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
  19 #include <stdlib.h>
  20 #include <unistd.h>
  21 #include <errno.h>
  22 #include <string.h>
  23 #include <sys/stat.h>
  24 #include <fcntl.h>
  25 #include <libgen.h>
  26 #include <sched.h>
  27 #include <linux/limits.h>
  28
  29 #include "capabilities.h"
  30 #include "elf.h"
  31 #include "fs.h"
  32 #include "jail.h"
  33 #include "log.h"
  34
  35 #include <libubox/uloop.h>
  36
  37 #define STACK_SIZE      (1024 * 1024)
  38 #define OPT_ARGS        "S:C:n:r:w:d:psuloc"
  39
  40 static struct {
  41         char *name;
  42         char **jail_argv;
  43         char *seccomp;
  44         char *capabilities;
  45         int no_new_privs;
  46         int namespace;
  47         int procfs;
  48         int ronly;
  49         int sysfs;
  50 } opts;
  51
  52 extern int pivot_root(const char *new_root, const char *put_old);
  53
  54 int debug = 0;
  55
  56 static char child_stack[STACK_SIZE];
  57
  58 static int mkdir_p(char *dir, mode_t mask)
  59 {
  60         char *l = strrchr(dir, '/');
  61         int ret;
  62
  63         if (!l)
  64                 return 0;
  65
  66         *l = '\0';
  67
  68         if (mkdir_p(dir, mask))
  69                 return -1;
  70
  71         *l = '/';
  72
  73         ret = mkdir(dir, mask);
  74         if (ret && errno == EEXIST)
  75                 return 0;
  76
  77         if (ret)
  78                 ERROR("mkdir failed on %s: %s\n", dir, strerror(errno));
  79
  80         return ret;
  81 }
  82
  83 int mount_bind(const char *root, const char *path, int readonly, int error)
  84 {
  85         struct stat s;
  86         char new[PATH_MAX];
  87         int fd;
  88
  89         if (stat(path, &s)) {
  90                 ERROR("stat(%s) failed: %s\n", path, strerror(errno));
  91                 return error;
  92         }
  93
  94         snprintf(new, sizeof(new), "%s%s", root, path);
  95         if (S_ISDIR(s.st_mode)) {
  96                 mkdir_p(new, 0755);
  97         } else {
  98                 mkdir_p(dirname(new), 0755);
  99                 snprintf(new, sizeof(new), "%s%s", root, path);
 100                 fd = creat(new, 0644);
 101                 if (fd == -1) {
 102                         ERROR("failed to create %s: %s\n", new, strerror(errno));
 103                         return -1;
 104                 }
 105                 close(fd);
 106         }
 107
 108         if (mount(path, new, NULL, MS_BIND, NULL)) {
 109                 ERROR("failed to mount -B %s %s: %s\n", path, new, strerror(errno));
 110                 return -1;
 111         }
 112
 113         if (readonly && mount(NULL, new, NULL, MS_BIND | MS_REMOUNT | MS_RDONLY, NULL)) {
 114                 ERROR("failed to remount ro %s: %s\n", new, strerror(errno));
 115                 return -1;
 116         }
 117
 118         DEBUG("mount -B %s %s\n", path, new);
 119
 120         return 0;
 121 }
 122
 123 static int build_jail_fs(void)
 124 {
 125         char jail_root[] = "/tmp/ujail-XXXXXX";
 126         if (mkdtemp(jail_root) == NULL) {
 127                 ERROR("mkdtemp(jail_root) failed: %s\n", strerror(errno));
 128                 return -1;
 129         }
 130
 131         if (mount("tmpfs", jail_root, "tmpfs", MS_NOATIME, "mode=0755")) {
 132                 ERROR("tmpfs mount failed %s\n", strerror(errno));
 133                 return -1;
 134         }
 135
 136         if (chdir(jail_root)) {
 137                 ERROR("failed to chdir() in the jail root\n");
 138                 return -1;
 139         }
 140
 141         if (add_path_and_deps(*opts.jail_argv, 1, -1, 0)) {
 142                 ERROR("failed to load dependencies\n");
 143                 return -1;
 144         }
 145
 146         if (opts.seccomp && add_path_and_deps("libpreload-seccomp.so", 1, -1, 1)) {
 147                 ERROR("failed to load libpreload-seccomp.so\n");
 148                 return -1;
 149         }
 150
 151         if (mount_all(jail_root)) {
 152                 ERROR("mount_all() failed\n");
 153                 return -1;
 154         }
 155
 156         char dirbuf[sizeof(jail_root) + 4];
 157         snprintf(dirbuf, sizeof(dirbuf), "%s/old", jail_root);
 158         mkdir(dirbuf, 0755);
 159
 160         if (pivot_root(jail_root, dirbuf) == -1) {
 161                 ERROR("pivot_root failed: %s\n", strerror(errno));
 162                 return -1;
 163         }
 164
 165         snprintf(dirbuf, sizeof(dirbuf), "/old%s", jail_root);
 166         rmdir(dirbuf);
 167         umount2("/old", MNT_DETACH);
 168         rmdir("/old");
 169
 170         if (opts.procfs) {
 171                 mkdir("/proc", 0755);
 172                 mount("proc", "/proc", "proc", MS_NOATIME | MS_NODEV | MS_NOEXEC | MS_NOSUID, 0);
 173         }
 174         if (opts.sysfs) {
 175                 mkdir("/sys", 0755);
 176                 mount("sysfs", "/sys", "sysfs", MS_NOATIME | MS_NODEV | MS_NOEXEC | MS_NOSUID, 0);
 177         }
 178         if (opts.ronly)
 179                 mount(NULL, "/", NULL, MS_RDONLY | MS_REMOUNT, 0);
 180
 181         return 0;
 182 }
 183
 184 #define MAX_ENVP        8
 185 static char** build_envp(const char *seccomp)
 186 {
 187         static char *envp[MAX_ENVP];
 188         static char preload_var[PATH_MAX];
 189         static char seccomp_var[PATH_MAX];
 190         static char debug_var[] = "LD_DEBUG=all";
 191         const char *preload_lib = find_lib("libpreload-seccomp.so");
 192         int count = 0;
 193
 194         if (seccomp && !preload_lib) {
 195                 ERROR("failed to add preload-lib to env\n");
 196                 return NULL;
 197         }
 198         if (seccomp) {
 199                 snprintf(seccomp_var, sizeof(seccomp_var), "SECCOMP_FILE=%s", seccomp);
 200                 envp[count++] = seccomp_var;
 201                 snprintf(preload_var, sizeof(preload_var), "LD_PRELOAD=%s", preload_lib);
 202                 envp[count++] = preload_var;
 203         }
 204         if (debug > 1)
 205                 envp[count++] = debug_var;
 206
 207         return envp;
 208 }
 209
 210 static void usage(void)
 211 {
 212         fprintf(stderr, "ujail <options> -- <binary> <params ...>\n");
 213         fprintf(stderr, "  -d <num>\tshow debug log (increase num to increase verbosity)\n");
 214         fprintf(stderr, "  -S <file>\tseccomp filter config\n");
 215         fprintf(stderr, "  -C <file>\tcapabilities drop config\n");
 216         fprintf(stderr, "  -c\t\tset PR_SET_NO_NEW_PRIVS\n");
 217         fprintf(stderr, "  -n <name>\tthe name of the jail\n");
 218         fprintf(stderr, "namespace jail options:\n");
 219         fprintf(stderr, "  -r <file>\treadonly files that should be staged\n");
 220         fprintf(stderr, "  -w <file>\twriteable files that should be staged\n");
 221         fprintf(stderr, "  -p\t\tjail has /proc\n");
 222         fprintf(stderr, "  -s\t\tjail has /sys\n");
 223         fprintf(stderr, "  -l\t\tjail has /dev/log\n");
 224         fprintf(stderr, "  -u\t\tjail has a ubus socket\n");
 225         fprintf(stderr, "  -o\t\tremont jail root (/) read only\n");
 226         fprintf(stderr, "\nWarning: by default root inside the jail is the same\n\
 227 and he has the same powers as root outside the jail,\n\
 228 thus he can escape the jail and/or break stuff.\n\
 229 Please use seccomp/capabilities (-S/-C) to restrict his powers\n\n\
 230 If you use none of the namespace jail options,\n\
 231 ujail will not use namespace/build a jail,\n\
 232 and will only drop capabilities/apply seccomp filter.\n\n");
 233 }
 234
 235 static int exec_jail(void)
 236 {
 237         char **envp = build_envp(opts.seccomp);
 238         if (!envp)
 239                 exit(EXIT_FAILURE);
 240
 241         if (opts.capabilities && drop_capabilities(opts.capabilities))
 242                 exit(EXIT_FAILURE);
 243
 244         if (opts.no_new_privs && prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0)) {
 245                 ERROR("prctl(PR_SET_NO_NEW_PRIVS) failed: %s\n", strerror(errno));
 246                 exit(EXIT_FAILURE);
 247         }
 248
 249         INFO("exec-ing %s\n", *opts.jail_argv);
 250         execve(*opts.jail_argv, opts.jail_argv, envp);
 251         /* we get there only if execve fails */
 252         ERROR("failed to execve %s: %s\n", *opts.jail_argv, strerror(errno));
 253         exit(EXIT_FAILURE);
 254 }
 255
 256 static int spawn_jail(void *_notused)
 257 {
 258         if (opts.name && sethostname(opts.name, strlen(opts.name))) {
 259                 ERROR("failed to sethostname: %s\n", strerror(errno));
 260         }
 261
 262         if (build_jail_fs()) {
 263                 ERROR("failed to build jail fs");
 264                 exit(EXIT_FAILURE);
 265         }
 266
 267         return exec_jail();
 268 }
 269
 270 static int jail_running = 1;
 271 static int jail_return_code = 0;
 272
 273 static void jail_process_handler(struct uloop_process *c, int ret)
 274 {
 275         if (WIFEXITED(ret)) {
 276                 jail_return_code = WEXITSTATUS(ret);
 277                 INFO("jail (%d) exited with exit: %d\n", c->pid, jail_return_code);
 278         } else {
 279                 jail_return_code = WTERMSIG(ret);
 280                 INFO("jail (%d) exited with signal: %d\n", c->pid, jail_return_code);
 281         }
 282         jail_running = 0;
 283         uloop_end();
 284 }
 285
 286 static struct uloop_process jail_process = {
 287         .cb = jail_process_handler,
 288 };
 289
 290 int main(int argc, char **argv)
 291 {
 292         uid_t uid = getuid();
 293         char log[] = "/dev/log";
 294         char ubus[] = "/var/run/ubus.sock";
 295         int ch;
 296
 297         if (uid) {
 298                 ERROR("not root, aborting: %s\n", strerror(errno));
 299                 return EXIT_FAILURE;
 300         }
 301
 302         umask(022);
 303         mount_list_init();
 304         init_library_search();
 305
 306         while ((ch = getopt(argc, argv, OPT_ARGS)) != -1) {
 307                 switch (ch) {
 308                 case 'd':
 309                         debug = atoi(optarg);
 310                         break;
 311                 case 'p':
 312                         opts.namespace = 1;
 313                         opts.procfs = 1;
 314                         break;
 315                 case 'o':
 316                         opts.namespace = 1;
 317                         opts.ronly = 1;
 318                         break;
 319                 case 's':
 320                         opts.namespace = 1;
 321                         opts.sysfs = 1;
 322                         break;
 323                 case 'S':
 324                         opts.seccomp = optarg;
 325                         add_mount(optarg, 1, -1);
 326                         break;
 327                 case 'C':
 328                         opts.capabilities = optarg;
 329                         add_mount(optarg, 1, -1);
 330                         break;
 331                 case 'c':
 332                         opts.no_new_privs = 1;
 333                         break;
 334                 case 'n':
 335                         opts.name = optarg;
 336                         break;
 337                 case 'r':
 338                         opts.namespace = 1;
 339                         add_path_and_deps(optarg, 1, 0, 0);
 340                         break;
 341                 case 'w':
 342                         opts.namespace = 1;
 343                         add_path_and_deps(optarg, 0, 0, 0);
 344                         break;
 345                 case 'u':
 346                         opts.namespace = 1;
 347                         add_mount(ubus, 0, -1);
 348                         break;
 349                 case 'l':
 350                         opts.namespace = 1;
 351                         add_mount(log, 0, -1);
 352                         break;
 353                 }
 354         }
 355
 356         /* no <binary> param found */
 357         if (argc - optind < 1) {
 358                 usage();
 359                 return EXIT_FAILURE;
 360         }
 361         if (!(opts.namespace||opts.capabilities||opts.seccomp)) {
 362                 ERROR("Not using namespaces, capabilities or seccomp !!!\n\n");
 363                 usage();
 364                 return EXIT_FAILURE;
 365         }
 366         DEBUG("Using namespaces(%d), capabilities(%d), seccomp(%d)\n",
 367                 opts.namespace,
 368                 opts.capabilities != 0,
 369                 opts.seccomp != 0);
 370
 371         opts.jail_argv = &argv[optind];
 372
 373         if (opts.name)
 374                 prctl(PR_SET_NAME, opts.name, NULL, NULL, NULL);
 375
 376         uloop_init();
 377         if (opts.namespace) {
 378                 jail_process.pid = clone(spawn_jail,
 379                         child_stack + STACK_SIZE,
 380                         CLONE_NEWUTS | CLONE_NEWPID | CLONE_NEWNS | CLONE_NEWIPC | SIGCHLD, NULL);
 381         } else {
 382                 jail_process.pid = fork();
 383         }
 384
 385         if (jail_process.pid > 0) {
 386                 /* parent process */
 387                 uloop_process_add(&jail_process);
 388                 uloop_run();
 389                 uloop_done();
 390                 if (jail_running) {
 391                         DEBUG("uloop interrupted, killing jail process\n");
 392                         kill(jail_process.pid, SIGTERM);
 393                         waitpid(jail_process.pid, NULL, 0);
 394                 }
 395                 return jail_return_code;
 396         } else if (jail_process.pid == 0) {
 397                 /* fork child process */
 398                 return exec_jail();
 399         } else {
 400                 ERROR("failed to clone/fork: %s\n", strerror(errno));
 401                 return EXIT_FAILURE;
 402         }
 403 }