1
/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4
This file is part of systemd.
6
Copyright 2010 Lennart Poettering
8
systemd is free software; you can redistribute it and/or modify it
9
under the terms of the GNU General Public License as published by
10
the Free Software Foundation; either version 2 of the License, or
11
(at your option) any later version.
13
systemd is distributed in the hope that it will be useful, but
14
WITHOUT ANY WARRANTY; without even the implied warranty of
15
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16
General Public License for more details.
18
You should have received a copy of the GNU General Public License
19
along with systemd; If not, see <http://www.gnu.org/licenses/>.
25
#include <sys/types.h>
26
#include <sys/syscall.h>
27
#include <sys/mount.h>
33
#include <sys/prctl.h>
34
#include <sys/capability.h>
36
#include <sys/epoll.h>
38
#include <sys/signalfd.h>
42
#include <systemd/sd-daemon.h>
47
#include "cgroup-util.h"
49
#include "loopback-setup.h"
51
static char *arg_directory = NULL;
52
static char *arg_user = NULL;
53
static bool arg_private_network = false;
55
static int help(void) {
57
printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
58
"Spawn a minimal namespace container for debugging, testing and building.\n\n"
59
" -h --help Show this help\n"
60
" -D --directory=NAME Root directory for the container\n"
61
" -u --user=USER Run the command under specified user or uid\n"
62
" --private-network Disable network in container\n",
63
program_invocation_short_name);
68
static int parse_argv(int argc, char *argv[]) {
71
ARG_PRIVATE_NETWORK = 0x100
74
static const struct option options[] = {
75
{ "help", no_argument, NULL, 'h' },
76
{ "directory", required_argument, NULL, 'D' },
77
{ "user", required_argument, NULL, 'u' },
78
{ "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK },
87
while ((c = getopt_long(argc, argv, "+hD:u:", options, NULL)) >= 0) {
97
if (!(arg_directory = strdup(optarg))) {
98
log_error("Failed to duplicate root directory.");
106
if (!(arg_user = strdup(optarg))) {
107
log_error("Failed to duplicate user name.");
113
case ARG_PRIVATE_NETWORK:
114
arg_private_network = true;
121
log_error("Unknown option code %c", c);
129
static int mount_all(const char *dest) {
131
typedef struct MountPoint {
140
static const MountPoint mount_table[] = {
141
{ "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
142
{ "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */
143
{ "/proc/sys", "/proc/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
144
{ "/sys", "/sys", "bind", NULL, MS_BIND, true }, /* Bind mount first */
145
{ "/sys", "/sys", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */
146
{ "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID, true },
147
{ "/dev/pts", "/dev/pts", "bind", NULL, MS_BIND, true },
148
{ "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV, true },
150
{ "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND, false }, /* Bind mount first */
151
{ "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */
159
for (k = 0; k < ELEMENTSOF(mount_table); k++) {
162
if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
163
log_error("Out of memory");
171
if ((t = path_is_mount_point(where, false)) < 0) {
172
log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t));
181
mkdir_p(where, 0755);
183
if (mount(mount_table[k].what,
186
mount_table[k].flags,
187
mount_table[k].options) < 0 &&
188
mount_table[k].fatal) {
190
log_error("mount(%s) failed: %m", where);
199
/* Fix the timezone, if possible */
200
if (asprintf(&where, "%s/etc/localtime", dest) >= 0) {
202
if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0)
203
mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
208
if (asprintf(&where, "%s/etc/timezone", dest) >= 0) {
210
if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0)
211
mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
219
static int copy_devnodes(const char *dest, const char *console) {
221
static const char devnodes[] =
236
char *from = NULL, *to = NULL;
243
NULSTR_FOREACH(d, devnodes) {
246
asprintf(&from, "/dev/%s", d);
247
asprintf(&to, "%s/dev/%s", dest, d);
250
log_error("Failed to allocate devnode path");
263
if (stat(from, &st) < 0) {
265
if (errno != ENOENT) {
266
log_error("Failed to stat %s: %m", from);
271
} else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
273
log_error("%s is not a char or block device, cannot copy.", from);
277
} else if (mknod(to, st.st_mode, st.st_rdev) < 0) {
279
log_error("mknod(%s) failed: %m", dest);
288
if (stat(console, &st) < 0) {
290
log_error("Failed to stat %s: %m", console);
296
} else if (!S_ISCHR(st.st_mode)) {
298
log_error("/dev/console is not a char device.");
305
if (asprintf(&to, "%s/dev/console", dest) < 0) {
307
log_error("Out of memory");
314
/* We need to bind mount the right tty to /dev/console since
315
* ptys can only exist on pts file systems. To have something
316
* to bind mount things on we create a device node first, that
317
* has the right major/minor (note that the major minor
318
* doesn't actually matter here, since we mount it over
321
if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0)
322
log_error("mknod for /dev/console failed: %m");
324
if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
325
log_error("bind mount for /dev/console failed: %m");
333
if ((k = chmod_and_chown(console, 0600, 0, 0)) < 0) {
334
log_error("Failed to correct access mode for TTY: %s", strerror(-k));
346
static int drop_capabilities(void) {
347
static const unsigned long retain[] = {
357
CAP_NET_BIND_SERVICE,
373
for (l = 0; l <= cap_last_cap(); l++) {
376
for (i = 0; i < ELEMENTSOF(retain); i++)
380
if (i < ELEMENTSOF(retain))
383
if (prctl(PR_CAPBSET_DROP, l) < 0) {
384
log_error("PR_CAPBSET_DROP failed: %m");
392
static int is_os_tree(const char *path) {
395
/* We use /bin/sh as flag file if something is an OS */
397
if (asprintf(&p, "%s/bin/sh", path) < 0)
403
return r < 0 ? 0 : 1;
406
static int process_pty(int master, sigset_t *mask) {
408
char in_buffer[LINE_MAX], out_buffer[LINE_MAX];
409
size_t in_buffer_full = 0, out_buffer_full = 0;
410
struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev;
411
bool stdin_readable = false, stdout_writable = false, master_readable = false, master_writable = false;
412
int ep = -1, signal_fd = -1, r;
414
fd_nonblock(STDIN_FILENO, 1);
415
fd_nonblock(STDOUT_FILENO, 1);
416
fd_nonblock(master, 1);
418
if ((signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC)) < 0) {
419
log_error("signalfd(): %m");
424
if ((ep = epoll_create1(EPOLL_CLOEXEC)) < 0) {
425
log_error("Failed to create epoll: %m");
431
stdin_ev.events = EPOLLIN|EPOLLET;
432
stdin_ev.data.fd = STDIN_FILENO;
435
stdout_ev.events = EPOLLOUT|EPOLLET;
436
stdout_ev.data.fd = STDOUT_FILENO;
439
master_ev.events = EPOLLIN|EPOLLOUT|EPOLLET;
440
master_ev.data.fd = master;
443
signal_ev.events = EPOLLIN;
444
signal_ev.data.fd = signal_fd;
446
if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0 ||
447
epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 ||
448
epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 ||
449
epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) {
450
log_error("Failed to regiser fds in epoll: %m");
456
struct epoll_event ev[16];
460
if ((nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1)) < 0) {
462
if (errno == EINTR || errno == EAGAIN)
465
log_error("epoll_wait(): %m");
472
for (i = 0; i < nfds; i++) {
473
if (ev[i].data.fd == STDIN_FILENO) {
475
if (ev[i].events & (EPOLLIN|EPOLLHUP))
476
stdin_readable = true;
478
} else if (ev[i].data.fd == STDOUT_FILENO) {
480
if (ev[i].events & (EPOLLOUT|EPOLLHUP))
481
stdout_writable = true;
483
} else if (ev[i].data.fd == master) {
485
if (ev[i].events & (EPOLLIN|EPOLLHUP))
486
master_readable = true;
488
if (ev[i].events & (EPOLLOUT|EPOLLHUP))
489
master_writable = true;
491
} else if (ev[i].data.fd == signal_fd) {
492
struct signalfd_siginfo sfsi;
495
if ((n = read(signal_fd, &sfsi, sizeof(sfsi))) != sizeof(sfsi)) {
498
log_error("Failed to read from signalfd: invalid block size");
503
if (errno != EINTR && errno != EAGAIN) {
504
log_error("Failed to read from signalfd: %m");
510
if (sfsi.ssi_signo == SIGWINCH) {
513
/* The window size changed, let's forward that. */
514
if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0)
515
ioctl(master, TIOCSWINSZ, &ws);
524
while ((stdin_readable && in_buffer_full <= 0) ||
525
(master_writable && in_buffer_full > 0) ||
526
(master_readable && out_buffer_full <= 0) ||
527
(stdout_writable && out_buffer_full > 0)) {
529
if (stdin_readable && in_buffer_full < LINE_MAX) {
531
if ((k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full)) < 0) {
533
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
534
stdin_readable = false;
536
log_error("read(): %m");
541
in_buffer_full += (size_t) k;
544
if (master_writable && in_buffer_full > 0) {
546
if ((k = write(master, in_buffer, in_buffer_full)) < 0) {
548
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
549
master_writable = false;
551
log_error("write(): %m");
557
assert(in_buffer_full >= (size_t) k);
558
memmove(in_buffer, in_buffer + k, in_buffer_full - k);
563
if (master_readable && out_buffer_full < LINE_MAX) {
565
if ((k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full)) < 0) {
567
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
568
master_readable = false;
570
log_error("read(): %m");
575
out_buffer_full += (size_t) k;
578
if (stdout_writable && out_buffer_full > 0) {
580
if ((k = write(STDOUT_FILENO, out_buffer, out_buffer_full)) < 0) {
582
if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
583
stdout_writable = false;
585
log_error("write(): %m");
591
assert(out_buffer_full >= (size_t) k);
592
memmove(out_buffer, out_buffer + k, out_buffer_full - k);
593
out_buffer_full -= k;
601
close_nointr_nofail(ep);
604
close_nointr_nofail(signal_fd);
609
int main(int argc, char *argv[]) {
611
int r = EXIT_FAILURE, k;
612
char *oldcg = NULL, *newcg = NULL;
614
const char *console = NULL;
615
struct termios saved_attr, raw_attr;
617
bool saved_attr_valid = false;
620
log_parse_environment();
623
if ((r = parse_argv(argc, argv)) <= 0)
629
p = path_make_absolute_cwd(arg_directory);
633
arg_directory = get_current_dir_name();
635
if (!arg_directory) {
636
log_error("Failed to determine path");
640
path_kill_slashes(arg_directory);
642
if (geteuid() != 0) {
643
log_error("Need to be root.");
647
if (sd_booted() <= 0) {
648
log_error("Not running on a systemd system.");
652
if (path_equal(arg_directory, "/")) {
653
log_error("Spawning container on root directory not supported.");
657
if (is_os_tree(arg_directory) <= 0) {
658
log_error("Directory %s doesn't look like an OS root directory. Refusing.", arg_directory);
662
if ((k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg)) < 0) {
663
log_error("Failed to determine current cgroup: %s", strerror(-k));
667
if (asprintf(&newcg, "%s/nspawn-%lu", oldcg, (unsigned long) getpid()) < 0) {
668
log_error("Failed to allocate cgroup path.");
672
if ((k = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, newcg, 0)) < 0) {
673
log_error("Failed to create cgroup: %s", strerror(-k));
677
if ((master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY)) < 0) {
678
log_error("Failed to acquire pseudo tty: %m");
682
if (!(console = ptsname(master))) {
683
log_error("Failed to determine tty name: %m");
687
log_info("Spawning namespace container on %s (console is %s).", arg_directory, console);
689
if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0)
690
ioctl(master, TIOCSWINSZ, &ws);
692
if (unlockpt(master) < 0) {
693
log_error("Failed to unlock tty: %m");
697
if (tcgetattr(STDIN_FILENO, &saved_attr) < 0) {
698
log_error("Failed to get terminal attributes: %m");
702
saved_attr_valid = true;
704
raw_attr = saved_attr;
705
cfmakeraw(&raw_attr);
706
raw_attr.c_lflag &= ~ECHO;
708
if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) {
709
log_error("Failed to set terminal attributes: %m");
713
assert_se(sigemptyset(&mask) == 0);
714
sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
715
assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
717
pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
720
log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
722
log_error("clone() failed: %m");
731
const char *home = NULL;
732
uid_t uid = (uid_t) -1;
733
gid_t gid = (gid_t) -1;
734
const char *envp[] = {
735
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
736
"container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
744
envp[2] = strv_find_prefix(environ, "TERM=");
746
close_nointr_nofail(master);
748
close_nointr(STDIN_FILENO);
749
close_nointr(STDOUT_FILENO);
750
close_nointr(STDERR_FILENO);
752
close_all_fds(NULL, 0);
754
reset_all_signal_handlers();
756
assert_se(sigemptyset(&mask) == 0);
757
assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
762
if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0)
765
/* Mark / as private, in case somebody marked it shared */
766
if (mount(NULL, "/", NULL, MS_PRIVATE|MS_REC, NULL) < 0)
769
if (mount_all(arg_directory) < 0)
772
if (copy_devnodes(arg_directory, console) < 0)
775
if (chdir(arg_directory) < 0) {
776
log_error("chdir(%s) failed: %m", arg_directory);
780
if (open_terminal("dev/console", O_RDWR) != STDIN_FILENO ||
781
dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
782
dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
785
if (mount(arg_directory, "/", "bind", MS_BIND|MS_MOVE, NULL) < 0) {
786
log_error("mount(MS_MOVE) failed: %m");
790
if (chroot(".") < 0) {
791
log_error("chroot() failed: %m");
795
if (chdir("/") < 0) {
796
log_error("chdir() failed: %m");
804
if (drop_capabilities() < 0)
809
if (get_user_creds((const char**)&arg_user, &uid, &gid, &home) < 0) {
810
log_error("get_user_creds() failed: %m");
814
if (mkdir_parents(home, 0775) < 0) {
815
log_error("mkdir_parents() failed: %m");
819
if (safe_mkdir(home, 0775, uid, gid) < 0) {
820
log_error("safe_mkdir() failed: %m");
824
if (initgroups((const char*)arg_user, gid) < 0) {
825
log_error("initgroups() failed: %m");
829
if (setresgid(gid, gid, gid) < 0) {
830
log_error("setregid() failed: %m");
834
if (setresuid(uid, uid, uid) < 0) {
835
log_error("setreuid() failed: %m");
840
if ((asprintf((char**)(envp + 3), "HOME=%s", home? home: "/root") < 0) ||
841
(asprintf((char**)(envp + 4), "USER=%s", arg_user? arg_user : "root") < 0) ||
842
(asprintf((char**)(envp + 5), "LOGNAME=%s", arg_user? arg_user : "root") < 0)) {
843
log_error("Out of memory");
847
if ((hn = file_name_from_path(arg_directory)))
848
sethostname(hn, strlen(hn));
851
execvpe(argv[optind], argv + optind, (char**) envp);
853
chdir(home ? home : "/root");
854
execle("/bin/bash", "-bash", NULL, (char**) envp);
857
log_error("execv() failed: %m");
863
if (process_pty(master, &mask) < 0)
866
if (saved_attr_valid) {
867
tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
868
saved_attr_valid = false;
871
r = wait_for_terminate_and_warn(argc > optind ? argv[optind] : "bash", pid);
877
if (saved_attr_valid)
878
tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
881
close_nointr_nofail(master);
884
cg_attach(SYSTEMD_CGROUP_CONTROLLER, oldcg, 0);
887
cg_kill_recursive_and_wait(SYSTEMD_CGROUP_CONTROLLER, newcg, true);