diff options
Diffstat (limited to 'src/nspawn')
| -rw-r--r-- | src/nspawn/nspawn-mount.c | 214 | ||||
| -rw-r--r-- | src/nspawn/nspawn-mount.h | 2 | ||||
| -rw-r--r-- | src/nspawn/nspawn-register.c | 17 | ||||
| -rw-r--r-- | src/nspawn/nspawn-seccomp.c | 6 | ||||
| -rw-r--r-- | src/nspawn/nspawn-setuid.c | 7 | ||||
| -rw-r--r-- | src/nspawn/nspawn.c | 186 |
6 files changed, 319 insertions, 113 deletions
diff --git a/src/nspawn/nspawn-mount.c b/src/nspawn/nspawn-mount.c index 9f4903c842..803caef3dd 100644 --- a/src/nspawn/nspawn-mount.c +++ b/src/nspawn/nspawn-mount.c @@ -23,6 +23,8 @@ #include "alloc-util.h" #include "cgroup-util.h" #include "escape.h" +#include "fd-util.h" +#include "fileio.h" #include "fs-util.h" #include "label.h" #include "mkdir.h" @@ -181,13 +183,15 @@ int tmpfs_mount_parse(CustomMount **l, unsigned *n, const char *s) { static int tmpfs_patch_options( const char *options, - bool userns, uid_t uid_shift, uid_t uid_range, + bool userns, + uid_t uid_shift, uid_t uid_range, + bool patch_ids, const char *selinux_apifs_context, char **ret) { char *buf = NULL; - if (userns && uid_shift != 0) { + if ((userns && uid_shift != 0) || patch_ids) { assert(uid_shift != UID_INVALID); if (options) @@ -218,7 +222,13 @@ static int tmpfs_patch_options( } #endif + if (!buf && options) { + buf = strdup(options); + if (!buf) + return -ENOMEM; + } *ret = buf; + return !!buf; } @@ -271,7 +281,15 @@ int mount_sysfs(const char *dest) { return log_error_errno(errno, "Failed to remove %s: %m", full); x = prefix_roota(top, "/fs/kdbus"); - (void) mkdir(x, 0755); + (void) mkdir_p(x, 0755); + + /* Create mountpoint for cgroups. Otherwise we are not allowed since we + * remount /sys read-only. + */ + if (cg_ns_supported()) { + x = prefix_roota(top, "/fs/cgroup"); + (void) mkdir_p(x, 0755); + } if (mount(NULL, top, NULL, MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_REMOUNT, NULL) < 0) return log_error_errno(errno, "Failed to make %s read-only: %m", top); @@ -297,18 +315,19 @@ int mount_all(const char *dest, } MountPoint; static const MountPoint mount_table[] = { - { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true, true, false }, - { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true, true, false }, /* Bind mount first */ - { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_REMOUNT, true, true, false }, /* Then, make it r/o */ - { "tmpfs", "/sys", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV, true, false, true }, - { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true, false, false }, - { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true, false, false }, - { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false, false }, - { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false, false }, - { "tmpfs", "/tmp", "tmpfs", "mode=1777", MS_STRICTATIME, true, false, false }, + { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true, true, false }, + { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true, true, false }, /* Bind mount first ...*/ + { "/proc/sys/net", "/proc/sys/net", NULL, NULL, MS_BIND, true, true, true }, /* (except for this) */ + { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_REMOUNT, true, true, false }, /* ... then, make it r/o */ + { "tmpfs", "/sys", "tmpfs", "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV, true, false, true }, + { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true, false, false }, + { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true, false, false }, + { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false, false }, + { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true, false, false }, + { "tmpfs", "/tmp", "tmpfs", "mode=1777", MS_STRICTATIME, true, false, false }, #ifdef HAVE_SELINUX - { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false, 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, false }, /* Then, make it r/o */ + { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false, 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, false }, /* Then, make it r/o */ #endif }; @@ -348,7 +367,7 @@ int mount_all(const char *dest, o = mount_table[k].options; if (streq_ptr(mount_table[k].type, "tmpfs")) { - r = tmpfs_patch_options(o, use_userns, uid_shift, uid_range, selinux_apifs_context, &options); + r = tmpfs_patch_options(o, use_userns, uid_shift, uid_range, false, selinux_apifs_context, &options); if (r < 0) return log_oom(); if (r > 0) @@ -485,7 +504,7 @@ static int mount_tmpfs( 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, userns, uid_shift, uid_range, selinux_apifs_context, &buf); + r = tmpfs_patch_options(m->options, userns, uid_shift, uid_range, false, selinux_apifs_context, &buf); if (r < 0) return log_oom(); options = r > 0 ? buf : m->options; @@ -600,6 +619,48 @@ int mount_custom( return 0; } +/* Retrieve existing subsystems. This function is called in a new cgroup + * namespace. + */ +static int get_controllers(Set *subsystems) { + _cleanup_fclose_ FILE *f = NULL; + char line[LINE_MAX]; + + assert(subsystems); + + f = fopen("/proc/self/cgroup", "re"); + if (!f) + return errno == ENOENT ? -ESRCH : -errno; + + FOREACH_LINE(line, f, return -errno) { + int r; + char *e, *l, *p; + + truncate_nl(line); + + l = strchr(line, ':'); + if (!l) + continue; + + l++; + e = strchr(l, ':'); + if (!e) + continue; + + *e = 0; + + if (streq(l, "") || streq(l, "name=systemd")) + continue; + + p = strdup(l); + r = set_consume(subsystems, p); + if (r < 0) + return r; + } + + return 0; +} + static int mount_legacy_cgroup_hierarchy(const char *dest, const char *controller, const char *hierarchy, bool read_only) { char *to; int r; @@ -628,11 +689,107 @@ static int mount_legacy_cgroup_hierarchy(const char *dest, const char *controlle return 1; } -static int mount_legacy_cgroups( - const char *dest, +/* Mount a legacy cgroup hierarchy when cgroup namespaces are supported. */ +static int mount_legacy_cgns_supported( bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context) { + _cleanup_set_free_free_ Set *controllers = NULL; + const char *cgroup_root = "/sys/fs/cgroup", *c; + int r; + + (void) mkdir_p(cgroup_root, 0755); + + /* Mount a tmpfs to /sys/fs/cgroup if it's not mounted there yet. */ + r = path_is_mount_point(cgroup_root, AT_SYMLINK_FOLLOW); + 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; + + /* When cgroup namespaces are enabled and user namespaces are + * used then the mount of the cgroupfs is done *inside* the new + * user namespace. We're root in the new user namespace and the + * kernel will happily translate our uid/gid to the correct + * uid/gid as seen from e.g. /proc/1/mountinfo. So we simply + * pass uid 0 and not uid_shift to tmpfs_patch_options(). + */ + r = tmpfs_patch_options("mode=755", userns, 0, uid_range, true, selinux_apifs_context, &options); + if (r < 0) + return log_oom(); + 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"); + } + + if (cg_unified() > 0) + goto skip_controllers; + + controllers = set_new(&string_hash_ops); + if (!controllers) + return log_oom(); + + r = get_controllers(controllers); + if (r < 0) + return log_error_errno(r, "Failed to determine cgroup controllers: %m"); + + for (;;) { + _cleanup_free_ const char *controller = NULL; + + controller = set_steal_first(controllers); + if (!controller) + break; + + r = mount_legacy_cgroup_hierarchy("", controller, controller, !userns); + if (r < 0) + return r; + + /* When multiple hierarchies are co-mounted, make their + * constituting individual hierarchies a symlink to the + * co-mount. + */ + c = controller; + for (;;) { + _cleanup_free_ char *target = NULL, *tok = NULL; + + r = extract_first_word(&c, &tok, ",", 0); + if (r < 0) + return log_error_errno(r, "Failed to extract co-mounted cgroup controller: %m"); + if (r == 0) + break; + + target = prefix_root("/sys/fs/cgroup", tok); + if (!target) + return log_oom(); + + if (streq(controller, tok)) + break; + + r = symlink_idempotent(controller, target); + if (r == -EINVAL) + return log_error_errno(r, "Invalid existing symlink for combined hierarchy: %m"); + if (r < 0) + return log_error_errno(r, "Failed to create symlink for combined hierarchy: %m"); + } + } + +skip_controllers: + r = mount_legacy_cgroup_hierarchy("", "none,name=systemd,xattr", "systemd", false); + if (r < 0) + return r; + + if (!userns) { + 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; +} + +/* Mount legacy cgroup hierarchy when cgroup namespaces are unsupported. */ +static int mount_legacy_cgns_unsupported( + const char *dest, + bool userns, uid_t uid_shift, uid_t uid_range, + const char *selinux_apifs_context) { _cleanup_set_free_free_ Set *controllers = NULL; const char *cgroup_root; int r; @@ -648,7 +805,7 @@ static int mount_legacy_cgroups( if (r == 0) { _cleanup_free_ char *options = NULL; - r = tmpfs_patch_options("mode=755", userns, uid_shift, uid_range, selinux_apifs_context, &options); + r = tmpfs_patch_options("mode=755", userns, uid_shift, uid_range, false, selinux_apifs_context, &options); if (r < 0) return log_oom(); @@ -707,10 +864,8 @@ static int mount_legacy_cgroups( return r; r = symlink_idempotent(combined, target); - if (r == -EINVAL) { - log_error("Invalid existing symlink for combined hierarchy"); - return r; - } + if (r == -EINVAL) + return log_error_errno(r, "Invalid existing symlink for combined hierarchy: %m"); if (r < 0) return log_error_errno(r, "Failed to create symlink for combined hierarchy: %m"); } @@ -761,12 +916,15 @@ int mount_cgroups( const char *dest, bool unified_requested, bool userns, uid_t uid_shift, uid_t uid_range, - const char *selinux_apifs_context) { + const char *selinux_apifs_context, + bool use_cgns) { if (unified_requested) return mount_unified_cgroups(dest); - else - return mount_legacy_cgroups(dest, userns, uid_shift, uid_range, selinux_apifs_context); + else if (use_cgns && cg_ns_supported()) + return mount_legacy_cgns_supported(userns, uid_shift, uid_range, selinux_apifs_context); + + return mount_legacy_cgns_unsupported(dest, userns, uid_shift, uid_range, selinux_apifs_context); } int mount_systemd_cgroup_writable( @@ -834,7 +992,7 @@ int setup_volatile_state( return log_error_errno(errno, "Failed to create %s: %m", directory); options = "mode=755"; - r = tmpfs_patch_options(options, userns, uid_shift, uid_range, selinux_apifs_context, &buf); + r = tmpfs_patch_options(options, userns, uid_shift, uid_range, false, selinux_apifs_context, &buf); if (r < 0) return log_oom(); if (r > 0) @@ -870,7 +1028,7 @@ int setup_volatile( return log_error_errno(errno, "Failed to create temporary directory: %m"); options = "mode=755"; - r = tmpfs_patch_options(options, userns, uid_shift, uid_range, selinux_apifs_context, &buf); + r = tmpfs_patch_options(options, userns, uid_shift, uid_range, false, selinux_apifs_context, &buf); if (r < 0) return log_oom(); if (r > 0) diff --git a/src/nspawn/nspawn-mount.h b/src/nspawn/nspawn-mount.h index 0daf145412..0eff8e1006 100644 --- a/src/nspawn/nspawn-mount.h +++ b/src/nspawn/nspawn-mount.h @@ -58,7 +58,7 @@ int custom_mount_compare(const void *a, const void *b); int mount_all(const char *dest, bool use_userns, bool in_userns, bool use_netns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context); int mount_sysfs(const char *dest); -int mount_cgroups(const char *dest, bool unified_requested, bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context); +int mount_cgroups(const char *dest, bool unified_requested, bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context, bool use_cgns); int mount_systemd_cgroup_writable(const char *dest, bool unified_requested); int mount_custom(const char *dest, CustomMount *mounts, unsigned n, bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context); diff --git a/src/nspawn/nspawn-register.c b/src/nspawn/nspawn-register.c index 7fd711b8a4..e5b76a0c5d 100644 --- a/src/nspawn/nspawn-register.c +++ b/src/nspawn/nspawn-register.c @@ -104,7 +104,7 @@ int register_machine( return bus_log_create_error(r); } - r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict"); + r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "closed"); if (r < 0) return bus_log_create_error(r); @@ -112,31 +112,20 @@ int register_machine( * 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)", 11, + r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 2, /* 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", - /* Allow /run/systemd/inaccessible/{chr,blk} - * devices inside the container */ - "/run/systemd/inaccessible/chr", "rwm", - "/run/systemd/inaccessible/blk", "rwm"); + "char-pts", "rw"); if (r < 0) return bus_log_create_error(r); diff --git a/src/nspawn/nspawn-seccomp.c b/src/nspawn/nspawn-seccomp.c index 54db1b47f8..3ab7160ebe 100644 --- a/src/nspawn/nspawn-seccomp.c +++ b/src/nspawn/nspawn-seccomp.c @@ -119,10 +119,8 @@ static int seccomp_add_default_syscall_filter(scmp_filter_ctx ctx, r = seccomp_rule_add(ctx, SCMP_ACT_ERRNO(EPERM), blacklist[i].syscall_num, 0); if (r == -EFAULT) continue; /* unknown syscall */ - if (r < 0) { - log_error_errno(r, "Failed to block syscall: %m"); - return r; - } + if (r < 0) + return log_error_errno(r, "Failed to block syscall: %m"); } return 0; diff --git a/src/nspawn/nspawn-setuid.c b/src/nspawn/nspawn-setuid.c index ee15a47e93..b8e8e091c8 100644 --- a/src/nspawn/nspawn-setuid.c +++ b/src/nspawn/nspawn-setuid.c @@ -124,14 +124,12 @@ int change_uid_gid(const char *user, char **_home) { fd = -1; if (!fgets(line, sizeof(line), f)) { - if (!ferror(f)) { log_error("Failed to resolve user %s.", user); return -ESRCH; } - log_error_errno(errno, "Failed to read from getent: %m"); - return -errno; + return log_error_errno(errno, "Failed to read from getent: %m"); } truncate_nl(line); @@ -214,8 +212,7 @@ int change_uid_gid(const char *user, char **_home) { return -ESRCH; } - log_error_errno(errno, "Failed to read from getent: %m"); - return -errno; + return log_error_errno(errno, "Failed to read from getent: %m"); } truncate_nl(line); diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index e4be0a2251..6cc1b9177d 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -61,9 +61,9 @@ #include "fs-util.h" #include "gpt.h" #include "hostname-util.h" +#include "id128-util.h" #include "log.h" #include "loopback-setup.h" -#include "machine-id-setup.h" #include "machine-image.h" #include "macro.h" #include "missing.h" @@ -76,10 +76,10 @@ #include "nspawn-network.h" #include "nspawn-patch-uid.h" #include "nspawn-register.h" +#include "nspawn-seccomp.h" #include "nspawn-settings.h" #include "nspawn-setuid.h" #include "nspawn-stub-pid1.h" -#include "nspawn-seccomp.h" #include "parse-util.h" #include "path-util.h" #include "process-util.h" @@ -101,9 +101,11 @@ #include "util.h" /* Note that devpts's gid= parameter parses GIDs as signed values, hence we stay away from the upper half of the 32bit - * UID range here */ + * UID range here. We leave a bit of room at the lower end and a lot of room at the upper end, so that other subsystems + * may have their own allocation ranges too. */ #define UID_SHIFT_PICK_MIN ((uid_t) UINT32_C(0x00080000)) #define UID_SHIFT_PICK_MAX ((uid_t) UINT32_C(0x6FFF0000)) + /* nspawn is listening on the socket at the path in the constant nspawn_notify_socket_path * nspawn_notify_socket_path is relative to the container * the init process in the container pid can send messages to nspawn following the sd_notify(3) protocol */ @@ -192,6 +194,7 @@ static int arg_settings_trusted = -1; static char **arg_parameters = NULL; static const char *arg_container_service_name = "systemd-nspawn"; static bool arg_notify_ready = false; +static bool arg_use_cgns = true; static void help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" @@ -277,7 +280,6 @@ static void help(void) { , program_invocation_short_name); } - static int custom_mounts_prepare(void) { unsigned i; int r; @@ -593,9 +595,12 @@ static int parse_argv(int argc, char *argv[]) { case ARG_UUID: r = sd_id128_from_string(optarg, &arg_uuid); - if (r < 0) { - log_error("Invalid UUID: %s", optarg); - return r; + if (r < 0) + return log_error_errno(r, "Invalid UUID: %s", optarg); + + if (sd_id128_is_null(arg_uuid)) { + log_error("Machine UUID may not be all zeroes."); + return -EINVAL; } arg_settings_mask |= SETTING_MACHINE_ID; @@ -1100,6 +1105,12 @@ static int parse_argv(int argc, char *argv[]) { if (e) arg_container_service_name = e; + r = getenv_bool("SYSTEMD_NSPAWN_USE_CGNS"); + if (r < 0) + arg_use_cgns = cg_ns_supported(); + else + arg_use_cgns = r; + return 1; } @@ -1265,20 +1276,9 @@ static int setup_resolv_conf(const char *dest) { return 0; } -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) { + sd_id128_t rnd = SD_ID128_NULL; const char *from, *to; - sd_id128_t rnd = {}; - char as_uuid[37]; int r; if (arg_share_system) @@ -1294,18 +1294,16 @@ static int setup_boot_id(const char *dest) { 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); + r = id128_write(from, ID128_UUID, rnd, false); 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"); + log_warning_errno(errno, "Failed to make boot id read-only, ignoring: %m"); - unlink(from); + (void) unlink(from); return r; } @@ -1803,17 +1801,18 @@ static int dissect_image( char **root_device, bool *root_device_rw, char **home_device, bool *home_device_rw, char **srv_device, bool *srv_device_rw, + char **esp_device, bool *secondary) { #ifdef HAVE_BLKID - int home_nr = -1, srv_nr = -1; + int home_nr = -1, srv_nr = -1, esp_nr = -1; #ifdef GPT_ROOT_NATIVE int root_nr = -1; #endif #ifdef GPT_ROOT_SECONDARY int secondary_root_nr = -1; #endif - _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *generic = NULL; + _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *esp = NULL, *generic = NULL; _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL; _cleanup_udev_device_unref_ struct udev_device *d = NULL; _cleanup_blkid_free_probe_ blkid_probe b = NULL; @@ -1831,6 +1830,7 @@ static int dissect_image( assert(root_device); assert(home_device); assert(srv_device); + assert(esp_device); assert(secondary); assert(arg_image); @@ -2044,6 +2044,16 @@ static int dissect_image( r = free_and_strdup(&srv, node); if (r < 0) return log_oom(); + } else if (sd_id128_equal(type_id, GPT_ESP)) { + + if (esp && nr >= esp_nr) + continue; + + esp_nr = nr; + + r = free_and_strdup(&esp, node); + if (r < 0) + return log_oom(); } #ifdef GPT_ROOT_NATIVE else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { @@ -2161,6 +2171,11 @@ static int dissect_image( *srv_device_rw = srv_rw; } + if (esp) { + *esp_device = esp; + esp = NULL; + } + return 0; #else log_error("--image= is not supported, compiled without blkid support."); @@ -2231,33 +2246,37 @@ static int mount_device(const char *what, const char *where, const char *directo } static int setup_machine_id(const char *directory) { + const char *etc_machine_id; + sd_id128_t id; int r; - const char *etc_machine_id, *t; - _cleanup_free_ char *s = NULL; - etc_machine_id = prefix_roota(directory, "/etc/machine-id"); + /* If the UUID in the container is already set, then that's what counts, and we use. If it isn't set, and the + * caller passed --uuid=, then we'll pass it in the $container_uuid env var to PID 1 of the container. The + * assumption is that PID 1 will then write it to /etc/machine-id to make it persistent. If --uuid= is not + * passed we generate a random UUID, and pass it via $container_uuid. In effect this means that /etc/machine-id + * in the container and our idea of the container UUID will always be in sync (at least if PID 1 in the + * container behaves nicely). */ - r = read_one_line_file(etc_machine_id, &s); - if (r < 0) - return log_error_errno(r, "Failed to read machine ID from %s: %m", etc_machine_id); + etc_machine_id = prefix_roota(directory, "/etc/machine-id"); - t = strstrip(s); + r = id128_read(etc_machine_id, ID128_PLAIN, &id); + if (r < 0) { + if (!IN_SET(r, -ENOENT, -ENOMEDIUM)) /* If the file is missing or empty, we don't mind */ + return log_error_errno(r, "Failed to read machine ID from container image: %m"); - if (!isempty(t)) { - r = sd_id128_from_string(t, &arg_uuid); - if (r < 0) - return log_error_errno(r, "Failed to parse machine ID from %s: %m", etc_machine_id); - } else { if (sd_id128_is_null(arg_uuid)) { r = sd_id128_randomize(&arg_uuid); if (r < 0) - return log_error_errno(r, "Failed to generate random machine ID: %m"); + return log_error_errno(r, "Failed to acquire randomized machine UUID: %m"); + } + } else { + if (sd_id128_is_null(id)) { + log_error("Machine ID in container image is zero, refusing."); + return -EINVAL; } - } - r = machine_id_setup(directory, arg_uuid); - if (r < 0) - return log_error_errno(r, "Failed to setup machine ID: %m"); + arg_uuid = id; + } return 0; } @@ -2289,7 +2308,8 @@ static int mount_devices( const char *where, const char *root_device, bool root_device_rw, const char *home_device, bool home_device_rw, - const char *srv_device, bool srv_device_rw) { + const char *srv_device, bool srv_device_rw, + const char *esp_device) { int r; assert(where); @@ -2312,6 +2332,27 @@ static int mount_devices( return log_error_errno(r, "Failed to mount server data directory: %m"); } + if (esp_device) { + const char *mp, *x; + + /* Mount the ESP to /efi if it exists and is empty. If it doesn't exist, use /boot instead. */ + + mp = "/efi"; + x = strjoina(arg_directory, mp); + r = dir_is_empty(x); + if (r == -ENOENT) { + mp = "/boot"; + x = strjoina(arg_directory, mp); + r = dir_is_empty(x); + } + + if (r > 0) { + r = mount_device(esp_device, arg_directory, mp, true); + if (r < 0) + return log_error_errno(r, "Failed to mount ESP: %m"); + } + } + return 0; } @@ -2594,9 +2635,25 @@ static int inner_child( return -ESRCH; } - r = mount_systemd_cgroup_writable("", arg_unified_cgroup_hierarchy); - if (r < 0) - return r; + if (arg_use_cgns && cg_ns_supported()) { + r = unshare(CLONE_NEWCGROUP); + if (r < 0) + return log_error_errno(errno, "Failed to unshare cgroup namespace"); + r = mount_cgroups( + "", + arg_unified_cgroup_hierarchy, + arg_userns_mode != USER_NAMESPACE_NO, + arg_uid_shift, + arg_uid_range, + arg_selinux_apifs_context, + arg_use_cgns); + if (r < 0) + return r; + } else { + r = mount_systemd_cgroup_writable("", arg_unified_cgroup_hierarchy); + if (r < 0) + return r; + } r = reset_uid_gid(); if (r < 0) @@ -2662,9 +2719,9 @@ static int inner_child( (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) return log_oom(); - assert(!sd_id128_equal(arg_uuid, SD_ID128_NULL)); + assert(!sd_id128_is_null(arg_uuid)); - if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0) + if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_to_uuid_string(arg_uuid, as_uuid)) < 0) return log_oom(); if (fdset_size(fds) > 0) { @@ -2785,6 +2842,7 @@ static int outer_child( const char *root_device, bool root_device_rw, const char *home_device, bool home_device_rw, const char *srv_device, bool srv_device_rw, + const char *esp_device, bool interactive, bool secondary, int pid_socket, @@ -2846,7 +2904,8 @@ static int outer_child( r = mount_devices(directory, root_device, root_device_rw, home_device, home_device_rw, - srv_device, srv_device_rw); + srv_device, srv_device_rw, + esp_device); if (r < 0) return r; @@ -2978,15 +3037,18 @@ static int outer_child( if (r < 0) return r; - r = mount_cgroups( - directory, - arg_unified_cgroup_hierarchy, - arg_userns_mode != USER_NAMESPACE_NO, - arg_uid_shift, - arg_uid_range, - arg_selinux_apifs_context); - if (r < 0) - return r; + if (!arg_use_cgns || !cg_ns_supported()) { + r = mount_cgroups( + directory, + arg_unified_cgroup_hierarchy, + arg_userns_mode != USER_NAMESPACE_NO, + arg_uid_shift, + arg_uid_range, + arg_selinux_apifs_context, + arg_use_cgns); + if (r < 0) + return r; + } r = mount_move_root(directory); if (r < 0) @@ -3449,7 +3511,7 @@ static int load_settings(void) { int main(int argc, char *argv[]) { - _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL, *console = NULL; + _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL, *esp_device = NULL, *console = NULL; bool root_device_rw = true, home_device_rw = true, srv_device_rw = true; _cleanup_close_ int master = -1, image_fd = -1; _cleanup_fdset_free_ FDSet *fds = NULL; @@ -3558,7 +3620,7 @@ int main(int argc, char *argv[]) { } if (r < 0) { log_error_errno(r, "Failed to lock %s: %m", arg_directory); - return r; + goto finish; } if (arg_template) { @@ -3631,6 +3693,7 @@ int main(int argc, char *argv[]) { &root_device, &root_device_rw, &home_device, &home_device_rw, &srv_device, &srv_device_rw, + &esp_device, &secondary); if (r < 0) goto finish; @@ -3805,6 +3868,7 @@ int main(int argc, char *argv[]) { root_device, root_device_rw, home_device, home_device_rw, srv_device, srv_device_rw, + esp_device, interactive, secondary, pid_socket_pair[1], |