sys-utils/nsenter.c

   1 /*
   2  * nsenter(1) - command-line interface for setns(2)
   3  *
   4  * Copyright (C) 2012-2013 Eric Biederman <ebiederm@xmission.com>
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the
   8  * Free Software Foundation; version 2.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  18  */
  19
  20 #include <dirent.h>
  21 #include <errno.h>
  22 #include <getopt.h>
  23 #include <sched.h>
  24 #include <stdio.h>
  25 #include <stdlib.h>
  26 #include <stdbool.h>
  27 #include <unistd.h>
  28 #include <assert.h>
  29 #include <sys/types.h>
  30 #include <sys/wait.h>
  31 #include <grp.h>
  32 #include <sys/stat.h>
  33
  34 #ifdef HAVE_LIBSELINUX
  35 # include <selinux/selinux.h>
  36 #endif
  37
  38 #include "strutils.h"
  39 #include "nls.h"
  40 #include "c.h"
  41 #include "closestream.h"
  42 #include "namespace.h"
  43 #include "exec_shell.h"
  44
  45 static struct namespace_file {
  46         int nstype;
  47         const char *name;
  48         int fd;
  49 } namespace_files[] = {
  50         /* Careful the order is significant in this array.
  51          *
  52          * The user namespace comes either first or last: first if
  53          * you're using it to increase your privilege and last if
  54          * you're using it to decrease.  We enter the namespaces in
  55          * two passes starting initially from offset 1 and then offset
  56          * 0 if that fails.
  57          */
  58         { .nstype = CLONE_NEWUSER,  .name = "ns/user", .fd = -1 },
  59         { .nstype = CLONE_NEWCGROUP,.name = "ns/cgroup", .fd = -1 },
  60         { .nstype = CLONE_NEWIPC,   .name = "ns/ipc",  .fd = -1 },
  61         { .nstype = CLONE_NEWUTS,   .name = "ns/uts",  .fd = -1 },
  62         { .nstype = CLONE_NEWNET,   .name = "ns/net",  .fd = -1 },
  63         { .nstype = CLONE_NEWPID,   .name = "ns/pid",  .fd = -1 },
  64         { .nstype = CLONE_NEWNS,    .name = "ns/mnt",  .fd = -1 },
  65         { .nstype = 0, .name = NULL, .fd = -1 }
  66 };
  67
  68 static void __attribute__((__noreturn__)) usage(void)
  69 {
  70         FILE *out = stdout;
  71
  72         fputs(USAGE_HEADER, out);
  73         fprintf(out, _(" %s [options] [<program> [<argument>...]]\n"),
  74                 program_invocation_short_name);
  75
  76         fputs(USAGE_SEPARATOR, out);
  77         fputs(_("Run a program with namespaces of other processes.\n"), out);
  78
  79         fputs(USAGE_OPTIONS, out);
  80         fputs(_(" -a, --all              enter all namespaces\n"), out);
  81         fputs(_(" -t, --target <pid>     target process to get namespaces from\n"), out);
  82         fputs(_(" -m, --mount[=<file>]   enter mount namespace\n"), out);
  83         fputs(_(" -u, --uts[=<file>]     enter UTS namespace (hostname etc)\n"), out);
  84         fputs(_(" -i, --ipc[=<file>]     enter System V IPC namespace\n"), out);
  85         fputs(_(" -n, --net[=<file>]     enter network namespace\n"), out);
  86         fputs(_(" -p, --pid[=<file>]     enter pid namespace\n"), out);
  87         fputs(_(" -C, --cgroup[=<file>]  enter cgroup namespace\n"), out);
  88         fputs(_(" -U, --user[=<file>]    enter user namespace\n"), out);
  89         fputs(_(" -S, --setuid <uid>     set uid in entered namespace\n"), out);
  90         fputs(_(" -G, --setgid <gid>     set gid in entered namespace\n"), out);
  91         fputs(_("     --preserve-credentials do not touch uids or gids\n"), out);
  92         fputs(_(" -r, --root[=<dir>]     set the root directory\n"), out);
  93         fputs(_(" -w, --wd[=<dir>]       set the working directory\n"), out);
  94         fputs(_(" -F, --no-fork          do not fork before exec'ing <program>\n"), out);
  95 #ifdef HAVE_LIBSELINUX
  96         fputs(_(" -Z, --follow-context   set SELinux context according to --target PID\n"), out);
  97 #endif
  98
  99         fputs(USAGE_SEPARATOR, out);
 100         printf(USAGE_HELP_OPTIONS(24));
 101         printf(USAGE_MAN_TAIL("nsenter(1)"));
 102
 103         exit(EXIT_SUCCESS);
 104 }
 105
 106 static pid_t namespace_target_pid = 0;
 107 static int root_fd = -1;
 108 static int wd_fd = -1;
 109
 110 static void open_target_fd(int *fd, const char *type, const char *path)
 111 {
 112         char pathbuf[PATH_MAX];
 113
 114         if (!path && namespace_target_pid) {
 115                 snprintf(pathbuf, sizeof(pathbuf), "/proc/%u/%s",
 116                          namespace_target_pid, type);
 117                 path = pathbuf;
 118         }
 119         if (!path)
 120                 errx(EXIT_FAILURE,
 121                      _("neither filename nor target pid supplied for %s"),
 122                      type);
 123
 124         if (*fd >= 0)
 125                 close(*fd);
 126
 127         *fd = open(path, O_RDONLY);
 128         if (*fd < 0)
 129                 err(EXIT_FAILURE, _("cannot open %s"), path);
 130 }
 131
 132 static void open_namespace_fd(int nstype, const char *path)
 133 {
 134         struct namespace_file *nsfile;
 135
 136         for (nsfile = namespace_files; nsfile->nstype; nsfile++) {
 137                 if (nstype != nsfile->nstype)
 138                         continue;
 139
 140                 open_target_fd(&nsfile->fd, nsfile->name, path);
 141                 return;
 142         }
 143         /* This should never happen */
 144         assert(nsfile->nstype);
 145 }
 146
 147 static int get_ns_ino(const char *path, ino_t *ino)
 148 {
 149         struct stat st;
 150
 151         if (stat(path, &st) != 0)
 152                 return -errno;
 153         *ino = st.st_ino;
 154         return 0;
 155 }
 156
 157 static int is_same_namespace(pid_t a, pid_t b, const char *type)
 158 {
 159         char path[PATH_MAX];
 160         ino_t a_ino = 0, b_ino = 0;
 161
 162         snprintf(path, sizeof(path), "/proc/%u/%s", a, type);
 163         if (get_ns_ino(path, &a_ino) != 0)
 164                 err(EXIT_FAILURE, _("stat of %s failed"), path);
 165
 166         snprintf(path, sizeof(path), "/proc/%u/%s", b, type);
 167         if (get_ns_ino(path, &b_ino) != 0)
 168                 err(EXIT_FAILURE, _("stat of %s failed"), path);
 169
 170         return a_ino == b_ino;
 171 }
 172
 173 static void continue_as_child(void)
 174 {
 175         pid_t child = fork();
 176         int status;
 177         pid_t ret;
 178
 179         if (child < 0)
 180                 err(EXIT_FAILURE, _("fork failed"));
 181
 182         /* Only the child returns */
 183         if (child == 0)
 184                 return;
 185
 186         for (;;) {
 187                 ret = waitpid(child, &status, WUNTRACED);
 188                 if ((ret == child) && (WIFSTOPPED(status))) {
 189                         /* The child suspended so suspend us as well */
 190                         kill(getpid(), SIGSTOP);
 191                         kill(child, SIGCONT);
 192                 } else {
 193                         break;
 194                 }
 195         }
 196         /* Return the child's exit code if possible */
 197         if (WIFEXITED(status)) {
 198                 exit(WEXITSTATUS(status));
 199         } else if (WIFSIGNALED(status)) {
 200                 kill(getpid(), WTERMSIG(status));
 201         }
 202         exit(EXIT_FAILURE);
 203 }
 204
 205 int main(int argc, char *argv[])
 206 {
 207         enum {
 208                 OPT_PRESERVE_CRED = CHAR_MAX + 1
 209         };
 210         static const struct option longopts[] = {
 211                 { "all", no_argument, NULL, 'a' },
 212                 { "help", no_argument, NULL, 'h' },
 213                 { "version", no_argument, NULL, 'V'},
 214                 { "target", required_argument, NULL, 't' },
 215                 { "mount", optional_argument, NULL, 'm' },
 216                 { "uts", optional_argument, NULL, 'u' },
 217                 { "ipc", optional_argument, NULL, 'i' },
 218                 { "net", optional_argument, NULL, 'n' },
 219                 { "pid", optional_argument, NULL, 'p' },
 220                 { "user", optional_argument, NULL, 'U' },
 221                 { "cgroup", optional_argument, NULL, 'C' },
 222                 { "setuid", required_argument, NULL, 'S' },
 223                 { "setgid", required_argument, NULL, 'G' },
 224                 { "root", optional_argument, NULL, 'r' },
 225                 { "wd", optional_argument, NULL, 'w' },
 226                 { "no-fork", no_argument, NULL, 'F' },
 227                 { "preserve-credentials", no_argument, NULL, OPT_PRESERVE_CRED },
 228 #ifdef HAVE_LIBSELINUX
 229                 { "follow-context", no_argument, NULL, 'Z' },
 230 #endif
 231                 { NULL, 0, NULL, 0 }
 232         };
 233
 234         struct namespace_file *nsfile;
 235         int c, pass, namespaces = 0, setgroups_nerrs = 0, preserve_cred = 0;
 236         bool do_rd = false, do_wd = false, force_uid = false, force_gid = false;
 237         bool do_all = false;
 238         int do_fork = -1; /* unknown yet */
 239         uid_t uid = 0;
 240         gid_t gid = 0;
 241 #ifdef HAVE_LIBSELINUX
 242         bool selinux = 0;
 243 #endif
 244
 245         setlocale(LC_ALL, "");
 246         bindtextdomain(PACKAGE, LOCALEDIR);
 247         textdomain(PACKAGE);
 248         close_stdout_atexit();
 249
 250         while ((c =
 251                 getopt_long(argc, argv, "+ahVt:m::u::i::n::p::C::U::S:G:r::w::FZ",
 252                             longopts, NULL)) != -1) {
 253                 switch (c) {
 254                 case 'a':
 255                         do_all = true;
 256                         break;
 257                 case 't':
 258                         namespace_target_pid =
 259                             strtoul_or_err(optarg, _("failed to parse pid"));
 260                         break;
 261                 case 'm':
 262                         if (optarg)
 263                                 open_namespace_fd(CLONE_NEWNS, optarg);
 264                         else
 265                                 namespaces |= CLONE_NEWNS;
 266                         break;
 267                 case 'u':
 268                         if (optarg)
 269                                 open_namespace_fd(CLONE_NEWUTS, optarg);
 270                         else
 271                                 namespaces |= CLONE_NEWUTS;
 272                         break;
 273                 case 'i':
 274                         if (optarg)
 275                                 open_namespace_fd(CLONE_NEWIPC, optarg);
 276                         else
 277                                 namespaces |= CLONE_NEWIPC;
 278                         break;
 279                 case 'n':
 280                         if (optarg)
 281                                 open_namespace_fd(CLONE_NEWNET, optarg);
 282                         else
 283                                 namespaces |= CLONE_NEWNET;
 284                         break;
 285                 case 'p':
 286                         if (optarg)
 287                                 open_namespace_fd(CLONE_NEWPID, optarg);
 288                         else
 289                                 namespaces |= CLONE_NEWPID;
 290                         break;
 291                 case 'C':
 292                         if (optarg)
 293                                 open_namespace_fd(CLONE_NEWCGROUP, optarg);
 294                         else
 295                                 namespaces |= CLONE_NEWCGROUP;
 296                         break;
 297                 case 'U':
 298                         if (optarg)
 299                                 open_namespace_fd(CLONE_NEWUSER, optarg);
 300                         else
 301                                 namespaces |= CLONE_NEWUSER;
 302                         break;
 303                 case 'S':
 304                         uid = strtoul_or_err(optarg, _("failed to parse uid"));
 305                         force_uid = true;
 306                         break;
 307                 case 'G':
 308                         gid = strtoul_or_err(optarg, _("failed to parse gid"));
 309                         force_gid = true;
 310                         break;
 311                 case 'F':
 312                         do_fork = 0;
 313                         break;
 314                 case 'r':
 315                         if (optarg)
 316                                 open_target_fd(&root_fd, "root", optarg);
 317                         else
 318                                 do_rd = true;
 319                         break;
 320                 case 'w':
 321                         if (optarg)
 322                                 open_target_fd(&wd_fd, "cwd", optarg);
 323                         else
 324                                 do_wd = true;
 325                         break;
 326                 case OPT_PRESERVE_CRED:
 327                         preserve_cred = 1;
 328                         break;
 329 #ifdef HAVE_LIBSELINUX
 330                 case 'Z':
 331                         selinux = 1;
 332                         break;
 333 #endif
 334                 case 'h':
 335                         usage();
 336                 case 'V':
 337                         print_version(EXIT_SUCCESS);
 338                 default:
 339                         errtryhelp(EXIT_FAILURE);
 340                 }
 341         }
 342
 343 #ifdef HAVE_LIBSELINUX
 344         if (selinux && is_selinux_enabled() > 0) {
 345                 char *scon = NULL;
 346
 347                 if (!namespace_target_pid)
 348                         errx(EXIT_FAILURE, _("no target PID specified for --follow-context"));
 349                 if (getpidcon(namespace_target_pid, &scon) < 0)
 350                         errx(EXIT_FAILURE, _("failed to get %d SELinux context"),
 351                                         (int) namespace_target_pid);
 352                 if (setexeccon(scon) < 0)
 353                         errx(EXIT_FAILURE, _("failed to set exec context to '%s'"), scon);
 354                 freecon(scon);
 355         }
 356 #endif
 357
 358         if (do_all) {
 359                 if (!namespace_target_pid)
 360                         errx(EXIT_FAILURE, _("no target PID specified for --all"));
 361                 for (nsfile = namespace_files; nsfile->nstype; nsfile++) {
 362                         if (nsfile->fd >= 0)
 363                                 continue;       /* namespace already specified */
 364
 365                         /* It is not permitted to use setns(2) to reenter the caller's
 366                          * current user namespace; see setns(2) man page for more details.
 367                          */
 368                         if (nsfile->nstype & CLONE_NEWUSER
 369                             && is_same_namespace(getpid(), namespace_target_pid, nsfile->name))
 370                                 continue;
 371
 372                         namespaces |= nsfile->nstype;
 373                 }
 374         }
 375
 376         /*
 377          * Open remaining namespace and directory descriptors.
 378          */
 379         for (nsfile = namespace_files; nsfile->nstype; nsfile++)
 380                 if (nsfile->nstype & namespaces)
 381                         open_namespace_fd(nsfile->nstype, NULL);
 382         if (do_rd)
 383                 open_target_fd(&root_fd, "root", NULL);
 384         if (do_wd)
 385                 open_target_fd(&wd_fd, "cwd", NULL);
 386
 387         /*
 388          * Update namespaces variable to contain all requested namespaces
 389          */
 390         for (nsfile = namespace_files; nsfile->nstype; nsfile++) {
 391                 if (nsfile->fd < 0)
 392                         continue;
 393                 namespaces |= nsfile->nstype;
 394         }
 395
 396         /* for user namespaces we always set UID and GID (default is 0)
 397          * and clear root's groups if --preserve-credentials is no specified */
 398         if ((namespaces & CLONE_NEWUSER) && !preserve_cred) {
 399                 force_uid = true, force_gid = true;
 400
 401                 /* We call setgroups() before and after we enter user namespace,
 402                  * let's complain only if both fail */
 403                 if (setgroups(0, NULL) != 0)
 404                         setgroups_nerrs++;
 405         }
 406
 407         /*
 408          * Now that we know which namespaces we want to enter, enter
 409          * them.  Do this in two passes, not entering the user
 410          * namespace on the first pass.  So if we're deprivileging the
 411          * container we'll enter the user namespace last and if we're
 412          * privileging it then we enter the user namespace first
 413          * (because the initial setns will fail).
 414          */
 415         for (pass = 0; pass < 2; pass ++) {
 416                 for (nsfile = namespace_files + 1 - pass; nsfile->nstype; nsfile++) {
 417                         if (nsfile->fd < 0)
 418                                 continue;
 419                         if (nsfile->nstype == CLONE_NEWPID && do_fork == -1)
 420                                 do_fork = 1;
 421                         if (setns(nsfile->fd, nsfile->nstype)) {
 422                                 if (pass != 0)
 423                                         err(EXIT_FAILURE,
 424                                             _("reassociate to namespace '%s' failed"),
 425                                             nsfile->name);
 426                                 else
 427                                         continue;
 428                         }
 429
 430                         close(nsfile->fd);
 431                         nsfile->fd = -1;
 432                 }
 433         }
 434
 435         /* Remember the current working directory if I'm not changing it */
 436         if (root_fd >= 0 && wd_fd < 0) {
 437                 wd_fd = open(".", O_RDONLY);
 438                 if (wd_fd < 0)
 439                         err(EXIT_FAILURE,
 440                             _("cannot open current working directory"));
 441         }
 442
 443         /* Change the root directory */
 444         if (root_fd >= 0) {
 445                 if (fchdir(root_fd) < 0)
 446                         err(EXIT_FAILURE,
 447                             _("change directory by root file descriptor failed"));
 448
 449                 if (chroot(".") < 0)
 450                         err(EXIT_FAILURE, _("chroot failed"));
 451
 452                 close(root_fd);
 453                 root_fd = -1;
 454         }
 455
 456         /* Change the working directory */
 457         if (wd_fd >= 0) {
 458                 if (fchdir(wd_fd) < 0)
 459                         err(EXIT_FAILURE,
 460                             _("change directory by working directory file descriptor failed"));
 461
 462                 close(wd_fd);
 463                 wd_fd = -1;
 464         }
 465
 466         if (do_fork == 1)
 467                 continue_as_child();
 468
 469         if (force_uid || force_gid) {
 470                 if (force_gid && setgroups(0, NULL) != 0 && setgroups_nerrs)    /* drop supplementary groups */
 471                         err(EXIT_FAILURE, _("setgroups failed"));
 472                 if (force_gid && setgid(gid) < 0)               /* change GID */
 473                         err(EXIT_FAILURE, _("setgid failed"));
 474                 if (force_uid && setuid(uid) < 0)               /* change UID */
 475                         err(EXIT_FAILURE, _("setuid failed"));
 476         }
 477
 478         if (optind < argc) {
 479                 execvp(argv[optind], argv + optind);
 480                 errexec(argv[optind]);
 481         }
 482         exec_shell();
 483 }