |= CLONE_NEWUTS;
 667                         opts.hostname = optarg;
 668                         break;
 669                 case 'r':
 670                         opts.namespace |= CLONE_NEWNS;
 671                         add_path_and_deps(optarg, 1, 0, 0);
 672                         break;
 673                 case 'w':
 674                         opts.namespace |= CLONE_NEWNS;
 675                         add_path_and_deps(optarg, 0, 0, 0);
 676                         break;
 677                 case 'u':
 678                         opts.namespace |= CLONE_NEWNS;
 679                         add_mount(ubus, 0, -1);
 680                         break;
 681                 case 'l':
 682                         opts.namespace |= CLONE_NEWNS;
 683                         add_mount(log, 0, -1);
 684                         break;
 685                 case 'U':
 686                         opts.user = optarg;
 687                         break;
 688                 case 'G':
 689                         opts.group = optarg;
 690                         break;
 691                 case 'O':
 692                         opts.overlaydir = optarg;
 693                         break;
 694                 case 'T':
 695                         opts.tmpoverlaysize = optarg;
 696                         break;
 697                 }
 698         }
 699
 700         if (opts.namespace)
 701                 opts.namespace |= CLONE_NEWIPC | CLONE_NEWPID;
 702
 703         if (opts.tmpoverlaysize && strlen(opts.tmpoverlaysize) > 8) {
 704                 ERROR("size parameter too long: \"%s\"\n", opts.tmpoverlaysize);
 705                 return -1;
 706         }
 707
 708         /* no <binary> param found */
 709         if (argc - optind < 1) {
 710                 usage();
 711                 return EXIT_FAILURE;
 712         }
 713         if (!(opts.namespace||opts.capabilities||opts.seccomp)) {
 714                 ERROR("Not using namespaces, capabilities or seccomp !!!\n\n");
 715                 usage();
 716                 return EXIT_FAILURE;
 717         }
 718         DEBUG("Using namespaces(0x%08x), capabilities(%d), seccomp(%d)\n",
 719                 opts.namespace,
 720                 opts.capabilities != 0,
 721                 opts.seccomp != 0);
 722
 723         opts.jail_argv = &argv[optind];
 724
 725         get_jail_user(&opts.pw_uid, &opts.pw_gid, &opts.gr_gid);
 726
 727         if (!opts.extroot) {
 728                 if (opts.namespace && add_path_and_deps(*opts.jail_argv, 1, -1, 0)) {
 729                         ERROR("failed to load dependencies\n");
 730                         return -1;
 731                 }
 732         }
 733
 734         if (opts.namespace && opts.seccomp && add_path_and_deps("libpreload-seccomp.so", 1, -1, 1)) {
 735                 ERROR("failed to load libpreload-seccomp.so\n");
 736                 return -1;
 737         }
 738
 739         if (opts.name)
 740                 prctl(PR_SET_NAME, opts.name, NULL, NULL, NULL);
 741
 742         uloop_init();
 743
 744         sigfillset(&sigmask);
 745         for (i = 0; i < _NSIG; i++) {
 746                 struct sigaction s = { 0 };
 747
 748                 if (!sigismember(&sigmask, i))
 749                         continue;
 750                 if ((i == SIGCHLD) || (i == SIGPIPE))
 751                         continue;
 752
 753                 s.sa_handler = jail_handle_signal;
 754                 sigaction(i, &s, NULL);
 755         }
 756
 757         if (opts.namespace) {
 758                 if (opts.namespace & CLONE_NEWNS) {
 759                         add_mount("/dev/full", 0, -1);
 760                         add_mount("/dev/null", 0, -1);
 761                         add_mount("/dev/random", 0, -1);
 762                         add_mount("/dev/urandom", 0, -1);
 763                         add_mount("/dev/tty", 0, -1);
 764                         add_mount("/dev/zero", 0, -1);
 765                         add_mount("/dev/console", 0, -1);
 766
 767                         if (!opts.extroot && (opts.user || opts.group)) {
 768                                 add_mount("/etc/passwd", 0, -1);
 769                                 add_mount("/etc/group", 0, -1);
 770                         }
 771
 772                         if (!(opts.namespace & CLONE_NEWNET)) {
 773                                 add_mount("/etc/resolv.conf", 0, -1);
 774                         }
 775                 }
 776
 777                 if (pipe(&pipes[0]) < 0 || pipe(&pipes[2]) < 0)
 778                         return -1;
 779
 780                 jail_process.pid = clone(exec_jail, child_stack + STACK_SIZE, SIGCHLD | opts.namespace, &pipes);
 781         } else {
 782                 jail_process.pid = fork();
 783         }
 784
 785         if (jail_process.pid > 0) {
 786                 seteuid(0);
 787                 /* parent process */
 788                 close(pipes[1]);
 789                 close(pipes[2]);
 790                 if (read(pipes[0], sig_buf, 1) < 1) {
 791                         ERROR("can't read from child\n");
 792                         return -1;
 793                 }
 794                 close(pipes[0]);
 795                 if (opts.namespace & CLONE_NEWUSER) {
 796                         bool has_gr = (opts.gr_gid != -1);
 797                         if (write_setgroups(jail_process.pid, false)) {
 798                                 ERROR("can't write setgroups\n");
 799                                 return -1;
 800                         }
 801                         if (opts.pw_uid != -1) {
 802                                 write_uid_gid_map(jail_process.pid, 0, opts.pw_uid);
 803                                 write_uid_gid_map(jail_process.pid, 1, has_gr?opts.gr_gid:opts.pw_gid);
 804                         } else {
 805                                 write_uid_gid_map(jail_process.pid, 0, 65534);
 806                                 write_uid_gid_map(jail_process.pid, 1, has_gr?opts.gr_gid:65534);
 807                         }
 808                 }
 809
 810                 if (opts.namespace & CLONE_NEWNET) {
 811                         netns_fd = netns_open_pid(jail_process.pid);
 812                         netns_updown(jail_process.pid, true);
 813                 }
 814
 815                 sig_buf[0] = 'O';
 816                 if (write(pipes[3], sig_buf, 1) < 0) {
 817                         ERROR("can't write to child\n");
 818                         return -1;
 819                 }
 820                 close(pipes[3]);
 821                 uloop_process_add(&jail_process);
 822                 uloop_run();
 823                 if (jail_running) {
 824                         DEBUG("uloop interrupted, killing jail process\n");
 825                         kill(jail_process.pid, SIGTERM);
 826                         uloop_timeout_set(&jail_process_timeout, 1000);
 827                         uloop_run();
 828                 }
 829                 uloop_done();
 830                 if (opts.namespace & CLONE_NEWNET) {
 831                         setns(netns_fd, CLONE_NEWNET);
 832                         netns_updown(getpid(), false);
 833                         close(netns_fd);
 834                 }
 835                 return jail_return_code;
 836         } else if (jail_process.pid == 0) {
 837                 /* fork child process */
 838                 return exec_jail(NULL);
 839         } else {
 840                 ERROR("failed to clone/fork: %m\n");
 841                 return EXIT_FAILURE;
 842         }
 843 }