along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/
-#include <signal.h>
-#include <sched.h>
-#include <unistd.h>
-#include <sys/types.h>
-#include <sys/mount.h>
-#include <stdlib.h>
-#include <string.h>
-#include <stdio.h>
+#ifdef HAVE_BLKID
+#include <blkid/blkid.h>
+#endif
#include <errno.h>
-#include <sys/prctl.h>
#include <getopt.h>
-#include <grp.h>
-#include <linux/fs.h>
-#include <sys/socket.h>
-#include <linux/netlink.h>
-#include <net/if.h>
-#include <linux/veth.h>
-#include <sys/personality.h>
#include <linux/loop.h>
-#include <sys/file.h>
-
-#ifdef HAVE_SELINUX
-#include <selinux/selinux.h>
-#endif
-
+#include <sched.h>
#ifdef HAVE_SECCOMP
#include <seccomp.h>
#endif
-
-#ifdef HAVE_BLKID
-#include <blkid/blkid.h>
+#ifdef HAVE_SELINUX
+#include <selinux/selinux.h>
#endif
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/file.h>
+#include <sys/mount.h>
+#include <sys/personality.h>
+#include <sys/prctl.h>
+#include <sys/types.h>
+#include <unistd.h>
#include "sd-daemon.h"
-#include "sd-bus.h"
#include "sd-id128.h"
-#include "sd-netlink.h"
-#include "random-util.h"
-#include "log.h"
-#include "util.h"
-#include "mkdir.h"
-#include "rm-rf.h"
-#include "macro.h"
-#include "missing.h"
+
+#include "barrier.h"
+#include "base-filesystem.h"
+#include "blkid-util.h"
+#include "btrfs-util.h"
+#include "cap-list.h"
+#include "capability.h"
#include "cgroup-util.h"
-#include "strv.h"
-#include "path-util.h"
-#include "loopback-setup.h"
+#include "copy.h"
#include "dev-setup.h"
+#include "env-util.h"
+#include "event-util.h"
+#include "fd-util.h"
#include "fdset.h"
-#include "build.h"
#include "fileio.h"
-#include "bus-util.h"
-#include "bus-error.h"
-#include "ptyfwd.h"
-#include "env-util.h"
-#include "netlink-util.h"
-#include "udev-util.h"
-#include "blkid-util.h"
+#include "formats-util.h"
#include "gpt.h"
-#include "siphash24.h"
-#include "copy.h"
-#include "base-filesystem.h"
-#include "barrier.h"
-#include "event-util.h"
-#include "capability.h"
-#include "cap-list.h"
-#include "btrfs-util.h"
+#include "hostname-util.h"
+#include "log.h"
+#include "loopback-setup.h"
#include "machine-image.h"
-#include "list.h"
-#include "in-addr-util.h"
-#include "firewall-util.h"
-#include "local-addresses.h"
-#include "formats-util.h"
+#include "macro.h"
+#include "missing.h"
+#include "mkdir.h"
+#include "netlink-util.h"
+#include "nspawn-cgroup.h"
+#include "nspawn-expose-ports.h"
+#include "nspawn-mount.h"
+#include "nspawn-network.h"
+#include "nspawn-register.h"
+#include "nspawn-settings.h"
+#include "nspawn-setuid.h"
+#include "parse-util.h"
+#include "path-util.h"
#include "process-util.h"
-#include "terminal-util.h"
-#include "hostname-util.h"
-#include "signal-util.h"
-
+#include "ptyfwd.h"
+#include "random-util.h"
+#include "rm-rf.h"
#ifdef HAVE_SECCOMP
#include "seccomp-util.h"
#endif
-
-#include "nspawn.h"
-#include "nspawn-settings.h"
+#include "signal-util.h"
+#include "socket-util.h"
+#include "string-util.h"
+#include "strv.h"
+#include "terminal-util.h"
+#include "udev-util.h"
+#include "user-util.h"
+#include "util.h"
typedef enum ContainerStatus {
CONTAINER_TERMINATED,
, program_invocation_short_name);
}
-static CustomMount* custom_mount_add(CustomMount **l, unsigned *n, CustomMountType t) {
- CustomMount *c, *ret;
-
- assert(l);
- assert(n);
- assert(t >= 0);
- assert(t < _CUSTOM_MOUNT_TYPE_MAX);
-
- c = realloc(*l, (*n + 1) * sizeof(CustomMount));
- if (!c)
- return NULL;
-
- *l = c;
- ret = *l + *n;
- (*n)++;
-
- *ret = (CustomMount) { .type = t };
-
- return ret;
-}
-
-void custom_mount_free_all(CustomMount *l, unsigned n) {
- unsigned i;
-
- for (i = 0; i < n; i++) {
- CustomMount *m = l + i;
-
- free(m->source);
- free(m->destination);
- free(m->options);
-
- if (m->work_dir) {
- (void) rm_rf(m->work_dir, REMOVE_ROOT|REMOVE_PHYSICAL);
- free(m->work_dir);
- }
-
- strv_free(m->lower);
- }
-
- free(l);
-}
-
-static int custom_mount_compare(const void *a, const void *b) {
- const CustomMount *x = a, *y = b;
- int r;
-
- r = path_compare(x->destination, y->destination);
- if (r != 0)
- return r;
-
- if (x->type < y->type)
- return -1;
- if (x->type > y->type)
- return 1;
-
- return 0;
-}
static int custom_mounts_prepare(void) {
unsigned i;
return 0;
}
-static int set_sanitized_path(char **b, const char *path) {
- char *p;
-
- assert(b);
- assert(path);
-
- p = canonicalize_file_name(path);
- if (!p) {
- if (errno != ENOENT)
- return -errno;
-
- p = path_make_absolute_cwd(path);
- if (!p)
- return -ENOMEM;
- }
-
- free(*b);
- *b = path_kill_slashes(p);
- return 0;
-}
-
static int detect_unified_cgroup_hierarchy(void) {
const char *e;
int r;
return 0;
}
-VolatileMode volatile_mode_from_string(const char *s) {
- int b;
-
- if (isempty(s))
- return _VOLATILE_MODE_INVALID;
-
- b = parse_boolean(s);
- if (b > 0)
- return VOLATILE_YES;
- if (b == 0)
- return VOLATILE_NO;
-
- if (streq(s, "state"))
- return VOLATILE_STATE;
-
- return _VOLATILE_MODE_INVALID;
-}
-
-int expose_port_parse(ExposePort **l, const char *s) {
-
- const char *split, *e;
- uint16_t container_port, host_port;
- int protocol;
- ExposePort *p;
- int r;
-
- if ((e = startswith(s, "tcp:")))
- protocol = IPPROTO_TCP;
- else if ((e = startswith(s, "udp:")))
- protocol = IPPROTO_UDP;
- else {
- e = s;
- protocol = IPPROTO_TCP;
- }
-
- split = strchr(e, ':');
- if (split) {
- char v[split - e + 1];
-
- memcpy(v, e, split - e);
- v[split - e] = 0;
-
- r = safe_atou16(v, &host_port);
- if (r < 0 || host_port <= 0)
- return -EINVAL;
-
- r = safe_atou16(split + 1, &container_port);
- } else {
- r = safe_atou16(e, &container_port);
- host_port = container_port;
- }
-
- if (r < 0 || container_port <= 0)
- return -EINVAL;
-
- LIST_FOREACH(ports, p, arg_expose_ports)
- if (p->protocol == protocol && p->host_port == host_port)
- return -EEXIST;
-
- p = new(ExposePort, 1);
- if (!p)
- return -ENOMEM;
-
- p->protocol = protocol;
- p->host_port = host_port;
- p->container_port = container_port;
-
- LIST_PREPEND(ports, *l, p);
-
- return 0;
-}
-
-int bind_mount_parse(CustomMount **l, unsigned *n, const char *s, bool read_only) {
- _cleanup_free_ char *source = NULL, *destination = NULL, *opts = NULL;
- const char *p = s;
- CustomMount *m;
- int r;
-
- assert(l);
- assert(n);
-
- r = extract_many_words(&p, ":", EXTRACT_DONT_COALESCE_SEPARATORS, &source, &destination, NULL);
- if (r < 0)
- return r;
- if (r == 0)
- return -EINVAL;
-
- if (r == 1) {
- destination = strdup(source);
- if (!destination)
- return -ENOMEM;
- }
-
- if (r == 2 && !isempty(p)) {
- opts = strdup(p);
- if (!opts)
- return -ENOMEM;
- }
-
- if (!path_is_absolute(source))
- return -EINVAL;
-
- if (!path_is_absolute(destination))
- return -EINVAL;
-
- m = custom_mount_add(l, n, CUSTOM_MOUNT_BIND);
- if (!m)
- return log_oom();
-
- m->source = source;
- m->destination = destination;
- m->read_only = read_only;
- m->options = opts;
-
- source = destination = opts = NULL;
- return 0;
-}
-
-int tmpfs_mount_parse(CustomMount **l, unsigned *n, const char *s) {
- _cleanup_free_ char *path = NULL, *opts = NULL;
- const char *p = s;
- CustomMount *m;
- int r;
-
- assert(l);
- assert(n);
- assert(s);
-
- r = extract_first_word(&p, &path, ":", EXTRACT_DONT_COALESCE_SEPARATORS);
- if (r < 0)
- return r;
- if (r == 0)
- return -EINVAL;
-
- if (isempty(p))
- opts = strdup("mode=0755");
- else
- opts = strdup(p);
- if (!opts)
- return -ENOMEM;
-
- if (!path_is_absolute(path))
- return -EINVAL;
-
- m = custom_mount_add(l, n, CUSTOM_MOUNT_TMPFS);
- if (!m)
- return -ENOMEM;
-
- m->destination = path;
- m->options = opts;
-
- path = opts = NULL;
- return 0;
-}
-
static int parse_argv(int argc, char *argv[]) {
enum {
return 0;
case ARG_VERSION:
- puts(PACKAGE_STRING);
- puts(SYSTEMD_FEATURES);
- return 0;
+ return version();
case 'D':
- r = set_sanitized_path(&arg_directory, optarg);
+ r = parse_path_argument_and_warn(optarg, false, &arg_directory);
if (r < 0)
- return log_error_errno(r, "Invalid root directory: %m");
-
+ return r;
break;
case ARG_TEMPLATE:
- r = set_sanitized_path(&arg_template, optarg);
+ r = parse_path_argument_and_warn(optarg, false, &arg_template);
if (r < 0)
- return log_error_errno(r, "Invalid template directory: %m");
-
+ return r;
break;
case 'i':
- r = set_sanitized_path(&arg_image, optarg);
+ r = parse_path_argument_and_warn(optarg, false, &arg_image);
if (r < 0)
- return log_error_errno(r, "Invalid image path: %m");
-
+ return r;
break;
case 'x':
return 0;
}
-static int tmpfs_patch_options(const char *options, char **ret) {
- char *buf = NULL;
+static int userns_lchown(const char *p, uid_t uid, gid_t gid) {
+ assert(p);
- if (arg_userns && arg_uid_shift != 0) {
- assert(arg_uid_shift != UID_INVALID);
+ if (!arg_userns)
+ return 0;
- if (options)
- (void) asprintf(&buf, "%s,uid=" UID_FMT ",gid=" UID_FMT, options, arg_uid_shift, arg_uid_shift);
- else
- (void) asprintf(&buf, "uid=" UID_FMT ",gid=" UID_FMT, arg_uid_shift, arg_uid_shift);
- if (!buf)
- return -ENOMEM;
+ if (uid == UID_INVALID && gid == GID_INVALID)
+ return 0;
- options = buf;
- }
+ if (uid != UID_INVALID) {
+ uid += arg_uid_shift;
-#ifdef HAVE_SELINUX
- if (arg_selinux_apifs_context) {
- char *t;
+ if (uid < arg_uid_shift || uid >= arg_uid_shift + arg_uid_range)
+ return -EOVERFLOW;
+ }
- if (options)
- t = strjoin(options, ",context=\"", arg_selinux_apifs_context, "\"", NULL);
- else
- t = strjoin("context=\"", arg_selinux_apifs_context, "\"", NULL);
- if (!t) {
- free(buf);
- return -ENOMEM;
- }
+ if (gid != GID_INVALID) {
+ gid += (gid_t) arg_uid_shift;
- free(buf);
- buf = t;
+ if (gid < (gid_t) arg_uid_shift || gid >= (gid_t) (arg_uid_shift + arg_uid_range))
+ return -EOVERFLOW;
}
-#endif
- *ret = buf;
- return !!buf;
-}
+ if (lchown(p, uid, gid) < 0)
+ return -errno;
-static int mount_all(const char *dest, bool userns) {
-
- typedef struct MountPoint {
- const char *what;
- const char *where;
- const char *type;
- const char *options;
- unsigned long flags;
- bool fatal;
- bool userns;
- } MountPoint;
-
- static const MountPoint mount_table[] = {
- { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true, true },
- { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true, true }, /* Bind mount first */
- { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_REMOUNT, true, true }, /* Then, make it r/o */
- { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true, false },
- { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true, false },
- { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false },
- { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false },
- { "tmpfs", "/tmp", "tmpfs", "mode=1777", MS_STRICTATIME, true, false },
-#ifdef HAVE_SELINUX
- { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false, false }, /* Bind mount first */
- { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_REMOUNT, false, false }, /* Then, make it r/o */
-#endif
- };
+ return 0;
+}
- unsigned k;
- int r;
+static int userns_mkdir(const char *root, const char *path, mode_t mode, uid_t uid, gid_t gid) {
+ const char *q;
- for (k = 0; k < ELEMENTSOF(mount_table); k++) {
- _cleanup_free_ char *where = NULL, *options = NULL;
- const char *o;
+ q = prefix_roota(root, path);
+ if (mkdir(q, mode) < 0) {
+ if (errno == EEXIST)
+ return 0;
+ return -errno;
+ }
- if (userns != mount_table[k].userns)
- continue;
+ return userns_lchown(q, uid, gid);
+}
- where = prefix_root(dest, mount_table[k].where);
- if (!where)
- return log_oom();
+static int setup_timezone(const char *dest) {
+ _cleanup_free_ char *p = NULL, *q = NULL;
+ const char *where, *check, *what;
+ char *z, *y;
+ int r;
- r = path_is_mount_point(where, AT_SYMLINK_FOLLOW);
- if (r < 0 && r != -ENOENT)
- return log_error_errno(r, "Failed to detect whether %s is a mount point: %m", where);
+ assert(dest);
- /* Skip this entry if it is not a remount. */
- if (mount_table[k].what && r > 0)
- continue;
+ /* Fix the timezone, if possible */
+ r = readlink_malloc("/etc/localtime", &p);
+ if (r < 0) {
+ log_warning("/etc/localtime is not a symlink, not updating container timezone.");
+ return 0;
+ }
- r = mkdir_p(where, 0755);
- if (r < 0) {
- if (mount_table[k].fatal)
- return log_error_errno(r, "Failed to create directory %s: %m", where);
-
- log_warning_errno(r, "Failed to create directory %s: %m", where);
- continue;
- }
-
- o = mount_table[k].options;
- if (streq_ptr(mount_table[k].type, "tmpfs")) {
- r = tmpfs_patch_options(o, &options);
- if (r < 0)
- return log_oom();
- if (r > 0)
- o = options;
- }
-
- if (mount(mount_table[k].what,
- where,
- mount_table[k].type,
- mount_table[k].flags,
- o) < 0) {
-
- if (mount_table[k].fatal)
- return log_error_errno(errno, "mount(%s) failed: %m", where);
-
- log_warning_errno(errno, "mount(%s) failed, ignoring: %m", where);
- }
+ 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;
}
- return 0;
-}
-
-static int parse_mount_bind_options(const char *options, unsigned long *mount_flags, char **mount_opts) {
- const char *p = options;
- unsigned long flags = *mount_flags;
- char *opts = NULL;
-
- assert(options);
-
- for (;;) {
- _cleanup_free_ char *word = NULL;
- int r = extract_first_word(&p, &word, ",", 0);
- if (r < 0)
- return log_error_errno(r, "Failed to extract mount option: %m");
- if (r == 0)
- break;
+ where = prefix_roota(dest, "/etc/localtime");
+ r = readlink_malloc(where, &q);
+ if (r >= 0) {
+ y = path_startswith(q, "../usr/share/zoneinfo/");
+ if (!y)
+ y = path_startswith(q, "/usr/share/zoneinfo/");
- if (streq(word, "rbind"))
- flags |= MS_REC;
- else if (streq(word, "norbind"))
- flags &= ~MS_REC;
- else {
- log_error("Invalid bind mount option: %s", word);
- return -EINVAL;
- }
+ /* Already pointing to the right place? Then do nothing .. */
+ if (y && streq(y, z))
+ return 0;
}
- *mount_flags = flags;
- /* in the future mount_opts will hold string options for mount(2) */
- *mount_opts = opts;
-
- return 0;
-}
-
-static int mount_bind(const char *dest, CustomMount *m) {
- struct stat source_st, dest_st;
- const char *where;
- unsigned long mount_flags = MS_BIND | MS_REC;
- _cleanup_free_ char *mount_opts = NULL;
- int r;
-
- assert(m);
-
- if (m->options) {
- r = parse_mount_bind_options(m->options, &mount_flags, &mount_opts);
- if (r < 0)
- return r;
+ check = strjoina("/usr/share/zoneinfo/", z);
+ check = prefix_root(dest, check);
+ if (laccess(check, F_OK) < 0) {
+ log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
+ return 0;
}
- if (stat(m->source, &source_st) < 0)
- return log_error_errno(errno, "Failed to stat %s: %m", m->source);
-
- where = prefix_roota(dest, m->destination);
-
- if (stat(where, &dest_st) >= 0) {
- if (S_ISDIR(source_st.st_mode) && !S_ISDIR(dest_st.st_mode)) {
- log_error("Cannot bind mount directory %s on file %s.", m->source, where);
- return -EINVAL;
- }
-
- if (!S_ISDIR(source_st.st_mode) && S_ISDIR(dest_st.st_mode)) {
- log_error("Cannot bind mount file %s on directory %s.", m->source, where);
- return -EINVAL;
- }
-
- } else if (errno == ENOENT) {
- r = mkdir_parents_label(where, 0755);
- if (r < 0)
- return log_error_errno(r, "Failed to make parents of %s: %m", where);
- } else {
- log_error_errno(errno, "Failed to stat %s: %m", where);
- return -errno;
+ r = unlink(where);
+ if (r < 0 && errno != ENOENT) {
+ log_error_errno(errno, "Failed to remove existing timezone info %s in container: %m", where);
+ return 0;
}
- /* Create the mount point. Any non-directory file can be
- * mounted on any non-directory file (regular, fifo, socket,
- * char, block).
- */
- if (S_ISDIR(source_st.st_mode))
- r = mkdir_label(where, 0755);
- else
- r = touch(where);
- if (r < 0 && r != -EEXIST)
- return log_error_errno(r, "Failed to create mount point %s: %m", where);
-
- if (mount(m->source, where, NULL, mount_flags, mount_opts) < 0)
- return log_error_errno(errno, "mount(%s) failed: %m", where);
-
- if (m->read_only) {
- r = bind_remount_recursive(where, true);
- if (r < 0)
- return log_error_errno(r, "Read-only bind mount failed: %m");
+ what = strjoina("../usr/share/zoneinfo/", z);
+ if (symlink(what, where) < 0) {
+ log_error_errno(errno, "Failed to correct timezone of container: %m");
+ return 0;
}
- return 0;
-}
-
-static int mount_tmpfs(const char *dest, CustomMount *m) {
- const char *where, *options;
- _cleanup_free_ char *buf = NULL;
- int r;
-
- assert(dest);
- assert(m);
-
- where = prefix_roota(dest, m->destination);
-
- r = mkdir_p_label(where, 0755);
- if (r < 0 && r != -EEXIST)
- return log_error_errno(r, "Creating mount point for tmpfs %s failed: %m", where);
-
- r = tmpfs_patch_options(m->options, &buf);
+ r = userns_lchown(where, 0, 0);
if (r < 0)
- return log_oom();
- options = r > 0 ? buf : m->options;
-
- if (mount("tmpfs", where, "tmpfs", MS_NODEV|MS_STRICTATIME, options) < 0)
- return log_error_errno(errno, "tmpfs mount to %s failed: %m", where);
+ return log_warning_errno(r, "Failed to chown /etc/localtime: %m");
return 0;
}
-static char *joined_and_escaped_lower_dirs(char * const *lower) {
- _cleanup_strv_free_ char **sv = NULL;
-
- sv = strv_copy(lower);
- if (!sv)
- return NULL;
-
- strv_reverse(sv);
-
- if (!strv_shell_escape(sv, ",:"))
- return NULL;
-
- return strv_join(sv, ":");
-}
-
-static int mount_overlay(const char *dest, CustomMount *m) {
- _cleanup_free_ char *lower = NULL;
- const char *where, *options;
+static int setup_resolv_conf(const char *dest) {
+ const char *where = NULL;
int r;
assert(dest);
- assert(m);
-
- where = prefix_roota(dest, m->destination);
-
- r = mkdir_label(where, 0755);
- if (r < 0 && r != -EEXIST)
- return log_error_errno(r, "Creating mount point for overlay %s failed: %m", where);
-
- (void) mkdir_p_label(m->source, 0755);
- lower = joined_and_escaped_lower_dirs(m->lower);
- if (!lower)
- return log_oom();
-
- if (m->read_only) {
- _cleanup_free_ char *escaped_source = NULL;
-
- escaped_source = shell_escape(m->source, ",:");
- if (!escaped_source)
- return log_oom();
-
- options = strjoina("lowerdir=", escaped_source, ":", lower);
- } else {
- _cleanup_free_ char *escaped_source = NULL, *escaped_work_dir = NULL;
-
- assert(m->work_dir);
- (void) mkdir_label(m->work_dir, 0700);
+ if (arg_private_network)
+ return 0;
- escaped_source = shell_escape(m->source, ",:");
- if (!escaped_source)
- return log_oom();
- escaped_work_dir = shell_escape(m->work_dir, ",:");
- if (!escaped_work_dir)
- return log_oom();
+ /* Fix resolv.conf, if possible */
+ where = prefix_roota(dest, "/etc/resolv.conf");
- options = strjoina("lowerdir=", lower, ",upperdir=", escaped_source, ",workdir=", escaped_work_dir);
+ r = copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644, 0);
+ if (r < 0) {
+ /* If the file already exists as symlink, let's
+ * suppress the warning, under the assumption that
+ * resolved or something similar runs inside and the
+ * symlink points there.
+ *
+ * If the disk image is read-only, there's also no
+ * point in complaining.
+ */
+ log_full_errno(IN_SET(r, -ELOOP, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+ "Failed to copy /etc/resolv.conf to %s: %m", where);
+ return 0;
}
- if (mount("overlay", where, "overlay", m->read_only ? MS_RDONLY : 0, options) < 0)
- return log_error_errno(errno, "overlay mount to %s failed: %m", where);
+ r = userns_lchown(where, 0, 0);
+ if (r < 0)
+ log_warning_errno(r, "Failed to chown /etc/resolv.conf: %m");
return 0;
}
-static int mount_custom(const char *dest) {
- unsigned i;
- int r;
-
- assert(dest);
-
- for (i = 0; i < arg_n_custom_mounts; i++) {
- CustomMount *m = &arg_custom_mounts[i];
-
- switch (m->type) {
-
- case CUSTOM_MOUNT_BIND:
- r = mount_bind(dest, m);
- break;
-
- case CUSTOM_MOUNT_TMPFS:
- r = mount_tmpfs(dest, m);
- break;
-
- case CUSTOM_MOUNT_OVERLAY:
- r = mount_overlay(dest, m);
- break;
-
- default:
- assert_not_reached("Unknown custom mount type");
- }
+static char* id128_format_as_uuid(sd_id128_t id, char s[37]) {
+ assert(s);
- if (r < 0)
- return r;
- }
+ snprintf(s, 37,
+ "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
+ SD_ID128_FORMAT_VAL(id));
- return 0;
+ return s;
}
-static int mount_legacy_cgroup_hierarchy(const char *dest, const char *controller, const char *hierarchy, bool read_only) {
- char *to;
+static int setup_boot_id(const char *dest) {
+ const char *from, *to;
+ sd_id128_t rnd = {};
+ char as_uuid[37];
int r;
- to = strjoina(dest, "/sys/fs/cgroup/", hierarchy);
-
- r = path_is_mount_point(to, 0);
- if (r < 0 && r != -ENOENT)
- return log_error_errno(r, "Failed to determine if %s is mounted already: %m", to);
- if (r > 0)
+ if (arg_share_system)
return 0;
- mkdir_p(to, 0755);
-
- /* The superblock mount options of the mount point need to be
- * identical to the hosts', and hence writable... */
- if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, controller) < 0)
- return log_error_errno(errno, "Failed to mount to %s: %m", to);
+ /* Generate a new randomized boot ID, so that each boot-up of
+ * the container gets a new one */
- /* ... hence let's only make the bind mount read-only, not the
- * superblock. */
- if (read_only) {
- if (mount(NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0)
- return log_error_errno(errno, "Failed to remount %s read-only: %m", to);
- }
- return 1;
-}
+ from = prefix_roota(dest, "/run/proc-sys-kernel-random-boot-id");
+ to = prefix_roota(dest, "/proc/sys/kernel/random/boot_id");
-static int mount_legacy_cgroups(const char *dest) {
- _cleanup_set_free_free_ Set *controllers = NULL;
- const char *cgroup_root;
- int r;
+ r = sd_id128_randomize(&rnd);
+ if (r < 0)
+ return log_error_errno(r, "Failed to generate random boot id: %m");
- cgroup_root = prefix_roota(dest, "/sys/fs/cgroup");
+ id128_format_as_uuid(rnd, as_uuid);
- /* Mount a tmpfs to /sys/fs/cgroup if it's not mounted there yet. */
- r = path_is_mount_point(cgroup_root, AT_SYMLINK_FOLLOW);
+ r = write_string_file(from, as_uuid, WRITE_STRING_FILE_CREATE);
if (r < 0)
- return log_error_errno(r, "Failed to determine if /sys/fs/cgroup is already mounted: %m");
- if (r == 0) {
- _cleanup_free_ char *options = NULL;
+ return log_error_errno(r, "Failed to write boot id: %m");
- r = tmpfs_patch_options("mode=755", &options);
- if (r < 0)
- return log_oom();
+ if (mount(from, to, NULL, MS_BIND, NULL) < 0)
+ r = log_error_errno(errno, "Failed to bind mount boot id: %m");
+ else if (mount(NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL) < 0)
+ log_warning_errno(errno, "Failed to make boot id read-only: %m");
- if (mount("tmpfs", cgroup_root, "tmpfs", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, options) < 0)
- return log_error_errno(errno, "Failed to mount /sys/fs/cgroup: %m");
- }
+ unlink(from);
+ return r;
+}
- if (cg_unified() > 0)
- goto skip_controllers;
+static int copy_devnodes(const char *dest) {
- controllers = set_new(&string_hash_ops);
- if (!controllers)
- return log_oom();
+ static const char devnodes[] =
+ "null\0"
+ "zero\0"
+ "full\0"
+ "random\0"
+ "urandom\0"
+ "tty\0"
+ "net/tun\0";
- r = cg_kernel_controllers(controllers);
- if (r < 0)
- return log_error_errno(r, "Failed to determine cgroup controllers: %m");
+ const char *d;
+ int r = 0;
+ _cleanup_umask_ mode_t u;
- for (;;) {
- _cleanup_free_ char *controller = NULL, *origin = NULL, *combined = NULL;
+ assert(dest);
- controller = set_steal_first(controllers);
- if (!controller)
- break;
+ u = umask(0000);
- origin = prefix_root("/sys/fs/cgroup/", controller);
- if (!origin)
- return log_oom();
+ /* Create /dev/net, so that we can create /dev/net/tun in it */
+ if (userns_mkdir(dest, "/dev/net", 0755, 0, 0) < 0)
+ return log_error_errno(r, "Failed to create /dev/net directory: %m");
- r = readlink_malloc(origin, &combined);
- if (r == -EINVAL) {
- /* Not a symbolic link, but directly a single cgroup hierarchy */
+ NULSTR_FOREACH(d, devnodes) {
+ _cleanup_free_ char *from = NULL, *to = NULL;
+ struct stat st;
- r = mount_legacy_cgroup_hierarchy(dest, controller, controller, true);
- if (r < 0)
- return r;
+ from = strappend("/dev/", d);
+ to = prefix_root(dest, from);
- } else if (r < 0)
- return log_error_errno(r, "Failed to read link %s: %m", origin);
- else {
- _cleanup_free_ char *target = NULL;
+ if (stat(from, &st) < 0) {
- target = prefix_root(dest, origin);
- if (!target)
- return log_oom();
+ if (errno != ENOENT)
+ return log_error_errno(errno, "Failed to stat %s: %m", from);
- /* A symbolic link, a combination of controllers in one hierarchy */
+ } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
- if (!filename_is_valid(combined)) {
- log_warning("Ignoring invalid combined hierarchy %s.", combined);
- continue;
- }
+ log_error("%s is not a char or block device, cannot copy.", from);
+ return -EIO;
- r = mount_legacy_cgroup_hierarchy(dest, combined, combined, true);
- if (r < 0)
- return r;
+ } else {
+ if (mknod(to, st.st_mode, st.st_rdev) < 0) {
+ if (errno != EPERM)
+ return log_error_errno(errno, "mknod(%s) failed: %m", to);
- r = symlink_idempotent(combined, target);
- if (r == -EINVAL) {
- log_error("Invalid existing symlink for combined hierarchy");
- return r;
+ /* Some systems abusively restrict mknod but
+ * allow bind mounts. */
+ r = touch(to);
+ if (r < 0)
+ return log_error_errno(r, "touch (%s) failed: %m", to);
+ if (mount(from, to, NULL, MS_BIND, NULL) < 0)
+ return log_error_errno(errno, "Both mknod and bind mount (%s) failed: %m", to);
}
+
+ r = userns_lchown(to, 0, 0);
if (r < 0)
- return log_error_errno(r, "Failed to create symlink for combined hierarchy: %m");
+ return log_error_errno(r, "chown() of device node %s failed: %m", to);
}
}
-skip_controllers:
- r = mount_legacy_cgroup_hierarchy(dest, "none,name=systemd,xattr", "systemd", false);
- if (r < 0)
- return r;
-
- if (mount(NULL, cgroup_root, NULL, MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755") < 0)
- return log_error_errno(errno, "Failed to remount %s read-only: %m", cgroup_root);
-
- return 0;
+ return r;
}
-static int mount_unified_cgroups(const char *dest) {
+static int setup_pts(const char *dest) {
+ _cleanup_free_ char *options = NULL;
const char *p;
- int r;
- assert(dest);
-
- p = strjoina(dest, "/sys/fs/cgroup");
-
- r = path_is_mount_point(p, AT_SYMLINK_FOLLOW);
- if (r < 0)
- return log_error_errno(r, "Failed to determine if %s is mounted already: %m", p);
- if (r > 0) {
- p = strjoina(dest, "/sys/fs/cgroup/cgroup.procs");
- if (access(p, F_OK) >= 0)
- return 0;
- if (errno != ENOENT)
- return log_error_errno(errno, "Failed to determine if mount point %s contains the unified cgroup hierarchy: %m", p);
-
- log_error("%s is already mounted but not a unified cgroup hierarchy. Refusing.", p);
- return -EINVAL;
- }
-
- if (mount("cgroup", p, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, "__DEVEL__sane_behavior") < 0)
- return log_error_errno(errno, "Failed to mount unified cgroup hierarchy to %s: %m", p);
-
- return 0;
-}
-
-static int mount_cgroups(const char *dest) {
- if (arg_unified_cgroup_hierarchy)
- return mount_unified_cgroups(dest);
+#ifdef HAVE_SELINUX
+ if (arg_selinux_apifs_context)
+ (void) asprintf(&options,
+ "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT ",context=\"%s\"",
+ arg_uid_shift + TTY_GID,
+ arg_selinux_apifs_context);
else
- return mount_legacy_cgroups(dest);
-}
-
-static int mount_systemd_cgroup_writable(const char *dest) {
- _cleanup_free_ char *own_cgroup_path = NULL;
- const char *systemd_root, *systemd_own;
- int r;
-
- assert(dest);
-
- r = cg_pid_get_path(NULL, 0, &own_cgroup_path);
- if (r < 0)
- return log_error_errno(r, "Failed to determine our own cgroup path: %m");
-
- /* If we are living in the top-level, then there's nothing to do... */
- if (path_equal(own_cgroup_path, "/"))
- return 0;
-
- if (arg_unified_cgroup_hierarchy) {
- systemd_own = strjoina(dest, "/sys/fs/cgroup", own_cgroup_path);
- systemd_root = prefix_roota(dest, "/sys/fs/cgroup");
- } else {
- systemd_own = strjoina(dest, "/sys/fs/cgroup/systemd", own_cgroup_path);
- systemd_root = prefix_roota(dest, "/sys/fs/cgroup/systemd");
- }
-
- /* Make our own cgroup a (writable) bind mount */
- if (mount(systemd_own, systemd_own, NULL, MS_BIND, NULL) < 0)
- return log_error_errno(errno, "Failed to turn %s into a bind mount: %m", own_cgroup_path);
-
- /* And then remount the systemd cgroup root read-only */
- if (mount(NULL, systemd_root, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0)
- return log_error_errno(errno, "Failed to mount cgroup root read-only: %m");
-
- return 0;
-}
-
-static int userns_lchown(const char *p, uid_t uid, gid_t gid) {
- assert(p);
-
- if (!arg_userns)
- return 0;
-
- if (uid == UID_INVALID && gid == GID_INVALID)
- return 0;
-
- if (uid != UID_INVALID) {
- uid += arg_uid_shift;
-
- if (uid < arg_uid_shift || uid >= arg_uid_shift + arg_uid_range)
- return -EOVERFLOW;
- }
-
- if (gid != GID_INVALID) {
- gid += (gid_t) arg_uid_shift;
-
- if (gid < (gid_t) arg_uid_shift || gid >= (gid_t) (arg_uid_shift + arg_uid_range))
- return -EOVERFLOW;
- }
-
- if (lchown(p, uid, gid) < 0)
- return -errno;
-
- return 0;
-}
-
-static int userns_mkdir(const char *root, const char *path, mode_t mode, uid_t uid, gid_t gid) {
- const char *q;
-
- q = prefix_roota(root, path);
- if (mkdir(q, mode) < 0) {
- if (errno == EEXIST)
- return 0;
- return -errno;
- }
-
- return userns_lchown(q, uid, gid);
-}
-
-static int setup_timezone(const char *dest) {
- _cleanup_free_ char *p = NULL, *q = NULL;
- const char *where, *check, *what;
- char *z, *y;
- int r;
-
- assert(dest);
-
- /* Fix the timezone, if possible */
- 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 = prefix_roota(dest, "/etc/localtime");
- r = readlink_malloc(where, &q);
- if (r >= 0) {
- y = path_startswith(q, "../usr/share/zoneinfo/");
- if (!y)
- y = path_startswith(q, "/usr/share/zoneinfo/");
-
- /* Already pointing to the right place? Then do nothing .. */
- if (y && streq(y, z))
- return 0;
- }
-
- check = strjoina("/usr/share/zoneinfo/", z);
- check = prefix_root(dest, check);
- if (laccess(check, F_OK) < 0) {
- log_warning("Timezone %s does not exist in container, not updating container timezone.", z);
- return 0;
- }
-
- r = unlink(where);
- if (r < 0 && errno != ENOENT) {
- log_error_errno(errno, "Failed to remove existing timezone info %s in container: %m", where);
- return 0;
- }
-
- what = strjoina("../usr/share/zoneinfo/", z);
- if (symlink(what, where) < 0) {
- log_error_errno(errno, "Failed to correct timezone of container: %m");
- return 0;
- }
-
- r = userns_lchown(where, 0, 0);
- if (r < 0)
- return log_warning_errno(r, "Failed to chown /etc/localtime: %m");
-
- return 0;
-}
-
-static int setup_resolv_conf(const char *dest) {
- const char *where = NULL;
- int r;
-
- assert(dest);
-
- if (arg_private_network)
- return 0;
-
- /* Fix resolv.conf, if possible */
- where = prefix_roota(dest, "/etc/resolv.conf");
-
- r = copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644, 0);
- if (r < 0) {
- /* If the file already exists as symlink, let's
- * suppress the warning, under the assumption that
- * resolved or something similar runs inside and the
- * symlink points there.
- *
- * If the disk image is read-only, there's also no
- * point in complaining.
- */
- log_full_errno(IN_SET(r, -ELOOP, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
- "Failed to copy /etc/resolv.conf to %s: %m", where);
- return 0;
- }
-
- r = userns_lchown(where, 0, 0);
- if (r < 0)
- log_warning_errno(r, "Failed to chown /etc/resolv.conf: %m");
-
- return 0;
-}
-
-static int setup_volatile_state(const char *directory) {
- _cleanup_free_ char *buf = NULL;
- const char *p, *options;
- int r;
-
- assert(directory);
-
- if (arg_volatile_mode != VOLATILE_STATE)
- return 0;
-
- /* --volatile=state means we simply overmount /var
- with a tmpfs, and the rest read-only. */
-
- r = bind_remount_recursive(directory, true);
- if (r < 0)
- return log_error_errno(r, "Failed to remount %s read-only: %m", directory);
-
- p = prefix_roota(directory, "/var");
- r = mkdir(p, 0755);
- if (r < 0 && errno != EEXIST)
- return log_error_errno(errno, "Failed to create %s: %m", directory);
-
- options = "mode=755";
- r = tmpfs_patch_options(options, &buf);
- if (r < 0)
- return log_oom();
- if (r > 0)
- options = buf;
-
- if (mount("tmpfs", p, "tmpfs", MS_STRICTATIME, options) < 0)
- return log_error_errno(errno, "Failed to mount tmpfs to /var: %m");
-
- return 0;
-}
-
-static int setup_volatile(const char *directory) {
- bool tmpfs_mounted = false, bind_mounted = false;
- char template[] = "/tmp/nspawn-volatile-XXXXXX";
- _cleanup_free_ char *buf = NULL;
- const char *f, *t, *options;
- int r;
-
- assert(directory);
-
- if (arg_volatile_mode != VOLATILE_YES)
- return 0;
-
- /* --volatile=yes means we mount a tmpfs to the root dir, and
- the original /usr to use inside it, and that read-only. */
-
- if (!mkdtemp(template))
- return log_error_errno(errno, "Failed to create temporary directory: %m");
+#endif
+ (void) asprintf(&options,
+ "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT,
+ arg_uid_shift + TTY_GID);
- options = "mode=755";
- r = tmpfs_patch_options(options, &buf);
- if (r < 0)
- return log_oom();
- if (r > 0)
- options = buf;
-
- if (mount("tmpfs", template, "tmpfs", MS_STRICTATIME, options) < 0) {
- r = log_error_errno(errno, "Failed to mount tmpfs for root directory: %m");
- goto fail;
- }
-
- tmpfs_mounted = true;
-
- f = prefix_roota(directory, "/usr");
- t = prefix_roota(template, "/usr");
-
- r = mkdir(t, 0755);
- if (r < 0 && errno != EEXIST) {
- r = log_error_errno(errno, "Failed to create %s: %m", t);
- goto fail;
- }
-
- if (mount(f, t, NULL, MS_BIND|MS_REC, NULL) < 0) {
- r = log_error_errno(errno, "Failed to create /usr bind mount: %m");
- goto fail;
- }
-
- bind_mounted = true;
-
- r = bind_remount_recursive(t, true);
- if (r < 0) {
- log_error_errno(r, "Failed to remount %s read-only: %m", t);
- goto fail;
- }
-
- if (mount(template, directory, NULL, MS_MOVE, NULL) < 0) {
- r = log_error_errno(errno, "Failed to move root mount: %m");
- goto fail;
- }
-
- (void) rmdir(template);
-
- return 0;
-
-fail:
- if (bind_mounted)
- (void) umount(t);
-
- if (tmpfs_mounted)
- (void) umount(template);
- (void) rmdir(template);
- return r;
-}
-
-static char* id128_format_as_uuid(sd_id128_t id, char s[37]) {
- assert(s);
-
- snprintf(s, 37,
- "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x",
- SD_ID128_FORMAT_VAL(id));
-
- return s;
-}
-
-static int setup_boot_id(const char *dest) {
- const char *from, *to;
- sd_id128_t rnd = {};
- char as_uuid[37];
- int r;
-
- if (arg_share_system)
- return 0;
-
- /* Generate a new randomized boot ID, so that each boot-up of
- * the container gets a new one */
-
- from = prefix_roota(dest, "/run/proc-sys-kernel-random-boot-id");
- to = prefix_roota(dest, "/proc/sys/kernel/random/boot_id");
-
- r = sd_id128_randomize(&rnd);
- if (r < 0)
- return log_error_errno(r, "Failed to generate random boot id: %m");
-
- id128_format_as_uuid(rnd, as_uuid);
-
- r = write_string_file(from, as_uuid, WRITE_STRING_FILE_CREATE);
- if (r < 0)
- return log_error_errno(r, "Failed to write boot id: %m");
-
- if (mount(from, to, NULL, MS_BIND, NULL) < 0)
- r = log_error_errno(errno, "Failed to bind mount boot id: %m");
- else if (mount(NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL) < 0)
- log_warning_errno(errno, "Failed to make boot id read-only: %m");
-
- unlink(from);
- return r;
-}
-
-static int copy_devnodes(const char *dest) {
-
- static const char devnodes[] =
- "null\0"
- "zero\0"
- "full\0"
- "random\0"
- "urandom\0"
- "tty\0"
- "net/tun\0";
-
- const char *d;
- int r = 0;
- _cleanup_umask_ mode_t u;
-
- assert(dest);
-
- u = umask(0000);
-
- /* Create /dev/net, so that we can create /dev/net/tun in it */
- if (userns_mkdir(dest, "/dev/net", 0755, 0, 0) < 0)
- return log_error_errno(r, "Failed to create /dev/net directory: %m");
-
- NULSTR_FOREACH(d, devnodes) {
- _cleanup_free_ char *from = NULL, *to = NULL;
- struct stat st;
-
- from = strappend("/dev/", d);
- to = prefix_root(dest, from);
-
- if (stat(from, &st) < 0) {
-
- if (errno != ENOENT)
- return log_error_errno(errno, "Failed to stat %s: %m", from);
-
- } 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);
- return -EIO;
-
- } else {
- if (mknod(to, st.st_mode, st.st_rdev) < 0) {
- if (errno != EPERM)
- return log_error_errno(errno, "mknod(%s) failed: %m", to);
-
- /* Some systems abusively restrict mknod but
- * allow bind mounts. */
- r = touch(to);
- if (r < 0)
- return log_error_errno(r, "touch (%s) failed: %m", to);
- if (mount(from, to, NULL, MS_BIND, NULL) < 0)
- return log_error_errno(errno, "Both mknod and bind mount (%s) failed: %m", to);
- }
-
- r = userns_lchown(to, 0, 0);
- if (r < 0)
- return log_error_errno(r, "chown() of device node %s failed: %m", to);
- }
- }
-
- return r;
-}
-
-static int setup_pts(const char *dest) {
- _cleanup_free_ char *options = NULL;
- const char *p;
-
-#ifdef HAVE_SELINUX
- if (arg_selinux_apifs_context)
- (void) asprintf(&options,
- "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT ",context=\"%s\"",
- arg_uid_shift + TTY_GID,
- arg_selinux_apifs_context);
- else
-#endif
- (void) asprintf(&options,
- "newinstance,ptmxmode=0666,mode=620,gid=" GID_FMT,
- arg_uid_shift + TTY_GID);
-
- if (!options)
+ if (!options)
return log_oom();
/* Mount /dev/pts itself */
if (userns_lchown(p, 0, 0) < 0)
return log_error_errno(errno, "Failed to chown /dev/pts: %m");
- /* Create /dev/ptmx symlink */
- p = prefix_roota(dest, "/dev/ptmx");
- if (symlink("pts/ptmx", p) < 0)
- return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m");
- if (userns_lchown(p, 0, 0) < 0)
- return log_error_errno(errno, "Failed to chown /dev/ptmx: %m");
-
- /* And fix /dev/pts/ptmx ownership */
- p = prefix_roota(dest, "/dev/pts/ptmx");
- if (userns_lchown(p, 0, 0) < 0)
- return log_error_errno(errno, "Failed to chown /dev/pts/ptmx: %m");
-
- return 0;
-}
-
-static int setup_dev_console(const char *dest, const char *console) {
- _cleanup_umask_ mode_t u;
- const char *to;
- int r;
-
- assert(dest);
- assert(console);
-
- u = umask(0000);
-
- r = chmod_and_chown(console, 0600, arg_uid_shift, arg_uid_shift);
- if (r < 0)
- return log_error_errno(r, "Failed to correct access mode for TTY: %m");
-
- /* We need to bind mount the right tty to /dev/console since
- * ptys can only exist on pts file systems. To have something
- * to bind mount things on we create a empty regular file. */
-
- to = prefix_roota(dest, "/dev/console");
- r = touch(to);
- if (r < 0)
- return log_error_errno(r, "touch() for /dev/console failed: %m");
-
- if (mount(console, to, NULL, MS_BIND, NULL) < 0)
- return log_error_errno(errno, "Bind mount for /dev/console failed: %m");
-
- return 0;
-}
-
-static int setup_kmsg(const char *dest, int kmsg_socket) {
- const char *from, *to;
- _cleanup_umask_ mode_t u;
- int fd, k;
- union {
- struct cmsghdr cmsghdr;
- uint8_t buf[CMSG_SPACE(sizeof(int))];
- } control = {};
- struct msghdr mh = {
- .msg_control = &control,
- .msg_controllen = sizeof(control),
- };
- struct cmsghdr *cmsg;
-
- assert(kmsg_socket >= 0);
-
- u = umask(0000);
-
- /* We create the kmsg FIFO as /run/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. */
- from = prefix_roota(dest, "/run/kmsg");
- to = prefix_roota(dest, "/proc/kmsg");
-
- if (mkfifo(from, 0600) < 0)
- return log_error_errno(errno, "mkfifo() for /run/kmsg failed: %m");
- if (mount(from, to, NULL, MS_BIND, NULL) < 0)
- return log_error_errno(errno, "Bind mount for /proc/kmsg failed: %m");
-
- fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
- if (fd < 0)
- return log_error_errno(errno, "Failed to open fifo: %m");
-
- cmsg = CMSG_FIRSTHDR(&mh);
- cmsg->cmsg_level = SOL_SOCKET;
- cmsg->cmsg_type = SCM_RIGHTS;
- cmsg->cmsg_len = CMSG_LEN(sizeof(int));
- memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
-
- mh.msg_controllen = cmsg->cmsg_len;
-
- /* Store away the fd in the socket, so that it stays open as
- * long as we run the child */
- k = sendmsg(kmsg_socket, &mh, MSG_NOSIGNAL);
- safe_close(fd);
-
- if (k < 0)
- return log_error_errno(errno, "Failed to send FIFO fd: %m");
-
- /* And now make the FIFO unavailable as /run/kmsg... */
- (void) unlink(from);
-
- return 0;
-}
-
-static int send_rtnl(int send_fd) {
- union {
- struct cmsghdr cmsghdr;
- uint8_t buf[CMSG_SPACE(sizeof(int))];
- } control = {};
- struct msghdr mh = {
- .msg_control = &control,
- .msg_controllen = sizeof(control),
- };
- struct cmsghdr *cmsg;
- _cleanup_close_ int fd = -1;
- ssize_t k;
-
- assert(send_fd >= 0);
-
- if (!arg_expose_ports)
- return 0;
-
- fd = socket(PF_NETLINK, SOCK_RAW|SOCK_CLOEXEC|SOCK_NONBLOCK, NETLINK_ROUTE);
- if (fd < 0)
- return log_error_errno(errno, "Failed to allocate container netlink: %m");
-
- cmsg = CMSG_FIRSTHDR(&mh);
- cmsg->cmsg_level = SOL_SOCKET;
- cmsg->cmsg_type = SCM_RIGHTS;
- cmsg->cmsg_len = CMSG_LEN(sizeof(int));
- memcpy(CMSG_DATA(cmsg), &fd, sizeof(int));
-
- mh.msg_controllen = cmsg->cmsg_len;
-
- /* Store away the fd in the socket, so that it stays open as
- * long as we run the child */
- k = sendmsg(send_fd, &mh, MSG_NOSIGNAL);
- if (k < 0)
- return log_error_errno(errno, "Failed to send netlink fd: %m");
-
- return 0;
-}
-
-static int flush_ports(union in_addr_union *exposed) {
- ExposePort *p;
- int r, af = AF_INET;
-
- assert(exposed);
-
- if (!arg_expose_ports)
- return 0;
-
- if (in_addr_is_null(af, exposed))
- return 0;
-
- log_debug("Lost IP address.");
-
- LIST_FOREACH(ports, p, arg_expose_ports) {
- r = fw_add_local_dnat(false,
- af,
- p->protocol,
- NULL,
- NULL, 0,
- NULL, 0,
- p->host_port,
- exposed,
- p->container_port,
- NULL);
- if (r < 0)
- log_warning_errno(r, "Failed to modify firewall: %m");
- }
-
- *exposed = IN_ADDR_NULL;
- return 0;
-}
-
-static int expose_ports(sd_netlink *rtnl, union in_addr_union *exposed) {
- _cleanup_free_ struct local_address *addresses = NULL;
- _cleanup_free_ char *pretty = NULL;
- union in_addr_union new_exposed;
- ExposePort *p;
- bool add;
- int af = AF_INET, r;
-
- assert(exposed);
-
- /* Invoked each time an address is added or removed inside the
- * container */
-
- if (!arg_expose_ports)
- return 0;
-
- r = local_addresses(rtnl, 0, af, &addresses);
- if (r < 0)
- return log_error_errno(r, "Failed to enumerate local addresses: %m");
-
- add = r > 0 &&
- addresses[0].family == af &&
- addresses[0].scope < RT_SCOPE_LINK;
-
- if (!add)
- return flush_ports(exposed);
-
- new_exposed = addresses[0].address;
- if (in_addr_equal(af, exposed, &new_exposed))
- return 0;
-
- in_addr_to_string(af, &new_exposed, &pretty);
- log_debug("New container IP is %s.", strna(pretty));
-
- LIST_FOREACH(ports, p, arg_expose_ports) {
-
- r = fw_add_local_dnat(true,
- af,
- p->protocol,
- NULL,
- NULL, 0,
- NULL, 0,
- p->host_port,
- &new_exposed,
- p->container_port,
- in_addr_is_null(af, exposed) ? NULL : exposed);
- if (r < 0)
- log_warning_errno(r, "Failed to modify firewall: %m");
- }
-
- *exposed = new_exposed;
- return 0;
-}
-
-void expose_port_free_all(ExposePort *p) {
-
- while (p) {
- ExposePort *q = p;
- LIST_REMOVE(ports, p, q);
- free(q);
- }
-}
-
-static int on_address_change(sd_netlink *rtnl, sd_netlink_message *m, void *userdata) {
- union in_addr_union *exposed = userdata;
-
- assert(rtnl);
- assert(m);
- assert(exposed);
-
- expose_ports(rtnl, exposed);
- return 0;
-}
-
-static int watch_rtnl(sd_event *event, int recv_fd, union in_addr_union *exposed, sd_netlink **ret) {
- union {
- struct cmsghdr cmsghdr;
- uint8_t buf[CMSG_SPACE(sizeof(int))];
- } control = {};
- struct msghdr mh = {
- .msg_control = &control,
- .msg_controllen = sizeof(control),
- };
- struct cmsghdr *cmsg;
- _cleanup_netlink_unref_ sd_netlink *rtnl = NULL;
- int fd, r;
- ssize_t k;
-
- assert(event);
- assert(recv_fd >= 0);
- assert(ret);
-
- if (!arg_expose_ports)
- return 0;
-
- k = recvmsg(recv_fd, &mh, MSG_NOSIGNAL);
- if (k < 0)
- return log_error_errno(errno, "Failed to recv netlink fd: %m");
-
- cmsg = CMSG_FIRSTHDR(&mh);
- assert(cmsg->cmsg_level == SOL_SOCKET);
- assert(cmsg->cmsg_type == SCM_RIGHTS);
- assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int)));
- memcpy(&fd, CMSG_DATA(cmsg), sizeof(int));
-
- r = sd_netlink_open_fd(&rtnl, fd);
- if (r < 0) {
- safe_close(fd);
- return log_error_errno(r, "Failed to create rtnl object: %m");
- }
-
- r = sd_netlink_add_match(rtnl, RTM_NEWADDR, on_address_change, exposed);
- if (r < 0)
- return log_error_errno(r, "Failed to subscribe to RTM_NEWADDR messages: %m");
-
- r = sd_netlink_add_match(rtnl, RTM_DELADDR, on_address_change, exposed);
- if (r < 0)
- return log_error_errno(r, "Failed to subscribe to RTM_DELADDR messages: %m");
-
- r = sd_netlink_attach_event(rtnl, event, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to add to even loop: %m");
-
- *ret = rtnl;
- rtnl = NULL;
-
- return 0;
-}
-
-static int setup_hostname(void) {
-
- if (arg_share_system)
- return 0;
-
- if (sethostname_idempotent(arg_machine) < 0)
- return -errno;
-
- return 0;
-}
-
-static int setup_journal(const char *directory) {
- sd_id128_t machine_id, this_id;
- _cleanup_free_ char *b = NULL, *d = NULL;
- const char *etc_machine_id, *p, *q;
- char *id;
- int r;
-
- /* Don't link journals in ephemeral mode */
- if (arg_ephemeral)
- return 0;
-
- etc_machine_id = prefix_roota(directory, "/etc/machine-id");
-
- r = read_one_line_file(etc_machine_id, &b);
- if (r == -ENOENT && arg_link_journal == LINK_AUTO)
- return 0;
- else if (r < 0)
- return log_error_errno(r, "Failed to read machine ID from %s: %m", etc_machine_id);
-
- id = strstrip(b);
- if (isempty(id) && arg_link_journal == LINK_AUTO)
- return 0;
-
- /* Verify validity */
- r = sd_id128_from_string(id, &machine_id);
- if (r < 0)
- return log_error_errno(r, "Failed to parse machine ID from %s: %m", etc_machine_id);
-
- r = sd_id128_get_machine(&this_id);
- if (r < 0)
- return log_error_errno(r, "Failed to retrieve machine ID: %m");
-
- if (sd_id128_equal(machine_id, this_id)) {
- log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR,
- "Host and machine ids are equal (%s): refusing to link journals", id);
- if (arg_link_journal == LINK_AUTO)
- return 0;
- return -EEXIST;
- }
-
- if (arg_link_journal == LINK_NO)
- return 0;
-
- r = userns_mkdir(directory, "/var", 0755, 0, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to create /var: %m");
-
- r = userns_mkdir(directory, "/var/log", 0755, 0, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to create /var/log: %m");
-
- r = userns_mkdir(directory, "/var/log/journal", 0755, 0, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to create /var/log/journal: %m");
-
- p = strjoina("/var/log/journal/", id);
- q = prefix_roota(directory, p);
-
- if (path_is_mount_point(p, 0) > 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(q, 0) > 0) {
- if (arg_link_journal != LINK_AUTO) {
- log_error("%s: already a mount point, refusing to use for journal", q);
- return -EEXIST;
- }
-
- 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)) {
-
- r = userns_mkdir(directory, p, 0755, 0, 0);
- if (r < 0)
- log_warning_errno(errno, "Failed to create directory %s: %m", q);
- return 0;
- }
-
- if (unlink(p) < 0)
- return log_error_errno(errno, "Failed to remove symlink %s: %m", p);
- } else if (r == -EINVAL) {
-
- if (arg_link_journal == LINK_GUEST &&
- rmdir(p) < 0) {
-
- if (errno == ENOTDIR) {
- log_error("%s already exists and is neither a symlink nor a directory", p);
- return r;
- } else {
- log_error_errno(errno, "Failed to remove %s: %m", p);
- return -errno;
- }
- }
- } else if (r != -ENOENT) {
- log_error_errno(errno, "readlink(%s) failed: %m", p);
- return r;
- }
-
- if (arg_link_journal == LINK_GUEST) {
-
- if (symlink(q, p) < 0) {
- if (arg_link_journal_try) {
- log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p);
- return 0;
- } else {
- log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p);
- return -errno;
- }
- }
-
- r = userns_mkdir(directory, p, 0755, 0, 0);
- if (r < 0)
- log_warning_errno(errno, "Failed to create directory %s: %m", q);
- return 0;
- }
-
- if (arg_link_journal == LINK_HOST) {
- /* don't create parents here -- if the host doesn't have
- * permanent journal set up, don't force it here */
- r = mkdir(p, 0755);
- if (r < 0) {
- if (arg_link_journal_try) {
- log_debug_errno(errno, "Failed to create %s, skipping journal setup: %m", p);
- return 0;
- } else {
- log_error_errno(errno, "Failed to create %s: %m", p);
- return r;
- }
- }
-
- } else if (access(p, F_OK) < 0)
- return 0;
-
- if (dir_is_empty(q) == 0)
- log_warning("%s is not empty, proceeding anyway.", q);
-
- r = userns_mkdir(directory, p, 0755, 0, 0);
- if (r < 0) {
- log_error_errno(errno, "Failed to create %s: %m", q);
- return r;
- }
-
- if (mount(p, q, NULL, MS_BIND, NULL) < 0)
- return log_error_errno(errno, "Failed to bind mount journal from host into guest: %m");
-
- return 0;
-}
-
-static int drop_capabilities(void) {
- return capability_bounding_set_drop(~arg_retain, false);
-}
-
-static int register_machine(pid_t pid, int local_ifindex) {
- _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
- _cleanup_bus_flush_close_unref_ sd_bus *bus = NULL;
- int r;
-
- if (!arg_register)
- return 0;
-
- r = sd_bus_default_system(&bus);
- if (r < 0)
- return log_error_errno(r, "Failed to open system bus: %m");
-
- if (arg_keep_unit) {
- r = sd_bus_call_method(
- bus,
- "org.freedesktop.machine1",
- "/org/freedesktop/machine1",
- "org.freedesktop.machine1.Manager",
- "RegisterMachineWithNetwork",
- &error,
- NULL,
- "sayssusai",
- arg_machine,
- SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
- "nspawn",
- "container",
- (uint32_t) pid,
- strempty(arg_directory),
- local_ifindex > 0 ? 1 : 0, local_ifindex);
- } else {
- _cleanup_bus_message_unref_ sd_bus_message *m = NULL;
- char **i;
- unsigned j;
-
- r = sd_bus_message_new_method_call(
- bus,
- &m,
- "org.freedesktop.machine1",
- "/org/freedesktop/machine1",
- "org.freedesktop.machine1.Manager",
- "CreateMachineWithNetwork");
- if (r < 0)
- return bus_log_create_error(r);
-
- r = sd_bus_message_append(
- m,
- "sayssusai",
- arg_machine,
- SD_BUS_MESSAGE_APPEND_ID128(arg_uuid),
- "nspawn",
- "container",
- (uint32_t) pid,
- strempty(arg_directory),
- local_ifindex > 0 ? 1 : 0, local_ifindex);
- if (r < 0)
- return bus_log_create_error(r);
-
- r = sd_bus_message_open_container(m, 'a', "(sv)");
- if (r < 0)
- return bus_log_create_error(r);
-
- if (!isempty(arg_slice)) {
- r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice);
- if (r < 0)
- return bus_log_create_error(r);
- }
-
- r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict");
- if (r < 0)
- return bus_log_create_error(r);
-
- /* If you make changes here, also make sure to update
- * systemd-nspawn@.service, to keep the device
- * policies in sync regardless if we are run with or
- * without the --keep-unit switch. */
- r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 9,
- /* Allow the container to
- * access and create the API
- * device nodes, so that
- * PrivateDevices= in the
- * container can work
- * fine */
- "/dev/null", "rwm",
- "/dev/zero", "rwm",
- "/dev/full", "rwm",
- "/dev/random", "rwm",
- "/dev/urandom", "rwm",
- "/dev/tty", "rwm",
- "/dev/net/tun", "rwm",
- /* Allow the container
- * access to ptys. However,
- * do not permit the
- * container to ever create
- * these device nodes. */
- "/dev/pts/ptmx", "rw",
- "char-pts", "rw");
- if (r < 0)
- return bus_log_create_error(r);
-
- for (j = 0; j < arg_n_custom_mounts; j++) {
- CustomMount *cm = &arg_custom_mounts[j];
-
- if (cm->type != CUSTOM_MOUNT_BIND)
- continue;
-
- r = is_device_node(cm->source);
- if (r < 0)
- return log_error_errno(r, "Failed to stat %s: %m", cm->source);
-
- if (r) {
- r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 1,
- cm->source, cm->read_only ? "r" : "rw");
- if (r < 0)
- return log_error_errno(r, "Failed to append message arguments: %m");
- }
- }
-
- if (arg_kill_signal != 0) {
- r = sd_bus_message_append(m, "(sv)", "KillSignal", "i", arg_kill_signal);
- if (r < 0)
- return bus_log_create_error(r);
-
- r = sd_bus_message_append(m, "(sv)", "KillMode", "s", "mixed");
- if (r < 0)
- return bus_log_create_error(r);
- }
-
- STRV_FOREACH(i, arg_property) {
- r = sd_bus_message_open_container(m, 'r', "sv");
- if (r < 0)
- return bus_log_create_error(r);
-
- r = bus_append_unit_property_assignment(m, *i);
- if (r < 0)
- return r;
-
- r = sd_bus_message_close_container(m);
- if (r < 0)
- return bus_log_create_error(r);
- }
-
- r = sd_bus_message_close_container(m);
- if (r < 0)
- return bus_log_create_error(r);
-
- r = sd_bus_call(bus, m, 0, &error, NULL);
- }
-
- if (r < 0) {
- log_error("Failed to register machine: %s", bus_error_message(&error, r));
- return r;
- }
-
- return 0;
-}
-
-static int terminate_machine(pid_t pid) {
- _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL;
- _cleanup_bus_message_unref_ sd_bus_message *reply = NULL;
- _cleanup_bus_flush_close_unref_ sd_bus *bus = NULL;
- const char *path;
- int r;
-
- if (!arg_register)
- return 0;
-
- /* If we are reusing the unit, then just exit, systemd will do
- * the right thing when we exit. */
- if (arg_keep_unit)
- return 0;
-
- r = sd_bus_default_system(&bus);
- if (r < 0)
- return log_error_errno(r, "Failed to open system bus: %m");
-
- r = sd_bus_call_method(
- bus,
- "org.freedesktop.machine1",
- "/org/freedesktop/machine1",
- "org.freedesktop.machine1.Manager",
- "GetMachineByPID",
- &error,
- &reply,
- "u",
- (uint32_t) pid);
- if (r < 0) {
- /* Note that the machine might already have been
- * cleaned up automatically, hence don't consider it a
- * failure if we cannot get the machine object. */
- log_debug("Failed to get machine: %s", bus_error_message(&error, r));
- return 0;
- }
-
- r = sd_bus_message_read(reply, "o", &path);
- if (r < 0)
- return bus_log_parse_error(r);
-
- r = sd_bus_call_method(
- bus,
- "org.freedesktop.machine1",
- path,
- "org.freedesktop.machine1.Machine",
- "Terminate",
- &error,
- NULL,
- NULL);
- if (r < 0) {
- log_debug("Failed to terminate machine: %s", bus_error_message(&error, r));
- return 0;
- }
-
- return 0;
-}
-
-static int reset_audit_loginuid(void) {
- _cleanup_free_ char *p = NULL;
- int r;
-
- if (arg_share_system)
- return 0;
-
- r = read_one_line_file("/proc/self/loginuid", &p);
- if (r == -ENOENT)
- return 0;
- if (r < 0)
- return log_error_errno(r, "Failed to read /proc/self/loginuid: %m");
-
- /* Already reset? */
- if (streq(p, "4294967295"))
- return 0;
-
- r = write_string_file("/proc/self/loginuid", "4294967295", 0);
- if (r < 0) {
- log_error_errno(r,
- "Failed to reset audit login UID. This probably means that your kernel is too\n"
- "old and you have audit enabled. Note that the auditing subsystem is known to\n"
- "be incompatible with containers on old kernels. Please make sure to upgrade\n"
- "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
- "using systemd-nspawn. Sleeping for 5s... (%m)");
-
- sleep(5);
- }
-
- return 0;
-}
-
-#define HOST_HASH_KEY SD_ID128_MAKE(1a,37,6f,c7,46,ec,45,0b,ad,a3,d5,31,06,60,5d,b1)
-#define CONTAINER_HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2)
-#define MACVLAN_HASH_KEY SD_ID128_MAKE(00,13,6d,bc,66,83,44,81,bb,0c,f9,51,1f,24,a6,6f)
-
-static int generate_mac(struct ether_addr *mac, sd_id128_t hash_key, uint64_t idx) {
- uint8_t result[8];
- size_t l, sz;
- uint8_t *v, *i;
- int r;
-
- l = strlen(arg_machine);
- sz = sizeof(sd_id128_t) + l;
- if (idx > 0)
- sz += sizeof(idx);
-
- v = alloca(sz);
-
- /* fetch some persistent data unique to the host */
- r = sd_id128_get_machine((sd_id128_t*) v);
- if (r < 0)
- return r;
-
- /* combine with some data unique (on this host) to this
- * container instance */
- i = mempcpy(v + sizeof(sd_id128_t), arg_machine, l);
- if (idx > 0) {
- idx = htole64(idx);
- memcpy(i, &idx, sizeof(idx));
- }
-
- /* Let's hash the host machine ID plus the container name. We
- * use a fixed, but originally randomly created hash key here. */
- siphash24(result, v, sz, hash_key.bytes);
-
- assert_cc(ETH_ALEN <= sizeof(result));
- memcpy(mac->ether_addr_octet, result, ETH_ALEN);
-
- /* see eth_random_addr in the kernel */
- mac->ether_addr_octet[0] &= 0xfe; /* clear multicast bit */
- mac->ether_addr_octet[0] |= 0x02; /* set local assignment bit (IEEE802) */
-
- return 0;
-}
-
-static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ], int *ifi) {
- _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL;
- _cleanup_netlink_unref_ sd_netlink *rtnl = NULL;
- struct ether_addr mac_host, mac_container;
- int r, i;
-
- if (!arg_private_network)
- return 0;
-
- if (!arg_network_veth)
- return 0;
-
- /* Use two different interface name prefixes depending whether
- * we are in bridge mode or not. */
- snprintf(iface_name, IFNAMSIZ - 1, "%s-%s",
- arg_network_bridge ? "vb" : "ve", arg_machine);
-
- r = generate_mac(&mac_container, CONTAINER_HASH_KEY, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to generate predictable MAC address for container side: %m");
-
- r = generate_mac(&mac_host, HOST_HASH_KEY, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to generate predictable MAC address for host side: %m");
-
- r = sd_netlink_open(&rtnl);
- if (r < 0)
- return log_error_errno(r, "Failed to connect to netlink: %m");
-
- r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to allocate netlink message: %m");
-
- r = sd_netlink_message_append_string(m, IFLA_IFNAME, iface_name);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink interface name: %m");
-
- r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, &mac_host);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink MAC address: %m");
-
- r = sd_netlink_message_open_container(m, IFLA_LINKINFO);
- if (r < 0)
- return log_error_errno(r, "Failed to open netlink container: %m");
-
- r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, "veth");
- if (r < 0)
- return log_error_errno(r, "Failed to open netlink container: %m");
+ /* Create /dev/ptmx symlink */
+ p = prefix_roota(dest, "/dev/ptmx");
+ if (symlink("pts/ptmx", p) < 0)
+ return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m");
+ if (userns_lchown(p, 0, 0) < 0)
+ return log_error_errno(errno, "Failed to chown /dev/ptmx: %m");
- r = sd_netlink_message_open_container(m, VETH_INFO_PEER);
- if (r < 0)
- return log_error_errno(r, "Failed to open netlink container: %m");
+ /* And fix /dev/pts/ptmx ownership */
+ p = prefix_roota(dest, "/dev/pts/ptmx");
+ if (userns_lchown(p, 0, 0) < 0)
+ return log_error_errno(errno, "Failed to chown /dev/pts/ptmx: %m");
- r = sd_netlink_message_append_string(m, IFLA_IFNAME, "host0");
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink interface name: %m");
+ return 0;
+}
- r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, &mac_container);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink MAC address: %m");
+static int setup_dev_console(const char *dest, const char *console) {
+ _cleanup_umask_ mode_t u;
+ const char *to;
+ int r;
- r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink namespace field: %m");
+ assert(dest);
+ assert(console);
- r = sd_netlink_message_close_container(m);
- if (r < 0)
- return log_error_errno(r, "Failed to close netlink container: %m");
+ u = umask(0000);
- r = sd_netlink_message_close_container(m);
+ r = chmod_and_chown(console, 0600, arg_uid_shift, arg_uid_shift);
if (r < 0)
- return log_error_errno(r, "Failed to close netlink container: %m");
+ return log_error_errno(r, "Failed to correct access mode for TTY: %m");
- r = sd_netlink_message_close_container(m);
- if (r < 0)
- return log_error_errno(r, "Failed to close netlink container: %m");
+ /* We need to bind mount the right tty to /dev/console since
+ * ptys can only exist on pts file systems. To have something
+ * to bind mount things on we create a empty regular file. */
- r = sd_netlink_call(rtnl, m, 0, NULL);
+ to = prefix_roota(dest, "/dev/console");
+ r = touch(to);
if (r < 0)
- return log_error_errno(r, "Failed to add new veth interfaces (host0, %s): %m", iface_name);
-
- i = (int) if_nametoindex(iface_name);
- if (i <= 0)
- return log_error_errno(errno, "Failed to resolve interface %s: %m", iface_name);
+ return log_error_errno(r, "touch() for /dev/console failed: %m");
- *ifi = i;
+ if (mount(console, to, NULL, MS_BIND, NULL) < 0)
+ return log_error_errno(errno, "Bind mount for /dev/console failed: %m");
return 0;
}
-static int setup_bridge(const char veth_name[], int *ifi) {
- _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL;
- _cleanup_netlink_unref_ sd_netlink *rtnl = NULL;
- int r, bridge;
+static int setup_kmsg(const char *dest, int kmsg_socket) {
+ const char *from, *to;
+ _cleanup_umask_ mode_t u;
+ int fd, r;
- if (!arg_private_network)
- return 0;
+ assert(kmsg_socket >= 0);
- if (!arg_network_veth)
- return 0;
+ u = umask(0000);
- if (!arg_network_bridge)
- return 0;
+ /* We create the kmsg FIFO as /run/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. */
+ from = prefix_roota(dest, "/run/kmsg");
+ to = prefix_roota(dest, "/proc/kmsg");
- bridge = (int) if_nametoindex(arg_network_bridge);
- if (bridge <= 0)
- return log_error_errno(errno, "Failed to resolve interface %s: %m", arg_network_bridge);
+ if (mkfifo(from, 0600) < 0)
+ return log_error_errno(errno, "mkfifo() for /run/kmsg failed: %m");
+ if (mount(from, to, NULL, MS_BIND, NULL) < 0)
+ return log_error_errno(errno, "Bind mount for /proc/kmsg failed: %m");
- *ifi = bridge;
+ fd = open(from, O_RDWR|O_NDELAY|O_CLOEXEC);
+ if (fd < 0)
+ return log_error_errno(errno, "Failed to open fifo: %m");
- r = sd_netlink_open(&rtnl);
- if (r < 0)
- return log_error_errno(r, "Failed to connect to netlink: %m");
+ /* Store away the fd in the socket, so that it stays open as
+ * long as we run the child */
+ r = send_one_fd(kmsg_socket, fd, 0);
+ safe_close(fd);
- r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, 0);
if (r < 0)
- return log_error_errno(r, "Failed to allocate netlink message: %m");
+ return log_error_errno(r, "Failed to send FIFO fd: %m");
- r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP);
- if (r < 0)
- return log_error_errno(r, "Failed to set IFF_UP flag: %m");
+ /* And now make the FIFO unavailable as /run/kmsg... */
+ (void) unlink(from);
- r = sd_netlink_message_append_string(m, IFLA_IFNAME, veth_name);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink interface name field: %m");
+ return 0;
+}
- r = sd_netlink_message_append_u32(m, IFLA_MASTER, bridge);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink master field: %m");
+static int on_address_change(sd_netlink *rtnl, sd_netlink_message *m, void *userdata) {
+ union in_addr_union *exposed = userdata;
- r = sd_netlink_call(rtnl, m, 0, NULL);
- if (r < 0)
- return log_error_errno(r, "Failed to add veth interface to bridge: %m");
+ assert(rtnl);
+ assert(m);
+ assert(exposed);
+ expose_port_execute(rtnl, arg_expose_ports, exposed);
return 0;
}
-static int parse_interface(struct udev *udev, const char *name) {
- _cleanup_udev_device_unref_ struct udev_device *d = NULL;
- char ifi_str[2 + DECIMAL_STR_MAX(int)];
- int ifi;
-
- ifi = (int) if_nametoindex(name);
- if (ifi <= 0)
- return log_error_errno(errno, "Failed to resolve interface %s: %m", name);
+static int setup_hostname(void) {
- sprintf(ifi_str, "n%i", ifi);
- d = udev_device_new_from_device_id(udev, ifi_str);
- if (!d)
- return log_error_errno(errno, "Failed to get udev device for interface %s: %m", name);
+ if (arg_share_system)
+ return 0;
- if (udev_device_get_is_initialized(d) <= 0) {
- log_error("Network interface %s is not initialized yet.", name);
- return -EBUSY;
- }
+ if (sethostname_idempotent(arg_machine) < 0)
+ return -errno;
- return ifi;
+ return 0;
}
-static int move_network_interfaces(pid_t pid) {
- _cleanup_udev_unref_ struct udev *udev = NULL;
- _cleanup_netlink_unref_ sd_netlink *rtnl = NULL;
- char **i;
+static int setup_journal(const char *directory) {
+ sd_id128_t machine_id, this_id;
+ _cleanup_free_ char *b = NULL, *d = NULL;
+ const char *etc_machine_id, *p, *q;
+ char *id;
int r;
- if (!arg_private_network)
- return 0;
-
- if (strv_isempty(arg_network_interfaces))
+ /* Don't link journals in ephemeral mode */
+ if (arg_ephemeral)
return 0;
- r = sd_netlink_open(&rtnl);
- if (r < 0)
- return log_error_errno(r, "Failed to connect to netlink: %m");
-
- udev = udev_new();
- if (!udev) {
- log_error("Failed to connect to udev.");
- return -ENOMEM;
- }
+ etc_machine_id = prefix_roota(directory, "/etc/machine-id");
- STRV_FOREACH(i, arg_network_interfaces) {
- _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL;
- int ifi;
+ r = read_one_line_file(etc_machine_id, &b);
+ if (r == -ENOENT && arg_link_journal == LINK_AUTO)
+ return 0;
+ else if (r < 0)
+ return log_error_errno(r, "Failed to read machine ID from %s: %m", etc_machine_id);
- ifi = parse_interface(udev, *i);
- if (ifi < 0)
- return ifi;
+ id = strstrip(b);
+ if (isempty(id) && arg_link_journal == LINK_AUTO)
+ return 0;
- r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, ifi);
- if (r < 0)
- return log_error_errno(r, "Failed to allocate netlink message: %m");
+ /* Verify validity */
+ r = sd_id128_from_string(id, &machine_id);
+ if (r < 0)
+ return log_error_errno(r, "Failed to parse machine ID from %s: %m", etc_machine_id);
- r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid);
- if (r < 0)
- return log_error_errno(r, "Failed to append namespace PID to netlink message: %m");
+ r = sd_id128_get_machine(&this_id);
+ if (r < 0)
+ return log_error_errno(r, "Failed to retrieve machine ID: %m");
- r = sd_netlink_call(rtnl, m, 0, NULL);
- if (r < 0)
- return log_error_errno(r, "Failed to move interface %s to namespace: %m", *i);
+ if (sd_id128_equal(machine_id, this_id)) {
+ log_full(arg_link_journal == LINK_AUTO ? LOG_WARNING : LOG_ERR,
+ "Host and machine ids are equal (%s): refusing to link journals", id);
+ if (arg_link_journal == LINK_AUTO)
+ return 0;
+ return -EEXIST;
}
- return 0;
-}
-
-static int setup_macvlan(pid_t pid) {
- _cleanup_udev_unref_ struct udev *udev = NULL;
- _cleanup_netlink_unref_ sd_netlink *rtnl = NULL;
- unsigned idx = 0;
- char **i;
- int r;
-
- if (!arg_private_network)
- return 0;
-
- if (strv_isempty(arg_network_macvlan))
+ if (arg_link_journal == LINK_NO)
return 0;
- r = sd_netlink_open(&rtnl);
+ r = userns_mkdir(directory, "/var", 0755, 0, 0);
if (r < 0)
- return log_error_errno(r, "Failed to connect to netlink: %m");
-
- udev = udev_new();
- if (!udev) {
- log_error("Failed to connect to udev.");
- return -ENOMEM;
- }
-
- STRV_FOREACH(i, arg_network_macvlan) {
- _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL;
- _cleanup_free_ char *n = NULL;
- struct ether_addr mac;
- int ifi;
+ return log_error_errno(r, "Failed to create /var: %m");
- ifi = parse_interface(udev, *i);
- if (ifi < 0)
- return ifi;
+ r = userns_mkdir(directory, "/var/log", 0755, 0, 0);
+ if (r < 0)
+ return log_error_errno(r, "Failed to create /var/log: %m");
- r = generate_mac(&mac, MACVLAN_HASH_KEY, idx++);
- if (r < 0)
- return log_error_errno(r, "Failed to create MACVLAN MAC address: %m");
+ r = userns_mkdir(directory, "/var/log/journal", 0755, 0, 0);
+ if (r < 0)
+ return log_error_errno(r, "Failed to create /var/log/journal: %m");
- r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to allocate netlink message: %m");
+ p = strjoina("/var/log/journal/", id);
+ q = prefix_roota(directory, p);
- r = sd_netlink_message_append_u32(m, IFLA_LINK, ifi);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink interface index: %m");
+ if (path_is_mount_point(p, 0) > 0) {
+ if (arg_link_journal != LINK_AUTO) {
+ log_error("%s: already a mount point, refusing to use for journal", p);
+ return -EEXIST;
+ }
- n = strappend("mv-", *i);
- if (!n)
- return log_oom();
+ return 0;
+ }
- strshorten(n, IFNAMSIZ-1);
+ if (path_is_mount_point(q, 0) > 0) {
+ if (arg_link_journal != LINK_AUTO) {
+ log_error("%s: already a mount point, refusing to use for journal", q);
+ return -EEXIST;
+ }
- r = sd_netlink_message_append_string(m, IFLA_IFNAME, n);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink interface name: %m");
+ return 0;
+ }
- r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, &mac);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink MAC address: %m");
+ 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)) {
- r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink namespace field: %m");
+ r = userns_mkdir(directory, p, 0755, 0, 0);
+ if (r < 0)
+ log_warning_errno(errno, "Failed to create directory %s: %m", q);
+ return 0;
+ }
- r = sd_netlink_message_open_container(m, IFLA_LINKINFO);
- if (r < 0)
- return log_error_errno(r, "Failed to open netlink container: %m");
+ if (unlink(p) < 0)
+ return log_error_errno(errno, "Failed to remove symlink %s: %m", p);
+ } else if (r == -EINVAL) {
- r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, "macvlan");
- if (r < 0)
- return log_error_errno(r, "Failed to open netlink container: %m");
+ if (arg_link_journal == LINK_GUEST &&
+ rmdir(p) < 0) {
- r = sd_netlink_message_append_u32(m, IFLA_MACVLAN_MODE, MACVLAN_MODE_BRIDGE);
- if (r < 0)
- return log_error_errno(r, "Failed to append macvlan mode: %m");
+ if (errno == ENOTDIR) {
+ log_error("%s already exists and is neither a symlink nor a directory", p);
+ return r;
+ } else {
+ log_error_errno(errno, "Failed to remove %s: %m", p);
+ return -errno;
+ }
+ }
+ } else if (r != -ENOENT) {
+ log_error_errno(errno, "readlink(%s) failed: %m", p);
+ return r;
+ }
- r = sd_netlink_message_close_container(m);
- if (r < 0)
- return log_error_errno(r, "Failed to close netlink container: %m");
+ if (arg_link_journal == LINK_GUEST) {
- r = sd_netlink_message_close_container(m);
- if (r < 0)
- return log_error_errno(r, "Failed to close netlink container: %m");
+ if (symlink(q, p) < 0) {
+ if (arg_link_journal_try) {
+ log_debug_errno(errno, "Failed to symlink %s to %s, skipping journal setup: %m", q, p);
+ return 0;
+ } else {
+ log_error_errno(errno, "Failed to symlink %s to %s: %m", q, p);
+ return -errno;
+ }
+ }
- r = sd_netlink_call(rtnl, m, 0, NULL);
+ r = userns_mkdir(directory, p, 0755, 0, 0);
if (r < 0)
- return log_error_errno(r, "Failed to add new macvlan interfaces: %m");
+ log_warning_errno(errno, "Failed to create directory %s: %m", q);
+ return 0;
}
- return 0;
-}
-
-static int setup_ipvlan(pid_t pid) {
- _cleanup_udev_unref_ struct udev *udev = NULL;
- _cleanup_netlink_unref_ sd_netlink *rtnl = NULL;
- char **i;
- int r;
-
- if (!arg_private_network)
- return 0;
+ if (arg_link_journal == LINK_HOST) {
+ /* don't create parents here -- if the host doesn't have
+ * permanent journal set up, don't force it here */
+ r = mkdir(p, 0755);
+ if (r < 0) {
+ if (arg_link_journal_try) {
+ log_debug_errno(errno, "Failed to create %s, skipping journal setup: %m", p);
+ return 0;
+ } else {
+ log_error_errno(errno, "Failed to create %s: %m", p);
+ return r;
+ }
+ }
- if (strv_isempty(arg_network_ipvlan))
+ } else if (access(p, F_OK) < 0)
return 0;
- r = sd_netlink_open(&rtnl);
- if (r < 0)
- return log_error_errno(r, "Failed to connect to netlink: %m");
+ if (dir_is_empty(q) == 0)
+ log_warning("%s is not empty, proceeding anyway.", q);
- udev = udev_new();
- if (!udev) {
- log_error("Failed to connect to udev.");
- return -ENOMEM;
+ r = userns_mkdir(directory, p, 0755, 0, 0);
+ if (r < 0) {
+ log_error_errno(errno, "Failed to create %s: %m", q);
+ return r;
}
- STRV_FOREACH(i, arg_network_ipvlan) {
- _cleanup_netlink_message_unref_ sd_netlink_message *m = NULL;
- _cleanup_free_ char *n = NULL;
- int ifi;
-
- ifi = parse_interface(udev, *i);
- if (ifi < 0)
- return ifi;
-
- r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to allocate netlink message: %m");
-
- r = sd_netlink_message_append_u32(m, IFLA_LINK, ifi);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink interface index: %m");
-
- n = strappend("iv-", *i);
- if (!n)
- return log_oom();
-
- strshorten(n, IFNAMSIZ-1);
+ if (mount(p, q, NULL, MS_BIND, NULL) < 0)
+ return log_error_errno(errno, "Failed to bind mount journal from host into guest: %m");
- r = sd_netlink_message_append_string(m, IFLA_IFNAME, n);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink interface name: %m");
+ return 0;
+}
- r = sd_netlink_message_append_u32(m, IFLA_NET_NS_PID, pid);
- if (r < 0)
- return log_error_errno(r, "Failed to add netlink namespace field: %m");
+static int drop_capabilities(void) {
+ return capability_bounding_set_drop(~arg_retain, false);
+}
- r = sd_netlink_message_open_container(m, IFLA_LINKINFO);
- if (r < 0)
- return log_error_errno(r, "Failed to open netlink container: %m");
+static int reset_audit_loginuid(void) {
+ _cleanup_free_ char *p = NULL;
+ int r;
- r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, "ipvlan");
- if (r < 0)
- return log_error_errno(r, "Failed to open netlink container: %m");
+ if (arg_share_system)
+ return 0;
- r = sd_netlink_message_append_u16(m, IFLA_IPVLAN_MODE, IPVLAN_MODE_L2);
- if (r < 0)
- return log_error_errno(r, "Failed to add ipvlan mode: %m");
+ r = read_one_line_file("/proc/self/loginuid", &p);
+ if (r == -ENOENT)
+ return 0;
+ if (r < 0)
+ return log_error_errno(r, "Failed to read /proc/self/loginuid: %m");
- r = sd_netlink_message_close_container(m);
- if (r < 0)
- return log_error_errno(r, "Failed to close netlink container: %m");
+ /* Already reset? */
+ if (streq(p, "4294967295"))
+ return 0;
- r = sd_netlink_message_close_container(m);
- if (r < 0)
- return log_error_errno(r, "Failed to close netlink container: %m");
+ r = write_string_file("/proc/self/loginuid", "4294967295", 0);
+ if (r < 0) {
+ log_error_errno(r,
+ "Failed to reset audit login UID. This probably means that your kernel is too\n"
+ "old and you have audit enabled. Note that the auditing subsystem is known to\n"
+ "be incompatible with containers on old kernels. Please make sure to upgrade\n"
+ "your kernel or to off auditing with 'audit=0' on the kernel command line before\n"
+ "using systemd-nspawn. Sleeping for 5s... (%m)");
- r = sd_netlink_call(rtnl, m, 0, NULL);
- if (r < 0)
- return log_error_errno(r, "Failed to add new ipvlan interfaces: %m");
+ sleep(5);
}
return 0;
log_debug_errno(errno, "Failed to remove loop %d: %m", nr);
}
-static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
- int pipe_fds[2];
- pid_t pid;
-
- assert(database);
- assert(key);
- assert(rpid);
-
- if (pipe2(pipe_fds, O_CLOEXEC) < 0)
- return log_error_errno(errno, "Failed to allocate pipe: %m");
-
- pid = fork();
- if (pid < 0)
- return log_error_errno(errno, "Failed to fork getent child: %m");
- else if (pid == 0) {
- int nullfd;
- char *empty_env = NULL;
-
- if (dup3(pipe_fds[1], STDOUT_FILENO, 0) < 0)
- _exit(EXIT_FAILURE);
-
- if (pipe_fds[0] > 2)
- safe_close(pipe_fds[0]);
- if (pipe_fds[1] > 2)
- safe_close(pipe_fds[1]);
-
- nullfd = open("/dev/null", O_RDWR);
- if (nullfd < 0)
- _exit(EXIT_FAILURE);
-
- if (dup3(nullfd, STDIN_FILENO, 0) < 0)
- _exit(EXIT_FAILURE);
-
- if (dup3(nullfd, STDERR_FILENO, 0) < 0)
- _exit(EXIT_FAILURE);
-
- if (nullfd > 2)
- safe_close(nullfd);
-
- (void) reset_all_signal_handlers();
- (void) reset_signal_mask();
- close_all_fds(NULL, 0);
-
- execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env);
- execle("/bin/getent", "getent", database, key, NULL, &empty_env);
- _exit(EXIT_FAILURE);
- }
-
- pipe_fds[1] = safe_close(pipe_fds[1]);
-
- *rpid = pid;
-
- return pipe_fds[0];
-}
-
-static int change_uid_gid(char **_home) {
- char line[LINE_MAX], *x, *u, *g, *h;
- const char *word, *state;
- _cleanup_free_ uid_t *uids = NULL;
- _cleanup_free_ char *home = NULL;
- _cleanup_fclose_ FILE *f = NULL;
- _cleanup_close_ int fd = -1;
- unsigned n_uids = 0;
- size_t sz = 0, l;
- uid_t uid;
- gid_t gid;
- pid_t pid;
- int r;
-
- assert(_home);
-
- if (!arg_user || streq(arg_user, "root") || streq(arg_user, "0")) {
- /* Reset everything fully to 0, just in case */
-
- r = reset_uid_gid();
- if (r < 0)
- return log_error_errno(r, "Failed to become root: %m");
-
- *_home = NULL;
- return 0;
- }
-
- /* First, get user credentials */
- fd = spawn_getent("passwd", arg_user, &pid);
- if (fd < 0)
- return fd;
-
- f = fdopen(fd, "r");
- if (!f)
- return log_oom();
- fd = -1;
-
- if (!fgets(line, sizeof(line), f)) {
-
- if (!ferror(f)) {
- log_error("Failed to resolve user %s.", arg_user);
- return -ESRCH;
- }
-
- log_error_errno(errno, "Failed to read from getent: %m");
- return -errno;
- }
-
- truncate_nl(line);
-
- wait_for_terminate_and_warn("getent passwd", pid, true);
-
- x = strchr(line, ':');
- if (!x) {
- log_error("/etc/passwd entry has invalid user field.");
- return -EIO;
- }
-
- u = strchr(x+1, ':');
- if (!u) {
- log_error("/etc/passwd entry has invalid password field.");
- return -EIO;
- }
-
- u++;
- g = strchr(u, ':');
- if (!g) {
- log_error("/etc/passwd entry has invalid UID field.");
- return -EIO;
- }
-
- *g = 0;
- g++;
- x = strchr(g, ':');
- if (!x) {
- log_error("/etc/passwd entry has invalid GID field.");
- return -EIO;
- }
-
- *x = 0;
- h = strchr(x+1, ':');
- if (!h) {
- log_error("/etc/passwd entry has invalid GECOS field.");
- return -EIO;
- }
-
- h++;
- x = strchr(h, ':');
- if (!x) {
- log_error("/etc/passwd entry has invalid home directory field.");
- return -EIO;
- }
-
- *x = 0;
-
- r = parse_uid(u, &uid);
- if (r < 0) {
- log_error("Failed to parse UID of user.");
- return -EIO;
- }
-
- r = parse_gid(g, &gid);
- if (r < 0) {
- log_error("Failed to parse GID of user.");
- return -EIO;
- }
-
- home = strdup(h);
- if (!home)
- return log_oom();
-
- /* Second, get group memberships */
- fd = spawn_getent("initgroups", arg_user, &pid);
- if (fd < 0)
- return fd;
-
- fclose(f);
- f = fdopen(fd, "r");
- if (!f)
- return log_oom();
- fd = -1;
-
- if (!fgets(line, sizeof(line), f)) {
- if (!ferror(f)) {
- log_error("Failed to resolve user %s.", arg_user);
- return -ESRCH;
- }
-
- log_error_errno(errno, "Failed to read from getent: %m");
- return -errno;
- }
-
- truncate_nl(line);
-
- wait_for_terminate_and_warn("getent initgroups", pid, true);
-
- /* Skip over the username and subsequent separator whitespace */
- x = line;
- x += strcspn(x, WHITESPACE);
- x += strspn(x, WHITESPACE);
-
- FOREACH_WORD(word, l, x, state) {
- char c[l+1];
-
- memcpy(c, word, l);
- c[l] = 0;
-
- if (!GREEDY_REALLOC(uids, sz, n_uids+1))
- return log_oom();
-
- r = parse_uid(c, &uids[n_uids++]);
- if (r < 0) {
- log_error("Failed to parse group data from getent.");
- return -EIO;
- }
- }
-
- r = mkdir_parents(home, 0775);
- if (r < 0)
- return log_error_errno(r, "Failed to make home root directory: %m");
-
- r = mkdir_safe(home, 0755, uid, gid);
- if (r < 0 && r != -EEXIST)
- return log_error_errno(r, "Failed to make home directory: %m");
-
- (void) fchown(STDIN_FILENO, uid, gid);
- (void) fchown(STDOUT_FILENO, uid, gid);
- (void) fchown(STDERR_FILENO, uid, gid);
-
- if (setgroups(n_uids, uids) < 0)
- return log_error_errno(errno, "Failed to set auxiliary groups: %m");
-
- if (setresgid(gid, gid, gid) < 0)
- return log_error_errno(errno, "setregid() failed: %m");
-
- if (setresuid(uid, uid, uid) < 0)
- return log_error_errno(errno, "setreuid() failed: %m");
-
- if (_home) {
- *_home = home;
- home = NULL;
- }
-
- return 0;
-}
-
/*
* Return values:
* < 0 : wait_for_terminate() failed to get the state of the
return r;
}
-static void nop_handler(int sig) {}
-
static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
pid_t pid;
}
if (i->type == IMAGE_RAW)
- r = set_sanitized_path(&arg_image, i->path);
+ r = free_and_strdup(&arg_image, i->path);
else
- r = set_sanitized_path(&arg_directory, i->path);
+ r = free_and_strdup(&arg_directory, i->path);
if (r < 0)
return log_error_errno(r, "Invalid image directory: %m");
}
}
- r = mount_all(NULL, true);
+ r = mount_all(NULL, arg_userns, true, arg_uid_shift, arg_private_network, arg_uid_range, arg_selinux_apifs_context);
+ if (r < 0)
+ return r;
+
+ r = mount_sysfs(NULL);
if (r < 0)
return r;
return -ESRCH;
}
- r = mount_systemd_cgroup_writable("");
+ r = mount_systemd_cgroup_writable("", arg_unified_cgroup_hierarchy);
if (r < 0)
return r;
if (arg_private_network)
loopback_setup();
- r = send_rtnl(rtnl_socket);
- if (r < 0)
- return r;
- rtnl_socket = safe_close(rtnl_socket);
+ if (arg_expose_ports) {
+ r = expose_port_send_rtnl(rtnl_socket);
+ if (r < 0)
+ return r;
+ rtnl_socket = safe_close(rtnl_socket);
+ }
if (drop_capabilities() < 0)
return log_error_errno(errno, "drop_capabilities() failed: %m");
return log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context);
#endif
- r = change_uid_gid(&home);
+ r = change_uid_gid(arg_user, &home);
if (r < 0)
return r;
if (mount(directory, directory, NULL, MS_BIND|MS_REC, NULL) < 0)
return log_error_errno(errno, "Failed to make bind mount: %m");
- r = setup_volatile(directory);
+ r = setup_volatile(directory, arg_volatile_mode, arg_userns, arg_uid_shift, arg_uid_range, arg_selinux_context);
if (r < 0)
return r;
- r = setup_volatile_state(directory);
+ r = setup_volatile_state(directory, arg_volatile_mode, arg_userns, arg_uid_shift, arg_uid_range, arg_selinux_context);
if (r < 0)
return r;
return log_error_errno(r, "Failed to make tree read-only: %m");
}
- r = mount_all(directory, false);
+ r = mount_all(directory, arg_userns, false, arg_private_network, arg_uid_shift, arg_uid_range, arg_selinux_apifs_context);
if (r < 0)
return r;
- if (copy_devnodes(directory) < 0)
+ r = copy_devnodes(directory);
+ if (r < 0)
return r;
dev_setup(directory, arg_uid_shift, arg_uid_shift);
- if (setup_pts(directory) < 0)
+ r = setup_pts(directory);
+ if (r < 0)
return r;
r = setup_propagate(directory);
if (r < 0)
return r;
- r = mount_custom(directory);
+ r = mount_custom(directory, arg_custom_mounts, arg_n_custom_mounts, arg_userns, arg_uid_shift, arg_uid_range, arg_selinux_apifs_context);
if (r < 0)
return r;
- r = mount_cgroups(directory);
+ r = mount_cgroups(directory, arg_unified_cgroup_hierarchy, arg_userns, arg_uid_shift, arg_uid_range, arg_selinux_apifs_context);
if (r < 0)
return r;
}
pid_socket = safe_close(pid_socket);
+ kmsg_socket = safe_close(kmsg_socket);
+ rtnl_socket = safe_close(rtnl_socket);
return 0;
}
return 0;
}
-static int chown_cgroup(pid_t pid) {
- _cleanup_free_ char *path = NULL, *fs = NULL;
- _cleanup_close_ int fd = -1;
- const char *fn;
- int r;
-
- r = cg_pid_get_path(NULL, pid, &path);
- if (r < 0)
- return log_error_errno(r, "Failed to get container cgroup path: %m");
-
- r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, path, NULL, &fs);
- if (r < 0)
- return log_error_errno(r, "Failed to get file system path for container cgroup: %m");
-
- fd = open(fs, O_RDONLY|O_CLOEXEC|O_DIRECTORY);
- if (fd < 0)
- return log_error_errno(errno, "Failed to open %s: %m", fs);
-
- FOREACH_STRING(fn,
- ".",
- "tasks",
- "notify_on_release",
- "cgroup.procs",
- "cgroup.clone_children",
- "cgroup.controllers",
- "cgroup.subtree_control",
- "cgroup.populated")
- if (fchownat(fd, fn, arg_uid_shift, arg_uid_shift, 0) < 0)
- log_full_errno(errno == ENOENT ? LOG_DEBUG : LOG_WARNING, errno,
- "Failed to chown() cgroup file %s, ignoring: %m", fn);
-
- return 0;
-}
-
-static int sync_cgroup(pid_t pid) {
- _cleanup_free_ char *cgroup = NULL;
- char tree[] = "/tmp/unifiedXXXXXX", pid_string[DECIMAL_STR_MAX(pid) + 1];
- bool undo_mount = false;
- const char *fn;
- int unified, r;
-
- unified = cg_unified();
- if (unified < 0)
- return log_error_errno(unified, "Failed to determine whether the unified hierachy is used: %m");
-
- if ((unified > 0) == arg_unified_cgroup_hierarchy)
- return 0;
-
- /* When the host uses the legacy cgroup setup, but the
- * container shall use the unified hierarchy, let's make sure
- * we copy the path from the name=systemd hierarchy into the
- * unified hierarchy. Similar for the reverse situation. */
-
- r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, pid, &cgroup);
- if (r < 0)
- return log_error_errno(r, "Failed to get control group of " PID_FMT ": %m", pid);
-
- /* In order to access the unified hierarchy we need to mount it */
- if (!mkdtemp(tree))
- return log_error_errno(errno, "Failed to generate temporary mount point for unified hierarchy: %m");
-
- if (unified)
- r = mount("cgroup", tree, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, "none,name=systemd,xattr");
- else
- r = mount("cgroup", tree, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, "__DEVEL__sane_behavior");
- if (r < 0) {
- r = log_error_errno(errno, "Failed to mount unified hierarchy: %m");
- goto finish;
- }
-
- undo_mount = true;
-
- fn = strjoina(tree, cgroup, "/cgroup.procs");
- (void) mkdir_parents(fn, 0755);
-
- sprintf(pid_string, PID_FMT, pid);
- r = write_string_file(fn, pid_string, 0);
- if (r < 0)
- log_error_errno(r, "Failed to move process: %m");
-
-finish:
- if (undo_mount)
- (void) umount(tree);
-
- (void) rmdir(tree);
- return r;
-}
-
-static int create_subcgroup(pid_t pid) {
- _cleanup_free_ char *cgroup = NULL;
- const char *child;
- int unified, r;
- CGroupMask supported;
-
- /* In the unified hierarchy inner nodes may only only contain
- * subgroups, but not processes. Hence, if we running in the
- * unified hierarchy and the container does the same, and we
- * did not create a scope unit for the container move us and
- * the container into two separate subcgroups. */
-
- if (!arg_keep_unit)
- return 0;
-
- if (!arg_unified_cgroup_hierarchy)
- return 0;
-
- unified = cg_unified();
- if (unified < 0)
- return log_error_errno(unified, "Failed to determine whether the unified hierachy is used: %m");
- if (unified == 0)
- return 0;
-
- r = cg_mask_supported(&supported);
- if (r < 0)
- return log_error_errno(r, "Failed to determine supported controllers: %m");
-
- r = cg_pid_get_path(SYSTEMD_CGROUP_CONTROLLER, 0, &cgroup);
- if (r < 0)
- return log_error_errno(r, "Failed to get our control group: %m");
-
- child = strjoina(cgroup, "/payload");
- r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, child, pid);
- if (r < 0)
- return log_error_errno(r, "Failed to create %s subcgroup: %m", child);
-
- child = strjoina(cgroup, "/supervisor");
- r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, child, 0);
- if (r < 0)
- return log_error_errno(r, "Failed to create %s subcgroup: %m", child);
-
- /* Try to enable as many controllers as possible for the new payload. */
- (void) cg_enable_everywhere(supported, supported, cgroup);
- return 0;
-}
-
static int load_settings(void) {
_cleanup_(settings_freep) Settings *settings = NULL;
_cleanup_fclose_ FILE *f = NULL;
}
if ((arg_settings_mask & SETTING_CAPABILITY) == 0) {
+ uint64_t plus;
- if (!arg_settings_trusted && settings->capability != 0)
- log_warning("Ignoring Capability= setting, file %s is not trusted.", p);
- else
- arg_retain |= settings->capability;
+ plus = settings->capability;
+ if (settings_private_network(settings))
+ plus |= (1ULL << CAP_NET_ADMIN);
+
+ if (!arg_settings_trusted && plus != 0) {
+ if (settings->capability != 0)
+ log_warning("Ignoring Capability= setting, file %s is not trusted.", p);
+ } else
+ arg_retain |= plus;
arg_retain &= ~settings->drop_capability;
}
if (!arg_settings_trusted)
log_warning("Ignoring network settings, file %s is not trusted.", p);
else {
+ arg_network_veth = settings_private_network(settings);
+ arg_private_network = settings_private_network(settings);
+
strv_free(arg_network_interfaces);
arg_network_interfaces = settings->network_interfaces;
settings->network_interfaces = NULL;
free(arg_network_bridge);
arg_network_bridge = settings->network_bridge;
settings->network_bridge = NULL;
-
- arg_network_veth = settings->network_veth > 0 || settings->network_bridge;
-
- arg_private_network = true; /* all these settings imply private networking */
}
}
goto finish;
}
- r = btrfs_subvol_snapshot(arg_directory, np, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE);
+ r = btrfs_subvol_snapshot(arg_directory, np, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE | BTRFS_SNAPSHOT_QUOTA);
if (r < 0) {
log_error_errno(r, "Failed to create snapshot %s from %s: %m", np, arg_directory);
goto finish;
}
if (arg_template) {
- r = btrfs_subvol_snapshot(arg_template, arg_directory, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE);
+ r = btrfs_subvol_snapshot(arg_template, arg_directory, (arg_read_only ? BTRFS_SNAPSHOT_READ_ONLY : 0) | BTRFS_SNAPSHOT_FALLBACK_COPY | BTRFS_SNAPSHOT_RECURSIVE | BTRFS_SNAPSHOT_QUOTA);
if (r == -EEXIST) {
if (!arg_quiet)
log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template);
} else {
const char *p;
- p = strjoina(arg_directory,
- argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/");
- if (access(p, F_OK) < 0) {
- log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory);
+ p = strjoina(arg_directory, "/usr/");
+ if (laccess(p, F_OK) < 0) {
+ log_error("Directory %s doesn't look like it has an OS tree. Refusing.", arg_directory);
r = -EINVAL;
goto finish;
}
ContainerStatus container_status;
_cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL;
static const struct sigaction sa = {
- .sa_handler = nop_handler,
+ .sa_handler = nop_signal_handler,
.sa_flags = SA_NOCLDSTOP,
};
int ifi = 0;
goto finish;
}
- if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
+ if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
r = log_error_errno(errno, "Failed to create kmsg socket pair: %m");
goto finish;
}
- if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0) {
+ if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0) {
r = log_error_errno(errno, "Failed to create rtnl socket pair: %m");
goto finish;
}
- if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, pid_socket_pair) < 0) {
+ if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, pid_socket_pair) < 0) {
r = log_error_errno(errno, "Failed to create pid socket pair: %m");
goto finish;
}
if (arg_userns)
- if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, uid_shift_socket_pair) < 0) {
+ if (socketpair(AF_UNIX, SOCK_SEQPACKET|SOCK_CLOEXEC, 0, uid_shift_socket_pair) < 0) {
r = log_error_errno(errno, "Failed to create uid shift socket pair: %m");
goto finish;
}
barrier_set_role(&barrier, BARRIER_PARENT);
- fdset_free(fds);
- fds = NULL;
+ fds = fdset_free(fds);
kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]);
pid_socket_pair[1] = safe_close(pid_socket_pair[1]);
+ uid_shift_socket_pair[1] = safe_close(uid_shift_socket_pair[1]);
/* Wait for the outer child. */
r = wait_for_terminate_and_warn("namespace helper", pid, NULL);
goto finish;
}
if (l != sizeof(pid)) {
- log_error("Short read while reading inner child PID: %m");
+ log_error("Short read while reading inner child PID.");
r = EIO;
goto finish;
}
goto finish;
}
if (l != sizeof(arg_uid_shift)) {
- log_error("Short read while reading UID shift: %m");
+ log_error("Short read while reading UID shift.");
r = EIO;
goto finish;
}
(void) barrier_place(&barrier); /* #2 */
}
- r = move_network_interfaces(pid);
- if (r < 0)
- goto finish;
+ if (arg_private_network) {
- r = setup_veth(pid, veth_name, &ifi);
- if (r < 0)
- goto finish;
+ r = move_network_interfaces(pid, arg_network_interfaces);
+ if (r < 0)
+ goto finish;
- r = setup_bridge(veth_name, &ifi);
- if (r < 0)
- goto finish;
+ if (arg_network_veth) {
+ r = setup_veth(arg_machine, pid, veth_name, !!arg_network_bridge);
+ if (r < 0)
+ goto finish;
+ else if (r > 0)
+ ifi = r;
- r = setup_macvlan(pid);
- if (r < 0)
- goto finish;
+ if (arg_network_bridge) {
+ r = setup_bridge(veth_name, arg_network_bridge);
+ if (r < 0)
+ goto finish;
+ if (r > 0)
+ ifi = r;
+ }
+ }
- r = setup_ipvlan(pid);
- if (r < 0)
- goto finish;
+ r = setup_macvlan(arg_machine, pid, arg_network_macvlan);
+ if (r < 0)
+ goto finish;
- r = register_machine(pid, ifi);
- if (r < 0)
- goto finish;
+ r = setup_ipvlan(arg_machine, pid, arg_network_ipvlan);
+ if (r < 0)
+ goto finish;
+ }
- r = sync_cgroup(pid);
- if (r < 0)
- goto finish;
+ if (arg_register) {
+ r = register_machine(
+ arg_machine,
+ pid,
+ arg_directory,
+ arg_uuid,
+ ifi,
+ arg_slice,
+ arg_custom_mounts, arg_n_custom_mounts,
+ arg_kill_signal,
+ arg_property,
+ arg_keep_unit);
+ if (r < 0)
+ goto finish;
+ }
- r = create_subcgroup(pid);
+ r = sync_cgroup(pid, arg_unified_cgroup_hierarchy);
if (r < 0)
goto finish;
- r = chown_cgroup(pid);
+ if (arg_keep_unit) {
+ r = create_subcgroup(pid, arg_unified_cgroup_hierarchy);
+ if (r < 0)
+ goto finish;
+ }
+
+ r = chown_cgroup(pid, arg_uid_shift);
if (r < 0)
goto finish;
}
/* Let the child know that we are ready and wait that the child is completely ready now. */
- if (!barrier_place_and_sync(&barrier)) { /* #5 */
- log_error("Client died too early.");
+ if (!barrier_place_and_sync(&barrier)) { /* #4 */
+ log_error("Child died too early.");
r = -ESRCH;
goto finish;
}
sd_event_add_signal(event, NULL, SIGCHLD, NULL, NULL);
if (arg_expose_ports) {
- r = watch_rtnl(event, rtnl_socket_pair[0], &exposed, &rtnl);
+ r = expose_port_watch_rtnl(event, rtnl_socket_pair[0], on_address_change, &exposed, &rtnl);
if (r < 0)
goto finish;
- (void) expose_ports(rtnl, &exposed);
+ (void) expose_port_execute(rtnl, arg_expose_ports, &exposed);
}
rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]);
- r = pty_forward_new(event, master, true, !interactive, &forward);
+ r = pty_forward_new(event, master, PTY_FORWARD_IGNORE_VHANGUP | (interactive ? 0 : PTY_FORWARD_READ_ONLY), &forward);
if (r < 0) {
log_error_errno(r, "Failed to create PTY forwarder: %m");
goto finish;
putc('\n', stdout);
/* Kill if it is not dead yet anyway */
- terminate_machine(pid);
+ if (arg_register && !arg_keep_unit)
+ terminate_machine(pid);
/* Normally redundant, but better safe than sorry */
kill(pid, SIGKILL);
break;
}
- flush_ports(&exposed);
+ expose_port_flush(arg_expose_ports, &exposed);
}
finish:
/* Try to flush whatever is still queued in the pty */
if (master >= 0)
- (void) copy_bytes(master, STDOUT_FILENO, (off_t) -1, false);
+ (void) copy_bytes(master, STDOUT_FILENO, (uint64_t) -1, false);
loop_remove(loop_nr, &image_fd);
if (remove_subvol && arg_directory) {
int k;
- k = btrfs_subvol_remove(arg_directory, true);
+ k = btrfs_subvol_remove(arg_directory, BTRFS_REMOVE_RECURSIVE|BTRFS_REMOVE_QUOTA);
if (k < 0)
log_warning_errno(k, "Cannot remove subvolume '%s', ignoring: %m", arg_directory);
}
(void) rm_rf(p, REMOVE_ROOT);
}
- flush_ports(&exposed);
+ expose_port_flush(arg_expose_ports, &exposed);
free(arg_directory);
free(arg_template);
projects / thirdparty / systemd.git / blobdiff