#include "log.h"
#include "util.h"
#include "mkdir.h"
+#include "macro.h"
#include "audit.h"
#include "missing.h"
#include "cgroup-util.h"
#include "loopback-setup.h"
#include "sd-id128.h"
#include "dev-setup.h"
+#include "fdset.h"
+#include "build.h"
typedef enum LinkJournal {
LINK_NO,
(1ULL << CAP_SYS_NICE) |
(1ULL << CAP_SYS_PTRACE) |
(1ULL << CAP_SYS_TTY_CONFIG) |
- (1ULL << CAP_SYS_RESOURCE);
+ (1ULL << CAP_SYS_RESOURCE) |
+ (1ULL << CAP_SYS_BOOT);
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"
+ " --version Print version string\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 int parse_argv(int argc, char *argv[]) {
enum {
- ARG_PRIVATE_NETWORK = 0x100,
+ ARG_VERSION = 0x100,
+ ARG_PRIVATE_NETWORK,
ARG_UUID,
ARG_READ_ONLY,
ARG_CAPABILITY,
static const struct option options[] = {
{ "help", no_argument, NULL, 'h' },
+ { "version", no_argument, NULL, ARG_VERSION },
{ "directory", required_argument, NULL, 'D' },
{ "user", required_argument, NULL, 'u' },
{ "controllers", required_argument, NULL, 'C' },
help();
return 0;
+ case ARG_VERSION:
+ puts(PACKAGE_STRING);
+ puts(SYSTEMD_FEATURES);
+ return 0;
+
case 'D':
free(arg_directory);
arg_directory = canonicalize_file_name(optarg);
{ "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true },
{ "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true },
{ "/dev/pts", "/dev/pts", NULL, NULL, MS_BIND, true },
+ { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
{ "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true },
#ifdef HAVE_SELINUX
{ "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */
unsigned k;
int r = 0;
- char *where;
for (k = 0; k < ELEMENTSOF(mount_table); k++) {
+ char _cleanup_free_ *where = NULL;
int t;
if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
t = path_is_mount_point(where, true);
if (t < 0) {
log_error("Failed to detect whether %s is a mount point: %s", where, strerror(-t));
- free(where);
if (r == 0)
r = t;
if (r == 0)
r = -errno;
}
-
- free(where);
}
return r;
}
static int setup_timezone(const char *dest) {
- char *where;
+ _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL;
+ char *z, *y;
+ int r;
assert(dest);
/* Fix the timezone, if possible */
- if (asprintf(&where, "%s/etc/localtime", dest) < 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;
+ }
+
+ where = strappend(dest, "/etc/localtime");
+ if (!where)
return log_oom();
- 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/");
- free(where);
- if (asprintf(&where, "%s/etc/timezone", dest) < 0)
+ /* 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)
return log_oom();
- 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;
+ }
- free(where);
+ what = strappend("../usr/share/zoneinfo/", z);
+ if (!what)
+ return log_oom();
+
+ unlink(where);
+ if (symlink(what, where) < 0) {
+ log_error("Failed to correct timezone of container: %m");
+ return 0;
+ }
return 0;
}
return 0;
/* Fix resolv.conf, if possible */
- if (asprintf(&where, "%s/etc/resolv.conf", dest) < 0) {
+ where = strappend(dest, "/etc/resolv.conf");
+ if (!where)
return log_oom();
- }
+ /* We don't really care for the results of this really. If it
+ * fails, it fails, but meh... */
if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0)
mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
return 0;
}
+static int setup_boot_id(const char *dest) {
+ char _cleanup_free_ *from = NULL, *to = NULL;
+ sd_id128_t rnd;
+ char as_uuid[37];
+ int r;
+
+ assert(dest);
+
+ /* Generate a new randomized boot ID, so that each boot-up of
+ * the container gets a new one */
+
+ from = strappend(dest, "/dev/proc-sys-kernel-random-boot-id");
+ to = strappend(dest, "/proc/sys/kernel/random/boot_id");
+ if (!from || !to)
+ return log_oom();
+
+ r = sd_id128_randomize(&rnd);
+ if (r < 0) {
+ log_error("Failed to generate random boot id: %s", strerror(-r));
+ return r;
+ }
+
+ snprintf(as_uuid, sizeof(as_uuid),
+ "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
+ SD_ID128_FORMAT_VAL(rnd));
+ char_array_0(as_uuid);
+
+ r = write_one_line_file(from, as_uuid);
+ if (r < 0) {
+ log_error("Failed to write boot id: %s", strerror(-r));
+ return r;
+ }
+
+ if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
+ log_error("Failed to bind mount boot id: %m");
+ r = -errno;
+ } else
+ mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+
+ unlink(from);
+ return r;
+}
+
static int copy_devnodes(const char *dest) {
static const char devnodes[] =
"random\0"
"urandom\0"
"tty\0"
- "ptmx\0"
- "rtc0\0";
+ "ptmx\0";
const char *d;
int r = 0;
- mode_t u;
+ mode_t _cleanup_umask_ u;
assert(dest);
NULSTR_FOREACH(d, devnodes) {
struct stat st;
- char *from = NULL, *to = NULL;
+ char _cleanup_free_ *from = NULL, *to = NULL;
asprintf(&from, "/dev/%s", d);
asprintf(&to, "%s/dev/%s", dest, d);
if (!from || !to) {
- log_error("Failed to allocate devnode path");
-
- free(from);
- free(to);
-
- from = to = NULL;
+ log_oom();
if (r == 0)
r = -ENOMEM;
} else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
- log_error("%s is not a char or block device, cannot copy.", from);
+ log_error("%s is not a char or block device, cannot copy", from);
if (r == 0)
r = -EIO;
if (r == 0)
r = -errno;
}
-
- free(from);
- free(to);
}
- umask(u);
-
return r;
}
static int setup_dev_console(const char *dest, const char *console) {
struct stat st;
- char *to = NULL;
+ char _cleanup_free_ *to = NULL;
int r;
- mode_t u;
+ mode_t _cleanup_umask_ u;
assert(dest);
assert(console);
if (stat(console, &st) < 0) {
log_error("Failed to stat %s: %m", console);
- r = -errno;
- goto finish;
+ return -errno;
} else if (!S_ISCHR(st.st_mode)) {
- log_error("/dev/console is not a char device.");
- r = -EIO;
- goto finish;
+ log_error("/dev/console is not a char device");
+ return -EIO;
}
r = chmod_and_chown(console, 0600, 0, 0);
if (r < 0) {
log_error("Failed to correct access mode for TTY: %s", strerror(-r));
- goto finish;
+ return r;
}
- if (asprintf(&to, "%s/dev/console", dest) < 0) {
- r = log_oom();
- goto finish;
- }
+ if (asprintf(&to, "%s/dev/console", dest) < 0)
+ return log_oom();
/* 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");
- r = -errno;
- goto finish;
+ return -errno;
}
if (mount(console, to, "bind", MS_BIND, NULL) < 0) {
log_error("Bind mount for /dev/console failed: %m");
- r = -errno;
- goto finish;
+ return -errno;
}
-finish:
- free(to);
- umask(u);
-
- return r;
+ return 0;
}
static int setup_kmsg(const char *dest, int kmsg_socket) {
- char *from = NULL, *to = NULL;
+ char _cleanup_free_ *from = NULL, *to = NULL;
int r, fd, k;
- mode_t u;
+ mode_t _cleanup_umask_ u;
union {
struct cmsghdr cmsghdr;
uint8_t buf[CMSG_SPACE(sizeof(int))];
* 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");
- r = -errno;
- goto finish;
+ return -errno;
}
r = chmod_and_chown(from, 0600, 0, 0);
if (r < 0) {
log_error("Failed to correct access mode for /dev/kmsg: %s", strerror(-r));
- goto finish;
+ return r;
}
if (mount(from, to, "bind", MS_BIND, NULL) < 0) {
log_error("Bind mount for /proc/kmsg failed: %m");
- r = -errno;
- goto finish;
+ return -errno;
}
fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
if (fd < 0) {
log_error("Failed to open fifo: %m");
- r = -errno;
- goto finish;
+ return -errno;
}
zero(mh);
if (k < 0) {
log_error("Failed to send FIFO fd: %m");
- r = -errno;
- goto finish;
+ return -errno;
}
/* And now make the FIFO unavailable as /dev/kmsg... */
unlink(from);
-
-finish:
- free(from);
- free(to);
- umask(u);
-
- return r;
+ return 0;
}
static int setup_hostname(void) {
static int setup_journal(const char *directory) {
sd_id128_t machine_id;
- 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");
- if (!p) {
- r = log_oom();
- goto finish;
- }
+ if (!p)
+ return log_oom();
r = read_one_line_file(p, &b);
- 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));
return r;
}
- l = strstrip(b);
- if (isempty(l) && arg_link_journal == LINK_AUTO) {
- r = 0;
- goto finish;
- }
+ id = strstrip(b);
+ if (isempty(id) && arg_link_journal == LINK_AUTO)
+ return 0;
- /* Verify validaty */
- r = sd_id128_from_string(l, &machine_id);
+ /* Verify validity */
+ r = sd_id128_from_string(id, &machine_id);
if (r < 0) {
- log_error("Failed to parse machine ID: %s", strerror(-r));
- goto finish;
+ log_error("Failed to parse machine ID from %s: %s", p, strerror(-r));
+ return r;
}
free(p);
- 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) {
if (arg_link_journal != LINK_AUTO) {
- log_error("Journal already a mount point, refusing.");
- r = -EEXIST;
- goto finish;
+ log_error("%s: already a mount point, refusing to use for journal", q);
+ return -EEXIST;
}
- r = 0;
- goto finish;
+ return 0;
}
r = readlink_and_make_absolute(p, &d);
arg_link_journal == LINK_AUTO) &&
path_equal(d, q)) {
- mkdir_p(q, 0755);
-
- r = 0;
- goto finish;
+ r = mkdir_p(q, 0755);
+ if (r < 0)
+ log_warning("failed to create directory %s: %m", q);
+ return 0;
}
if (unlink(p) < 0) {
log_error("Failed to remove symlink %s: %m", p);
- r = -errno;
- goto finish;
+ return -errno;
}
} else if (r == -EINVAL) {
if (arg_link_journal == LINK_GUEST &&
rmdir(p) < 0) {
- if (errno == ENOTDIR)
- log_error("%s already exists and is neither symlink nor directory.", p);
- else {
+ if (errno == ENOTDIR) {
+ log_error("%s already exists and is neither a symlink nor a directory", p);
+ return r;
+ } else {
log_error("Failed to remove %s: %m", p);
- r = -errno;
+ return -errno;
}
-
- goto finish;
}
} else if (r != -ENOENT) {
log_error("readlink(%s) failed: %m", p);
- goto finish;
+ return r;
}
if (arg_link_journal == LINK_GUEST) {
if (symlink(q, p) < 0) {
log_error("Failed to symlink %s to %s: %m", q, p);
- r = -errno;
- goto finish;
+ return -errno;
}
- mkdir_p(q, 0755);
-
- r = 0;
- goto finish;
+ r = mkdir_p(q, 0755);
+ if (r < 0)
+ log_warning("failed to create directory %s: %m", q);
+ return 0;
}
if (arg_link_journal == LINK_HOST) {
r = mkdir_p(p, 0755);
if (r < 0) {
log_error("Failed to create %s: %m", p);
- goto finish;
+ return r;
}
- } else if (access(p, F_OK) < 0) {
- r = 0;
- goto finish;
- }
+ } else if (access(p, F_OK) < 0)
+ return 0;
if (dir_is_empty(q) == 0) {
log_error("%s not empty.", q);
- r = -ENOTEMPTY;
- goto finish;
+ return -ENOTEMPTY;
}
r = mkdir_p(q, 0755);
if (r < 0) {
log_error("Failed to create %s: %m", q);
- goto finish;
+ return r;
}
if (mount(p, q, "bind", MS_BIND, NULL) < 0) {
log_error("Failed to bind mount journal from host into guest: %m");
- r = -errno;
- goto finish;
+ return -errno;
}
- r = 0;
-
-finish:
- free(p);
- free(q);
- free(d);
- free(b);
- return r;
-
+ return 0;
}
static int drop_capabilities(void) {
return r < 0 ? 0 : 1;
}
-static int process_pty(int master, sigset_t *mask) {
+static int process_pty(int master, pid_t pid, sigset_t *mask) {
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;
+ bool tried_orderly_shutdown = false;
+
+ assert(master >= 0);
+ assert(pid > 0);
+ assert(mask);
fd_nonblock(STDIN_FILENO, 1);
fd_nonblock(STDOUT_FILENO, 1);
goto finish;
}
- 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;
+ }
+ }
zero(stdout_ev);
stdout_ev.events = EPOLLOUT|EPOLLET;
signal_ev.events = EPOLLIN;
signal_ev.data.fd = signal_fd;
- if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0 ||
- epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 ||
+ if (epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 ||
epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 ||
epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) {
- log_error("Failed to regiser fds in epoll: %m");
+ log_error("Failed to register fds in epoll: %m");
r = -errno;
goto finish;
}
/* 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);
+
} else {
r = 0;
goto finish;
int r = EXIT_FAILURE, k;
char *oldcg = NULL, *newcg = NULL;
char **controller = NULL;
- int master = -1;
+ int master = -1, n_fd_passed;
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 };
+ FDSet *fds = NULL;
log_parse_environment();
log_open();
goto finish;
}
+ 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));
goto finish;
}
- 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;
- raw_attr = saved_attr;
- cfmakeraw(&raw_attr);
- raw_attr.c_lflag &= ~ECHO;
-
- if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) {
- log_error("Failed to set terminal attributes: %m");
- goto finish;
+ raw_attr = saved_attr;
+ cfmakeraw(&raw_attr);
+ raw_attr.c_lflag &= ~ECHO;
}
if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1);
assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0);
- pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
- if (pid < 0) {
- if (errno == EINVAL)
- log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
- else
- log_error("clone() failed: %m");
-
- goto finish;
- }
-
- if (pid == 0) {
- /* child */
-
- const char *home = NULL;
- uid_t uid = (uid_t) -1;
- gid_t gid = (gid_t) -1;
- 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 */
- NULL, /* TERM */
- NULL, /* HOME */
- NULL, /* USER */
- NULL, /* LOGNAME */
- NULL, /* container_uuid */
- NULL
- };
-
- envp[2] = strv_find_prefix(environ, "TERM=");
-
- close_nointr_nofail(master);
-
- close_nointr(STDIN_FILENO);
- close_nointr(STDOUT_FILENO);
- close_nointr(STDERR_FILENO);
-
- close_all_fds(&kmsg_socket_pair[1], 1);
-
- 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)
- goto child_fail;
+ for (;;) {
+ siginfo_t status;
- if (setsid() < 0) {
- log_error("setsid() failed: %m");
- goto child_fail;
+ if (saved_attr_valid) {
+ if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) {
+ log_error("Failed to set terminal attributes: %m");
+ goto finish;
+ }
}
- if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
- log_error("PR_SET_PDEATHSIG failed: %m");
- goto child_fail;
- }
+ pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
+ if (pid < 0) {
+ if (errno == EINVAL)
+ log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m");
+ else
+ log_error("clone() failed: %m");
- /* Mark everything as slave, so that we still
- * receive mounts from the real root, but don't
- * propagate mounts to the real root. */
- if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
- log_error("MS_SLAVE|MS_REC failed: %m");
- goto child_fail;
+ goto finish;
}
- /* Turn directory into bind mount */
- if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) {
- log_error("Failed to make bind mount.");
- goto child_fail;
- }
+ if (pid == 0) {
+ /* child */
+
+ const char *home = NULL;
+ uid_t uid = (uid_t) -1;
+ gid_t gid = (gid_t) -1;
+ unsigned n_env = 0;
+ 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 */
+ NULL, /* TERM */
+ NULL, /* HOME */
+ NULL, /* USER */
+ NULL, /* LOGNAME */
+ NULL, /* container_uuid */
+ NULL, /* LISTEN_FDS */
+ NULL, /* LISTEN_PID */
+ NULL
+ };
+
+ envp[2] = strv_find_prefix(environ, "TERM=");
+ n_env = 3;
+
+ close_nointr_nofail(master);
+ master = -1;
+
+ close_nointr(STDIN_FILENO);
+ close_nointr(STDOUT_FILENO);
+ close_nointr(STDERR_FILENO);
+
+ close_nointr_nofail(kmsg_socket_pair[0]);
+ kmsg_socket_pair[0] = -1;
+
+ reset_all_signal_handlers();
+
+ assert_se(sigemptyset(&mask) == 0);
+ assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
+
+ k = open_terminal(console, O_RDWR);
+ if (k != STDIN_FILENO) {
+ if (k >= 0) {
+ close_nointr_nofail(k);
+ k = -EINVAL;
+ }
- if (arg_read_only)
- if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) {
- log_error("Failed to make read-only.");
+ log_error("Failed to open console: %s", strerror(-k));
goto child_fail;
}
- if (mount_all(arg_directory) < 0)
- goto child_fail;
-
- if (copy_devnodes(arg_directory) < 0)
- goto child_fail;
-
- dev_setup(arg_directory);
+ if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
+ dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) {
+ log_error("Failed to duplicate console: %m");
+ goto child_fail;
+ }
- if (setup_dev_console(arg_directory, console) < 0)
- goto child_fail;
+ if (setsid() < 0) {
+ log_error("setsid() failed: %m");
+ goto child_fail;
+ }
- if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
- goto child_fail;
+ if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
+ log_error("PR_SET_PDEATHSIG failed: %m");
+ goto child_fail;
+ }
- close_nointr_nofail(kmsg_socket_pair[1]);
+ /* Mark everything as slave, so that we still
+ * receive mounts from the real root, but don't
+ * propagate mounts to the real root. */
+ if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) {
+ log_error("MS_SLAVE|MS_REC failed: %m");
+ goto child_fail;
+ }
- if (setup_timezone(arg_directory) < 0)
- goto child_fail;
+ /* Turn directory into bind mount */
+ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) {
+ log_error("Failed to make bind mount.");
+ goto child_fail;
+ }
- if (setup_resolv_conf(arg_directory) < 0)
- goto child_fail;
+ if (arg_read_only)
+ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) {
+ log_error("Failed to make read-only.");
+ goto child_fail;
+ }
- if (setup_journal(arg_directory) < 0)
- goto child_fail;
+ if (mount_all(arg_directory) < 0)
+ goto child_fail;
- if (chdir(arg_directory) < 0) {
- log_error("chdir(%s) failed: %m", arg_directory);
- goto child_fail;
- }
+ if (copy_devnodes(arg_directory) < 0)
+ goto child_fail;
- if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
- log_error("mount(MS_MOVE) failed: %m");
- goto child_fail;
- }
+ dev_setup(arg_directory);
- if (chroot(".") < 0) {
- log_error("chroot() failed: %m");
- goto child_fail;
- }
+ if (setup_dev_console(arg_directory, console) < 0)
+ goto child_fail;
- if (chdir("/") < 0) {
- log_error("chdir() failed: %m");
- goto child_fail;
- }
+ if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
+ goto child_fail;
- umask(0022);
+ close_nointr_nofail(kmsg_socket_pair[1]);
+ kmsg_socket_pair[1] = -1;
- loopback_setup();
+ if (setup_boot_id(arg_directory) < 0)
+ goto child_fail;
- if (drop_capabilities() < 0) {
- log_error("drop_capabilities() failed: %m");
- goto child_fail;
- }
+ if (setup_timezone(arg_directory) < 0)
+ goto child_fail;
- if (arg_user) {
+ if (setup_resolv_conf(arg_directory) < 0)
+ goto child_fail;
- if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) {
- log_error("get_user_creds() failed: %m");
+ if (setup_journal(arg_directory) < 0)
goto child_fail;
- }
- if (mkdir_parents_label(home, 0775) < 0) {
- log_error("mkdir_parents_label() failed: %m");
+ if (chdir(arg_directory) < 0) {
+ log_error("chdir(%s) failed: %m", arg_directory);
goto child_fail;
}
- if (mkdir_safe_label(home, 0775, uid, gid) < 0) {
- log_error("mkdir_safe_label() failed: %m");
+ if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
+ log_error("mount(MS_MOVE) failed: %m");
goto child_fail;
}
- if (initgroups((const char*)arg_user, gid) < 0) {
- log_error("initgroups() failed: %m");
+ if (chroot(".") < 0) {
+ log_error("chroot() failed: %m");
goto child_fail;
}
- if (setresgid(gid, gid, gid) < 0) {
- log_error("setregid() failed: %m");
+ if (chdir("/") < 0) {
+ log_error("chdir() failed: %m");
goto child_fail;
}
- if (setresuid(uid, uid, uid) < 0) {
- log_error("setreuid() failed: %m");
+ umask(0022);
+
+ loopback_setup();
+
+ if (drop_capabilities() < 0) {
+ log_error("drop_capabilities() 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)) {
- log_oom();
- goto child_fail;
- }
+ if (arg_user) {
+
+ /* 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 (mkdir_parents_label(home, 0775) < 0) {
+ log_error("mkdir_parents_label() failed: %m");
+ goto child_fail;
+ }
- if (arg_uuid) {
- if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) {
+ if (mkdir_safe_label(home, 0775, uid, gid) < 0) {
+ log_error("mkdir_safe_label() failed: %m");
+ goto child_fail;
+ }
+
+ if (initgroups((const char*)arg_user, gid) < 0) {
+ log_error("initgroups() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresgid(gid, gid, gid) < 0) {
+ log_error("setregid() failed: %m");
+ goto child_fail;
+ }
+
+ if (setresuid(uid, uid, uid) < 0) {
+ log_error("setreuid() 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 + 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)) {
log_oom();
goto child_fail;
}
- }
- setup_hostname();
+ if (arg_uuid) {
+ 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)) {
+ log_oom();
+ goto child_fail;
+ }
+ }
- if (arg_boot) {
- char **a;
- size_t l;
+ setup_hostname();
- /* Automatically search for the init system */
+ if (arg_boot) {
+ char **a;
+ size_t l;
- l = 1 + argc - optind;
- a = newa(char*, l + 1);
- memcpy(a + 1, argv + optind, l * sizeof(char*));
+ /* Automatically search for the init system */
- a[0] = (char*) "/usr/lib/systemd/systemd";
- execve(a[0], a, (char**) envp);
+ l = 1 + argc - optind;
+ a = newa(char*, l + 1);
+ memcpy(a + 1, argv + optind, l * sizeof(char*));
- a[0] = (char*) "/lib/systemd/systemd";
- execve(a[0], a, (char**) envp);
+ a[0] = (char*) "/usr/lib/systemd/systemd";
+ execve(a[0], a, (char**) envp);
- a[0] = (char*) "/sbin/init";
- execve(a[0], a, (char**) envp);
- } else if (argc > optind)
- execvpe(argv[optind], argv + optind, (char**) envp);
- else {
- chdir(home ? home : "/root");
- execle("/bin/bash", "-bash", NULL, (char**) envp);
+ a[0] = (char*) "/lib/systemd/systemd";
+ execve(a[0], a, (char**) envp);
+
+ a[0] = (char*) "/sbin/init";
+ execve(a[0], a, (char**) envp);
+ } else if (argc > optind)
+ execvpe(argv[optind], argv + optind, (char**) envp);
+ else {
+ chdir(home ? home : "/root");
+ execle("/bin/bash", "-bash", NULL, (char**) envp);
+ }
+
+ log_error("execv() failed: %m");
+
+ child_fail:
+ _exit(EXIT_FAILURE);
}
- log_error("execv() failed: %m");
+ fdset_free(fds);
+ fds = NULL;
- child_fail:
- _exit(EXIT_FAILURE);
- }
+ if (process_pty(master, pid, &mask) < 0)
+ goto finish;
- if (process_pty(master, &mask) < 0)
- goto finish;
+ if (saved_attr_valid)
+ tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
- if (saved_attr_valid) {
- tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr);
- saved_attr_valid = false;
- }
+ r = wait_for_terminate(pid, &status);
+ if (r < 0) {
+ r = EXIT_FAILURE;
+ break;
+ }
+
+ if (status.si_code == CLD_EXITED) {
+ if (status.si_status != 0) {
+ log_error("Container failed with error code %i.", status.si_status);
+ r = status.si_status;
+ break;
+ }
- r = wait_for_terminate_and_warn(argc > optind ? argv[optind] : "bash", pid);
+ log_debug("Container exited successfully.");
+ break;
+ } else if (status.si_code == CLD_KILLED &&
+ status.si_status == SIGINT) {
+ log_info("Container has been shut down.");
+ r = 0;
+ break;
+ } else if (status.si_code == CLD_KILLED &&
+ status.si_status == SIGHUP) {
+ log_info("Container is being rebooted.");
+ continue;
+ } else if (status.si_code == CLD_KILLED ||
+ status.si_code == CLD_DUMPED) {
- if (r < 0)
- r = EXIT_FAILURE;
+ log_error("Container terminated by signal %s.", signal_to_string(status.si_status));
+ r = EXIT_FAILURE;
+ break;
+ } else {
+ log_error("Container failed due to unknown reason.");
+ r = EXIT_FAILURE;
+ break;
+ }
+ }
finish:
if (saved_attr_valid)
free(oldcg);
free(newcg);
+ fdset_free(fds);
+
return r;
}