chiark / gitweb /
nspawn: if a file system comes pre-mounted, still do the read-only remounts
[elogind.git] / src / nspawn / nspawn.c
index ce727b6e1d8b3d663855a19a69fac4e52c8e2d2b..ed72c3ebf1cffc43fae0455694abda1038f76510 100644 (file)
 #include "missing.h"
 #include "cgroup-util.h"
 #include "strv.h"
+#include "path-util.h"
 #include "loopback-setup.h"
+#include "sd-id128.h"
+#include "dev-setup.h"
+
+typedef enum LinkJournal {
+        LINK_NO,
+        LINK_AUTO,
+        LINK_HOST,
+        LINK_GUEST
+} LinkJournal;
 
 static char *arg_directory = NULL;
 static char *arg_user = NULL;
 static char **arg_controllers = NULL;
+static char *arg_uuid = NULL;
 static bool arg_private_network = false;
+static bool arg_read_only = false;
+static bool arg_boot = false;
+static LinkJournal arg_link_journal = LINK_AUTO;
+static uint64_t arg_retain =
+        (1ULL << CAP_CHOWN) |
+        (1ULL << CAP_DAC_OVERRIDE) |
+        (1ULL << CAP_DAC_READ_SEARCH) |
+        (1ULL << CAP_FOWNER) |
+        (1ULL << CAP_FSETID) |
+        (1ULL << CAP_IPC_OWNER) |
+        (1ULL << CAP_KILL) |
+        (1ULL << CAP_LEASE) |
+        (1ULL << CAP_LINUX_IMMUTABLE) |
+        (1ULL << CAP_NET_BIND_SERVICE) |
+        (1ULL << CAP_NET_BROADCAST) |
+        (1ULL << CAP_NET_RAW) |
+        (1ULL << CAP_SETGID) |
+        (1ULL << CAP_SETFCAP) |
+        (1ULL << CAP_SETPCAP) |
+        (1ULL << CAP_SETUID) |
+        (1ULL << CAP_SYS_ADMIN) |
+        (1ULL << CAP_SYS_CHROOT) |
+        (1ULL << CAP_SYS_NICE) |
+        (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"
-               "  -D --directory=NAME   Root directory for the container\n"
-               "  -u --user=USER        Run the command under specified user or uid\n"
-               "  -C --controllers=LIST Put the container in specified comma-separated cgroup hierarchies\n"
-               "     --private-network  Disable network in container\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"
+               "  -C --controllers=LIST   Put the container in specified comma-separated cgroup hierarchies\n"
+               "     --uuid=UUID          Set a specific machine UUID for the container\n"
+               "     --private-network    Disable network in container\n"
+               "     --read-only          Mount the root directory read-only\n"
+               "     --capability=CAP     In addition to the default, retain specified capability\n"
+               "     --link-journal=MODE  Link up guest journal, one of no, auto, guest, host\n"
+               "  -j                      Equivalent to --link-journal=host\n",
                program_invocation_short_name);
 
         return 0;
@@ -74,7 +117,11 @@ static int help(void) {
 static int parse_argv(int argc, char *argv[]) {
 
         enum {
-                ARG_PRIVATE_NETWORK = 0x100
+                ARG_PRIVATE_NETWORK = 0x100,
+                ARG_UUID,
+                ARG_READ_ONLY,
+                ARG_CAPABILITY,
+                ARG_LINK_JOURNAL
         };
 
         static const struct option options[] = {
@@ -83,6 +130,11 @@ static int parse_argv(int argc, char *argv[]) {
                 { "user",            required_argument, NULL, 'u'                 },
                 { "controllers",     required_argument, NULL, 'C'                 },
                 { "private-network", no_argument,       NULL, ARG_PRIVATE_NETWORK },
+                { "boot",            no_argument,       NULL, 'b'                 },
+                { "uuid",            required_argument, NULL, ARG_UUID            },
+                { "read-only",       no_argument,       NULL, ARG_READ_ONLY       },
+                { "capability",      required_argument, NULL, ARG_CAPABILITY      },
+                { "link-journal",    required_argument, NULL, ARG_LINK_JOURNAL    },
                 { NULL,              0,                 NULL, 0                   }
         };
 
@@ -91,7 +143,7 @@ static int parse_argv(int argc, char *argv[]) {
         assert(argc >= 0);
         assert(argv);
 
-        while ((c = getopt_long(argc, argv, "+hD:u:C:", options, NULL)) >= 0) {
+        while ((c = getopt_long(argc, argv, "+hD:u:C:bj", options, NULL)) >= 0) {
 
                 switch (c) {
 
@@ -101,8 +153,9 @@ static int parse_argv(int argc, char *argv[]) {
 
                 case 'D':
                         free(arg_directory);
-                        if (!(arg_directory = strdup(optarg))) {
-                                log_error("Failed to duplicate root directory.");
+                        arg_directory = canonicalize_file_name(optarg);
+                        if (!arg_directory) {
+                                log_error("Failed to canonicalize root directory.");
                                 return -ENOMEM;
                         }
 
@@ -132,6 +185,63 @@ static int parse_argv(int argc, char *argv[]) {
                         arg_private_network = true;
                         break;
 
+                case 'b':
+                        arg_boot = true;
+                        break;
+
+                case ARG_UUID:
+                        arg_uuid = optarg;
+                        break;
+
+                case ARG_READ_ONLY:
+                        arg_read_only = true;
+                        break;
+
+                case ARG_CAPABILITY: {
+                        char *state, *word;
+                        size_t length;
+
+                        FOREACH_WORD_SEPARATOR(word, length, optarg, ",", state) {
+                                cap_value_t cap;
+                                char *t;
+
+                                t = strndup(word, length);
+                                if (!t)
+                                        return log_oom();
+
+                                if (cap_from_name(t, &cap) < 0) {
+                                        log_error("Failed to parse capability %s.", t);
+                                        free(t);
+                                        return -EINVAL;
+                                }
+
+                                free(t);
+                                arg_retain |= 1ULL << (uint64_t) cap;
+                        }
+
+                        break;
+                }
+
+                case 'j':
+                        arg_link_journal = LINK_GUEST;
+                        break;
+
+                case ARG_LINK_JOURNAL:
+                        if (streq(optarg, "auto"))
+                                arg_link_journal = LINK_AUTO;
+                        else if (streq(optarg, "no"))
+                                arg_link_journal = LINK_NO;
+                        else if (streq(optarg, "guest"))
+                                arg_link_journal = LINK_GUEST;
+                        else if (streq(optarg, "host"))
+                                arg_link_journal = LINK_HOST;
+                        else {
+                                log_error("Failed to parse link journal mode %s", optarg);
+                                return -EINVAL;
+                        }
+
+                        break;
+
                 case '?':
                         return -EINVAL;
 
@@ -157,16 +267,15 @@ static int mount_all(const char *dest) {
 
         static const MountPoint mount_table[] = {
                 { "proc",      "/proc",     "proc",  NULL,       MS_NOSUID|MS_NOEXEC|MS_NODEV, true  },
-                { "/proc/sys", "/proc/sys", "bind",  NULL,       MS_BIND, true                       },   /* Bind mount first */
-                { "/proc/sys", "/proc/sys", "bind",  NULL,       MS_BIND|MS_RDONLY|MS_REMOUNT, true  },   /* Then, make it r/o */
-                { "/sys",      "/sys",      "bind",  NULL,       MS_BIND,                      true  },   /* Bind mount first */
-                { "/sys",      "/sys",      "bind",  NULL,       MS_BIND|MS_RDONLY|MS_REMOUNT, true  },   /* Then, make it r/o */
+                { "/proc/sys", "/proc/sys", NULL,    NULL,       MS_BIND, true                       },   /* Bind mount first */
+                { NULL,        "/proc/sys", NULL,    NULL,       MS_BIND|MS_RDONLY|MS_REMOUNT, true  },   /* Then, make it r/o */
+                { "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",  "bind",  NULL,       MS_BIND,                      true  },
+                { "/dev/pts",  "/dev/pts",  NULL,    NULL,       MS_BIND,                      true  },
                 { "tmpfs",     "/run",      "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true  },
 #ifdef HAVE_SELINUX
-                { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND,                      false },  /* Bind mount first */
-                { "/sys/fs/selinux", "/sys/fs/selinux", "bind", NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false },  /* Then, make it r/o */
+                { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND,                      false },  /* Bind mount first */
+                { NULL,              "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false },  /* Then, make it r/o */
 #endif
         };
 
@@ -178,7 +287,7 @@ static int mount_all(const char *dest) {
                 int t;
 
                 if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) {
-                        log_error("Out of memory");
+                        log_oom();
 
                         if (r == 0)
                                 r = -ENOMEM;
@@ -186,7 +295,7 @@ static int mount_all(const char *dest) {
                         break;
                 }
 
-                t = path_is_mount_point(where, false);
+                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);
@@ -197,7 +306,11 @@ static int mount_all(const char *dest) {
                         continue;
                 }
 
-                mkdir_p(where, 0755);
+                /* Skip this entry if it is not a remount. */
+                if (mount_table[k].what && t > 0)
+                        continue;
+
+                mkdir_p_label(where, 0755);
 
                 if (mount(mount_table[k].what,
                           where,
@@ -224,20 +337,16 @@ static int setup_timezone(const char *dest) {
         assert(dest);
 
         /* Fix the timezone, if possible */
-        if (asprintf(&where, "%s/etc/localtime", dest) < 0) {
-                log_error("Out of memory");
-                return -ENOMEM;
-        }
+        if (asprintf(&where, "%s/etc/localtime", dest) < 0)
+                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);
 
         free(where);
 
-        if (asprintf(&where, "%s/etc/timezone", dest) < 0) {
-                log_error("Out of memory");
-                return -ENOMEM;
-        }
+        if (asprintf(&where, "%s/etc/timezone", dest) < 0)
+                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);
@@ -247,6 +356,27 @@ static int setup_timezone(const char *dest) {
         return 0;
 }
 
+static int setup_resolv_conf(const char *dest) {
+        char *where;
+
+        assert(dest);
+
+        if (arg_private_network)
+                return 0;
+
+        /* Fix resolv.conf, if possible */
+        if (asprintf(&where, "%s/etc/resolv.conf", dest) < 0) {
+                return log_oom();
+        }
+
+        if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0)
+                mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL);
+
+        free(where);
+
+        return 0;
+}
+
 static int copy_devnodes(const char *dest) {
 
         static const char devnodes[] =
@@ -347,8 +477,7 @@ static int setup_dev_console(const char *dest, const char *console) {
         }
 
         if (asprintf(&to, "%s/dev/console", dest) < 0) {
-                log_error("Out of memory");
-                r = -ENOMEM;
+                r = log_oom();
                 goto finish;
         }
 
@@ -394,15 +523,20 @@ static int setup_kmsg(const char *dest, int kmsg_socket) {
 
         u = umask(0000);
 
+        /* We create the kmsg FIFO as /dev/kmsg, but immediately
+         * delete it after bind mounting it to /proc/kmsg. While FIFOs
+         * on the reading side behave very similar to /proc/kmsg,
+         * their writing side behaves differently from /dev/kmsg in
+         * 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) {
-                log_error("Out of memory");
-                r = -ENOMEM;
+                r = log_oom();
                 goto finish;
         }
 
         if (asprintf(&to, "%s/proc/kmsg", dest) < 0) {
-                log_error("Out of memory");
-                r = -ENOMEM;
+                r = log_oom();
                 goto finish;
         }
 
@@ -456,6 +590,9 @@ static int setup_kmsg(const char *dest, int kmsg_socket) {
                 goto finish;
         }
 
+        /* And now make the FIFO unavailable as /dev/kmsg... */
+        unlink(from);
+
 finish:
         free(from);
         free(to);
@@ -464,50 +601,177 @@ finish:
         return r;
 }
 
-static int drop_capabilities(void) {
-        static const unsigned long retain[] = {
-                CAP_CHOWN,
-                CAP_DAC_OVERRIDE,
-                CAP_DAC_READ_SEARCH,
-                CAP_FOWNER,
-                CAP_FSETID,
-                CAP_IPC_OWNER,
-                CAP_KILL,
-                CAP_LEASE,
-                CAP_LINUX_IMMUTABLE,
-                CAP_NET_BIND_SERVICE,
-                CAP_NET_BROADCAST,
-                CAP_NET_RAW,
-                CAP_SETGID,
-                CAP_SETFCAP,
-                CAP_SETPCAP,
-                CAP_SETUID,
-                CAP_SYS_ADMIN,
-                CAP_SYS_CHROOT,
-                CAP_SYS_NICE,
-                CAP_SYS_PTRACE,
-                CAP_SYS_TTY_CONFIG
-        };
+static int setup_hostname(void) {
+        char *hn;
+        int r = 0;
 
-        unsigned long l;
+        hn = path_get_file_name(arg_directory);
+        if (hn) {
+                hn = strdup(hn);
+                if (!hn)
+                        return -ENOMEM;
 
-        for (l = 0; l <= cap_last_cap(); l++) {
-                unsigned i;
+                hostname_cleanup(hn);
 
-                for (i = 0; i < ELEMENTSOF(retain); i++)
-                        if (retain[i] == l)
-                                break;
+                if (!isempty(hn))
+                        if (sethostname(hn, strlen(hn)) < 0)
+                                r = -errno;
 
-                if (i < ELEMENTSOF(retain))
-                        continue;
+                free(hn);
+        }
+
+        return r;
+}
+
+static int setup_journal(const char *directory) {
+        sd_id128_t machine_id;
+        char *p = NULL, *b = NULL, *l, *q = NULL, *d = NULL;
+        int r;
+
+        if (arg_link_journal == LINK_NO)
+                return 0;
+
+        p = strappend(directory, "/etc/machine-id");
+        if (!p) {
+                r = log_oom();
+                goto finish;
+        }
+
+        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));
+                return r;
+        }
 
-                if (prctl(PR_CAPBSET_DROP, l) < 0) {
-                        log_error("PR_CAPBSET_DROP failed: %m");
-                        return -errno;
+        l = strstrip(b);
+        if (isempty(l) && arg_link_journal == LINK_AUTO) {
+                r = 0;
+                goto finish;
+        }
+
+        /* Verify validaty */
+        r = sd_id128_from_string(l, &machine_id);
+        if (r < 0) {
+                log_error("Failed to parse machine ID: %s", strerror(-r));
+                goto finish;
+        }
+
+        free(p);
+        p = strappend("/var/log/journal/", l);
+        q = strjoin(directory, "/var/log/journal/", l, NULL);
+        if (!p || !q) {
+                r = log_oom();
+                goto finish;
+        }
+
+        if (path_is_mount_point(p, false) > 0 ||
+            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;
                 }
+
+                r = 0;
+                goto finish;
         }
 
-        return 0;
+        r = readlink_and_make_absolute(p, &d);
+        if (r >= 0) {
+                if ((arg_link_journal == LINK_GUEST ||
+                     arg_link_journal == LINK_AUTO) &&
+                    path_equal(d, q)) {
+
+                        mkdir_p(q, 0755);
+
+                        r = 0;
+                        goto finish;
+                }
+
+                if (unlink(p) < 0) {
+                        log_error("Failed to remove symlink %s: %m", p);
+                        r = -errno;
+                        goto finish;
+                }
+        } 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 {
+                                log_error("Failed to remove %s: %m", p);
+                                r = -errno;
+                        }
+
+                        goto finish;
+                }
+        } else if (r != -ENOENT) {
+                log_error("readlink(%s) failed: %m", p);
+                goto finish;
+        }
+
+        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;
+                }
+
+                mkdir_p(q, 0755);
+
+                r = 0;
+                goto finish;
+        }
+
+        if (arg_link_journal == LINK_HOST) {
+                r = mkdir_p(p, 0755);
+                if (r < 0) {
+                        log_error("Failed to create %s: %m", p);
+                        goto finish;
+                }
+
+        } else if (access(p, F_OK) < 0) {
+                r = 0;
+                goto finish;
+        }
+
+        if (dir_is_empty(q) == 0) {
+                log_error("%s not empty.", q);
+                r = -ENOTEMPTY;
+                goto finish;
+        }
+
+        r = mkdir_p(q, 0755);
+        if (r < 0) {
+                log_error("Failed to create %s: %m", q);
+                goto finish;
+        }
+
+        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;
+        }
+
+        r = 0;
+
+finish:
+        free(p);
+        free(q);
+        free(d);
+        free(b);
+        return r;
+
+}
+
+static int drop_capabilities(void) {
+        return capability_bounding_set_drop(~arg_retain, false);
 }
 
 static int is_os_tree(const char *path) {
@@ -536,13 +800,15 @@ static int process_pty(int master, sigset_t *mask) {
         fd_nonblock(STDOUT_FILENO, 1);
         fd_nonblock(master, 1);
 
-        if ((signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC)) < 0) {
+        signal_fd = signalfd(-1, mask, SFD_NONBLOCK|SFD_CLOEXEC);
+        if (signal_fd < 0) {
                 log_error("signalfd(): %m");
                 r = -errno;
                 goto finish;
         }
 
-        if ((ep = epoll_create1(EPOLL_CLOEXEC)) < 0) {
+        ep = epoll_create1(EPOLL_CLOEXEC);
+        if (ep < 0) {
                 log_error("Failed to create epoll: %m");
                 r = -errno;
                 goto finish;
@@ -578,7 +844,8 @@ static int process_pty(int master, sigset_t *mask) {
                 ssize_t k;
                 int i, nfds;
 
-                if ((nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1)) < 0) {
+                nfds = epoll_wait(ep, ev, ELEMENTSOF(ev), -1);
+                if (nfds < 0) {
 
                         if (errno == EINTR || errno == EAGAIN)
                                 continue;
@@ -613,7 +880,8 @@ static int process_pty(int master, sigset_t *mask) {
                                 struct signalfd_siginfo sfsi;
                                 ssize_t n;
 
-                                if ((n = read(signal_fd, &sfsi, sizeof(sfsi))) != sizeof(sfsi)) {
+                                n = read(signal_fd, &sfsi, sizeof(sfsi));
+                                if (n != sizeof(sfsi)) {
 
                                         if (n >= 0) {
                                                 log_error("Failed to read from signalfd: invalid block size");
@@ -649,7 +917,8 @@ static int process_pty(int master, sigset_t *mask) {
 
                         if (stdin_readable && in_buffer_full < LINE_MAX) {
 
-                                if ((k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full)) < 0) {
+                                k = read(STDIN_FILENO, in_buffer + in_buffer_full, LINE_MAX - in_buffer_full);
+                                if (k < 0) {
 
                                         if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
                                                 stdin_readable = false;
@@ -664,7 +933,8 @@ static int process_pty(int master, sigset_t *mask) {
 
                         if (master_writable && in_buffer_full > 0) {
 
-                                if ((k = write(master, in_buffer, in_buffer_full)) < 0) {
+                                k = write(master, in_buffer, in_buffer_full);
+                                if (k < 0) {
 
                                         if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
                                                 master_writable = false;
@@ -683,7 +953,8 @@ static int process_pty(int master, sigset_t *mask) {
 
                         if (master_readable && out_buffer_full < LINE_MAX) {
 
-                                if ((k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full)) < 0) {
+                                k = read(master, out_buffer + out_buffer_full, LINE_MAX - out_buffer_full);
+                                if (k < 0) {
 
                                         if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
                                                 master_readable = false;
@@ -698,7 +969,8 @@ static int process_pty(int master, sigset_t *mask) {
 
                         if (stdout_writable && out_buffer_full > 0) {
 
-                                if ((k = write(STDOUT_FILENO, out_buffer, out_buffer_full)) < 0) {
+                                k = write(STDOUT_FILENO, out_buffer, out_buffer_full);
+                                if (k < 0) {
 
                                         if (errno == EAGAIN || errno == EPIPE || errno == ECONNRESET || errno == EIO)
                                                 stdout_writable = false;
@@ -743,7 +1015,8 @@ int main(int argc, char *argv[]) {
         log_parse_environment();
         log_open();
 
-        if ((r = parse_argv(argc, argv)) <= 0)
+        r = parse_argv(argc, argv);
+        if (r <= 0)
                 goto finish;
 
         if (arg_directory) {
@@ -782,7 +1055,8 @@ int main(int argc, char *argv[]) {
                 goto finish;
         }
 
-        if ((k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg)) < 0) {
+        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;
         }
@@ -798,18 +1072,20 @@ int main(int argc, char *argv[]) {
                 goto finish;
         }
 
-        STRV_FOREACH(controller,arg_controllers) {
+        STRV_FOREACH(controller, arg_controllers) {
                 k = cg_create_and_attach(*controller, newcg, 0);
                 if (k < 0)
                         log_warning("Failed to create cgroup in controller %s: %s", *controller, strerror(-k));
         }
 
-        if ((master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY)) < 0) {
+        master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY);
+        if (master < 0) {
                 log_error("Failed to acquire pseudo tty: %m");
                 goto finish;
         }
 
-        if (!(console = ptsname(master))) {
+        console = ptsname(master);
+        if (!console) {
                 log_error("Failed to determine tty name: %m");
                 goto finish;
         }
@@ -862,7 +1138,6 @@ int main(int argc, char *argv[]) {
         if (pid == 0) {
                 /* child */
 
-                const char *hn;
                 const char *home = NULL;
                 uid_t uid = (uid_t) -1;
                 gid_t gid = (gid_t) -1;
@@ -873,6 +1148,7 @@ int main(int argc, char *argv[]) {
                         NULL, /* HOME */
                         NULL, /* USER */
                         NULL, /* LOGNAME */
+                        NULL, /* container_uuid */
                         NULL
                 };
 
@@ -891,15 +1167,40 @@ int main(int argc, char *argv[]) {
                 assert_se(sigemptyset(&mask) == 0);
                 assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
 
-                if (setsid() < 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;
+
+                if (setsid() < 0) {
+                        log_error("setsid() failed: %m");
+                        goto child_fail;
+                }
+
+                if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) {
+                        log_error("PR_SET_PDEATHSIG failed: %m");
                         goto child_fail;
+                }
 
-                if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0)
+                /* 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;
+                }
 
-                /* Mark / as private, in case somebody marked it shared */
-                if (mount(NULL, "/", NULL, MS_PRIVATE|MS_REC, NULL) < 0)
+                /* 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 (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 (mount_all(arg_directory) < 0)
                         goto child_fail;
@@ -907,6 +1208,8 @@ int main(int argc, char *argv[]) {
                 if (copy_devnodes(arg_directory) < 0)
                         goto child_fail;
 
+                dev_setup(arg_directory);
+
                 if (setup_dev_console(arg_directory, console) < 0)
                         goto child_fail;
 
@@ -918,17 +1221,18 @@ int main(int argc, char *argv[]) {
                 if (setup_timezone(arg_directory) < 0)
                         goto child_fail;
 
+                if (setup_resolv_conf(arg_directory) < 0)
+                        goto child_fail;
+
+                if (setup_journal(arg_directory) < 0)
+                        goto child_fail;
+
                 if (chdir(arg_directory) < 0) {
                         log_error("chdir(%s) failed: %m", arg_directory);
                         goto child_fail;
                 }
 
-                if (open_terminal("dev/console", O_RDWR) != STDIN_FILENO ||
-                    dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO ||
-                    dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO)
-                        goto child_fail;
-
-                if (mount(arg_directory, "/", "bind", MS_BIND|MS_MOVE, NULL) < 0) {
+                if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) {
                         log_error("mount(MS_MOVE) failed: %m");
                         goto child_fail;
                 }
@@ -947,23 +1251,25 @@ int main(int argc, char *argv[]) {
 
                 loopback_setup();
 
-                if (drop_capabilities() < 0)
+                if (drop_capabilities() < 0) {
+                        log_error("drop_capabilities() failed: %m");
                         goto child_fail;
+                }
 
                 if (arg_user) {
 
-                        if (get_user_creds((const char**)&arg_user, &uid, &gid, &home) < 0) {
+                        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(home, 0775) < 0) {
-                                log_error("mkdir_parents() failed: %m");
+                        if (mkdir_parents_label(home, 0775) < 0) {
+                                log_error("mkdir_parents_label() failed: %m");
                                 goto child_fail;
                         }
 
-                        if (safe_mkdir(home, 0775, uid, gid) < 0) {
-                                log_error("safe_mkdir() failed: %m");
+                        if (mkdir_safe_label(home, 0775, uid, gid) < 0) {
+                                log_error("mkdir_safe_label() failed: %m");
                                 goto child_fail;
                         }
 
@@ -983,17 +1289,41 @@ int main(int argc, char *argv[]) {
                         }
                 }
 
-                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_error("Out of memory");
+                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 ((hn = file_name_from_path(arg_directory)))
-                        sethostname(hn, strlen(hn));
+                if (arg_uuid) {
+                        if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) {
+                                log_oom();
+                                goto child_fail;
+                        }
+                }
+
+                setup_hostname();
+
+                if (arg_boot) {
+                        char **a;
+                        size_t l;
+
+                        /* Automatically search for the init system */
+
+                        l = 1 + argc - optind;
+                        a = newa(char*, l + 1);
+                        memcpy(a + 1, argv + optind, l * sizeof(char*));
+
+                        a[0] = (char*) "/usr/lib/systemd/systemd";
+                        execve(a[0], a, (char**) envp);
+
+                        a[0] = (char*) "/lib/systemd/systemd";
+                        execve(a[0], a, (char**) envp);
 
-                if (argc > optind)
+                        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");