(1ULL << CAP_SYS_PTRACE) |
(1ULL << CAP_SYS_TTY_CONFIG) |
(1ULL << CAP_SYS_RESOURCE) |
(1ULL << CAP_SYS_PTRACE) |
(1ULL << CAP_SYS_TTY_CONFIG) |
(1ULL << CAP_SYS_RESOURCE) |
static int help(void) {
printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
"Spawn a minimal namespace container for debugging, testing and building.\n\n"
" -h --help Show this help\n"
static int help(void) {
printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
"Spawn a minimal namespace container for debugging, testing and building.\n\n"
" -h --help Show this help\n"
" -D --directory=NAME Root directory for the container\n"
" -b --boot Boot up full system (i.e. invoke init)\n"
" -u --user=USER Run the command under specified user or uid\n"
" -D --directory=NAME Root directory for the container\n"
" -b --boot Boot up full system (i.e. invoke init)\n"
" -u --user=USER Run the command under specified user or uid\n"
static const struct option options[] = {
{ "help", no_argument, NULL, 'h' },
static const struct option options[] = {
{ "help", no_argument, NULL, 'h' },
{ "directory", required_argument, NULL, 'D' },
{ "user", required_argument, NULL, 'u' },
{ "controllers", required_argument, NULL, 'C' },
{ "directory", required_argument, NULL, 'D' },
{ "user", required_argument, NULL, 'u' },
{ "controllers", required_argument, NULL, 'C' },
int t;
if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
int t;
if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
+ r = readlink_malloc("/etc/localtime", &p);
+ if (r < 0) {
+ log_warning("/etc/localtime is not a symlink, not updating container timezone.");
+ return 0;
+ }
+
+ z = path_startswith(p, "../usr/share/zoneinfo/");
+ if (!z)
+ z = path_startswith(p, "/usr/share/zoneinfo/");
+ if (!z) {
+ log_warning("/etc/localtime does not point into /usr/share/zoneinfo/, not updating container timezone.");
+ return 0;
+ }
+
- if (mount("/etc/localtime", where, "bind", MS_BIND, NULL) >= 0)
- mount("/etc/localtime", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+ r = readlink_malloc(where, &q);
+ if (r >= 0) {
+ y = path_startswith(q, "../usr/share/zoneinfo/");
+ if (!y)
+ y = path_startswith(q, "/usr/share/zoneinfo/");
- where = strappend(dest, "/etc/timezone");
- if (!where)
+ /* Already pointing to the right place? Then do nothing .. */
+ if (y && streq(y, z))
+ return 0;
+ }
+
+ check = strjoin(dest, "/usr/share/zoneinfo/", z, NULL);
+ if (!check)
- if (mount("/etc/timezone", where, "bind", MS_BIND, NULL) >= 0)
- mount("/etc/timezone", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+ if (access(check, F_OK) < 0) {
+ log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
+ return 0;
+ }
if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0)
mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0)
mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
* the container gets a new one */
from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
* the container gets a new one */
from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
r = sd_id128_randomize(&rnd);
if (r < 0) {
log_error("Failed to generate random boot id: %s", strerror(-r));
r = sd_id128_randomize(&rnd);
if (r < 0) {
log_error("Failed to generate random boot id: %s", strerror(-r));
r = write_one_line_file(from, as_uuid);
if (r < 0) {
log_error("Failed to write boot id: %s", strerror(-r));
r = write_one_line_file(from, as_uuid);
if (r < 0) {
log_error("Failed to write boot id: %s", strerror(-r));
}
if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
}
if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
unlink(from);
mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
unlink(from);
asprintf(&from, "/dev/%s", d);
asprintf(&to, "%s/dev/%s", dest, d);
if (!from || !to) {
asprintf(&from, "/dev/%s", d);
asprintf(&to, "%s/dev/%s", dest, d);
if (!from || !to) {
if (stat(console, &st) < 0) {
log_error("Failed to stat %s: %m", console);
if (stat(console, &st) < 0) {
log_error("Failed to stat %s: %m", console);
}
r = chmod_and_chown(console, 0600, 0, 0);
if (r < 0) {
log_error("Failed to correct access mode for TTY: %s", strerror(-r));
}
r = chmod_and_chown(console, 0600, 0, 0);
if (r < 0) {
log_error("Failed to correct access mode for TTY: %s", strerror(-r));
/* We need to bind mount the right tty to /dev/console since
* ptys can only exist on pts file systems. To have something
/* We need to bind mount the right tty to /dev/console since
* ptys can only exist on pts file systems. To have something
if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) {
log_error("mknod() for /dev/console failed: %m");
if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) {
log_error("mknod() for /dev/console failed: %m");
}
if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
log_error("Bind mount for /dev/console failed: %m");
}
if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
log_error("Bind mount for /dev/console failed: %m");
* that writing blocks when nothing is reading. In order to
* avoid any problems with containers deadlocking due to this
* we simply make /dev/kmsg unavailable to the container. */
* that writing blocks when nothing is reading. In order to
* avoid any problems with containers deadlocking due to this
* we simply make /dev/kmsg unavailable to the container. */
- if (asprintf(&from, "%s/dev/kmsg", dest) < 0) {
- r = log_oom();
- goto finish;
- }
-
- if (asprintf(&to, "%s/proc/kmsg", dest) < 0) {
- r = log_oom();
- goto finish;
- }
+ if (asprintf(&from, "%s/dev/kmsg", dest) < 0 ||
+ asprintf(&to, "%s/proc/kmsg", dest) < 0)
+ return log_oom();
if (mkfifo(from, 0600) < 0) {
log_error("mkfifo() for /dev/kmsg failed: %m");
if (mkfifo(from, 0600) < 0) {
log_error("mkfifo() for /dev/kmsg failed: %m");
}
r = chmod_and_chown(from, 0600, 0, 0);
if (r < 0) {
log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
}
r = chmod_and_chown(from, 0600, 0, 0);
if (r < 0) {
log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
}
if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
log_error("Bind mount for /proc/kmsg failed: %m");
}
if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
log_error("Bind mount for /proc/kmsg failed: %m");
}
fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
if (fd < 0) {
log_error("Failed to open fifo: %m");
}
fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
if (fd < 0) {
log_error("Failed to open fifo: %m");
- char *p = NULL, *b = NULL, *l, *q = NULL, *d = NULL;
+ char _cleanup_free_ *p = NULL, *b = NULL, *q = NULL, *d = NULL;
+ char *id;
int r;
if (arg_link_journal == LINK_NO)
return 0;
p = strappend(directory, "/etc/machine-id");
int r;
if (arg_link_journal == LINK_NO)
return 0;
p = strappend(directory, "/etc/machine-id");
- if (r == -ENOENT && arg_link_journal == LINK_AUTO) {
- r = 0;
- goto finish;
- } else if (r < 0) {
- log_error("Failed to read machine ID: %s", strerror(-r));
+ if (r == -ENOENT && arg_link_journal == LINK_AUTO)
+ return 0;
+ else if (r < 0) {
+ log_error("Failed to read machine ID from %s: %s", p, strerror(-r));
- p = strappend("/var/log/journal/", l);
- q = strjoin(directory, "/var/log/journal/", l, NULL);
- if (!p || !q) {
- r = log_oom();
- goto finish;
+ p = strappend("/var/log/journal/", id);
+ q = strjoin(directory, "/var/log/journal/", id, NULL);
+ if (!p || !q)
+ return log_oom();
+
+ if (path_is_mount_point(p, false) > 0) {
+ if (arg_link_journal != LINK_AUTO) {
+ log_error("%s: already a mount point, refusing to use for journal", p);
+ return -EEXIST;
+ }
+
+ return 0;
- if (path_is_mount_point(p, false) > 0 ||
- path_is_mount_point(q, false) > 0) {
+ if (path_is_mount_point(q, false) > 0) {
}
} else if (r != -ENOENT) {
log_error("readlink(%s) failed: %m", p);
}
} else if (r != -ENOENT) {
log_error("readlink(%s) failed: %m", p);
}
if (arg_link_journal == LINK_GUEST) {
if (symlink(q, p) < 0) {
log_error("Failed to symlink %s to %s: %m", q, p);
}
if (arg_link_journal == LINK_GUEST) {
if (symlink(q, p) < 0) {
log_error("Failed to symlink %s to %s: %m", q, p);
}
if (arg_link_journal == LINK_HOST) {
r = mkdir_p(p, 0755);
if (r < 0) {
log_error("Failed to create %s: %m", p);
}
if (arg_link_journal == LINK_HOST) {
r = mkdir_p(p, 0755);
if (r < 0) {
log_error("Failed to create %s: %m", p);
if (dir_is_empty(q) == 0) {
log_error("%s not empty.", q);
if (dir_is_empty(q) == 0) {
log_error("%s not empty.", q);
}
if (mount(p, q, "bind", MS_BIND, NULL) < 0) {
log_error("Failed to bind mount journal from host into guest: %m");
}
if (mount(p, q, "bind", MS_BIND, NULL) < 0) {
log_error("Failed to bind mount journal from host into guest: %m");
char in_buffer[LINE_MAX], out_buffer[LINE_MAX];
size_t in_buffer_full = 0, out_buffer_full = 0;
struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev;
bool stdin_readable = false, stdout_writable = false, master_readable = false, master_writable = false;
int ep = -1, signal_fd = -1, r;
char in_buffer[LINE_MAX], out_buffer[LINE_MAX];
size_t in_buffer_full = 0, out_buffer_full = 0;
struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev;
bool stdin_readable = false, stdout_writable = false, master_readable = false, master_writable = false;
int ep = -1, signal_fd = -1, r;
fd_nonblock(STDIN_FILENO, 1);
fd_nonblock(STDOUT_FILENO, 1);
fd_nonblock(STDIN_FILENO, 1);
fd_nonblock(STDOUT_FILENO, 1);
- zero(stdin_ev);
- stdin_ev.events = EPOLLIN|EPOLLET;
- stdin_ev.data.fd = STDIN_FILENO;
+ /* We read from STDIN only if this is actually a TTY,
+ * otherwise we assume non-interactivity. */
+ if (isatty(STDIN_FILENO)) {
+ zero(stdin_ev);
+ stdin_ev.events = EPOLLIN|EPOLLET;
+ stdin_ev.data.fd = STDIN_FILENO;
+
+ if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0) {
+ log_error("Failed to register STDIN in epoll: %m");
+ r = -errno;
+ goto finish;
+ }
+ }
- if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0 ||
- epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 ||
- epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 ||
+ if (epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0) {
+ if (errno != EPERM) {
+ log_error("Failed to register stdout in epoll: %m");
+ r = -errno;
+ goto finish;
+ }
+ /* stdout without epoll support. Likely redirected to regular file. */
+ stdout_writable = true;
+ }
+
+ if (epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 ||
epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) {
epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) {
/* The window size changed, let's forward that. */
if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0)
ioctl(master, TIOCSWINSZ, &ws);
/* The window size changed, let's forward that. */
if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0)
ioctl(master, TIOCSWINSZ, &ws);
+ } else if (sfsi.ssi_signo == SIGTERM && arg_boot && !tried_orderly_shutdown) {
+
+ log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination.");
+
+ /* This only works for systemd... */
+ tried_orderly_shutdown = true;
+ kill(pid, SIGRTMIN+3);
+
int r = EXIT_FAILURE, k;
char *oldcg = NULL, *newcg = NULL;
char **controller = NULL;
int r = EXIT_FAILURE, k;
char *oldcg = NULL, *newcg = NULL;
char **controller = NULL;
const char *console = NULL;
struct termios saved_attr, raw_attr;
sigset_t mask;
bool saved_attr_valid = false;
struct winsize ws;
int kmsg_socket_pair[2] = { -1, -1 };
const char *console = NULL;
struct termios saved_attr, raw_attr;
sigset_t mask;
bool saved_attr_valid = false;
struct winsize ws;
int kmsg_socket_pair[2] = { -1, -1 };
+ log_close();
+ n_fd_passed = sd_listen_fds(false);
+ if (n_fd_passed > 0) {
+ k = fdset_new_listen_fds(&fds, false);
+ if (k < 0) {
+ log_error("Failed to collect file descriptors: %s", strerror(-k));
+ goto finish;
+ }
+ }
+ fdset_close_others(fds);
+ log_open();
+
k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg);
if (k < 0) {
log_error("Failed to determine current cgroup: %s", strerror(-k));
k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg);
if (k < 0) {
log_error("Failed to determine current cgroup: %s", strerror(-k));
- if (tcgetattr(STDIN_FILENO, &saved_attr) < 0) {
- log_error("Failed to get terminal attributes: %m");
- goto finish;
- }
-
- saved_attr_valid = true;
+ if (tcgetattr(STDIN_FILENO, &saved_attr) >= 0) {
+ saved_attr_valid = true;
if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
log_error("Failed to create kmsg socket pair");
if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
log_error("Failed to create kmsg socket pair");
- if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) {
- log_error("Failed to set terminal attributes: %m");
+ if(pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
+ log_error("pipe2(): %m");
const char *envp[] = {
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
"container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
const char *envp[] = {
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
"container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */
close_nointr(STDIN_FILENO);
close_nointr(STDOUT_FILENO);
close_nointr(STDERR_FILENO);
close_nointr(STDIN_FILENO);
close_nointr(STDOUT_FILENO);
close_nointr(STDERR_FILENO);
reset_all_signal_handlers();
assert_se(sigemptyset(&mask) == 0);
assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
reset_all_signal_handlers();
assert_se(sigemptyset(&mask) == 0);
assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
- if (open_terminal(console, O_RDWR) != STDIN_FILENO ||
- dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
- dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
+ k = open_terminal(console, O_RDWR);
+ if (k != STDIN_FILENO) {
+ if (k >= 0) {
+ close_nointr_nofail(k);
+ k = -EINVAL;
+ }
+
+ log_error("Failed to open console: %s", strerror(-k));
+ goto child_fail;
+ }
+
+ if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
+ dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) {
+ log_error("Failed to duplicate console: %m");
+ /* Note that this resolves user names
+ * inside the container, and hence
+ * accesses the NSS modules from the
+ * container and not the host. This is
+ * a bit weird... */
+
if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) {
log_error("get_user_creds() failed: %m");
goto child_fail;
if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) {
log_error("get_user_creds() failed: %m");
goto child_fail;
+ } else {
+ /* Reset everything fully to 0, just in case */
+
+ if (setgroups(0, NULL) < 0) {
+ log_error("setgroups() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresgid(0, 0, 0) < 0) {
+ log_error("setregid() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresuid(0, 0, 0) < 0) {
+ log_error("setreuid() failed: %m");
+ goto child_fail;
+ }
- if ((asprintf((char**)(envp + 3), "HOME=%s", home ? home: "/root") < 0) ||
- (asprintf((char**)(envp + 4), "USER=%s", arg_user ? arg_user : "root") < 0) ||
- (asprintf((char**)(envp + 5), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
+ if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) ||
+ (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) ||
+ (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) {
- if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) {
+ if (asprintf((char**)(envp + n_env++), "container_uuid=%s", arg_uuid) < 0) {
+ log_oom();
+ goto child_fail;
+ }
+ }
+
+ if (fdset_size(fds) > 0) {
+ k = fdset_cloexec(fds, false);
+ if (k < 0) {
+ log_error("Failed to unset O_CLOEXEC for file descriptors.");
+ goto child_fail;
+ }
+
+ if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) ||
+ (asprintf((char **)(envp + n_env++), "LISTEN_PID=%lu", (unsigned long) getpid()) < 0)) {
- if (process_pty(master, &mask) < 0)
- goto finish;
+ log_info("Init process in the container running as PID %d", pid);
+ close_nointr_nofail(pipefd[0]);
+ close_nointr_nofail(pipefd[1]);