X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=e75cc28074ac1b01bca67b005d0588e2670c507b;hp=fcc0f17aee011a9542262bd0d18177d6dcdadf62;hb=a2da110b78abe4e4b1b6d8ae4ef78b087c4dcc8b;hpb=ee3a6a51e5b098aa0c9641ed71d275c459ad2f5a diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index fcc0f17ae..e75cc2807 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -40,10 +40,10 @@ #include #include #include -#include -#include #include #include +#include +#include #ifdef HAVE_SELINUX #include @@ -53,6 +53,10 @@ #include #endif +#ifdef HAVE_BLKID +#include +#endif + #include "sd-daemon.h" #include "sd-bus.h" #include "sd-id128.h" @@ -79,6 +83,21 @@ #include "def.h" #include "rtnl-util.h" #include "udev-util.h" +#include "blkid-util.h" +#include "gpt.h" +#include "siphash24.h" +#include "copy.h" +#include "base-filesystem.h" +#include "barrier.h" + +#ifdef HAVE_SECCOMP +#include "seccomp-util.h" +#endif + +typedef enum ContainerStatus { + CONTAINER_TERMINATED, + CONTAINER_REBOOTED +} ContainerStatus; typedef enum LinkJournal { LINK_NO, @@ -87,12 +106,18 @@ typedef enum LinkJournal { LINK_GUEST } LinkJournal; +typedef enum Volatile { + VOLATILE_NO, + VOLATILE_YES, + VOLATILE_STATE, +} Volatile; + static char *arg_directory = NULL; static char *arg_user = NULL; static sd_id128_t arg_uuid = {}; static char *arg_machine = NULL; -static char *arg_selinux_context = NULL; -static char *arg_selinux_apifs_context = NULL; +static const char *arg_selinux_context = NULL; +static const char *arg_selinux_apifs_context = NULL; static const char *arg_slice = NULL; static bool arg_private_network = false; static bool arg_read_only = false; @@ -127,13 +152,19 @@ static uint64_t arg_retain = (1ULL << CAP_MKNOD); static char **arg_bind = NULL; static char **arg_bind_ro = NULL; +static char **arg_tmpfs = NULL; static char **arg_setenv = NULL; static bool arg_quiet = false; static bool arg_share_system = false; static bool arg_register = true; static bool arg_keep_unit = false; static char **arg_network_interfaces = NULL; +static char **arg_network_macvlan = NULL; static bool arg_network_veth = false; +static const char *arg_network_bridge = NULL; +static unsigned long arg_personality = 0xffffffffLU; +static const char *arg_image = NULL; +static Volatile arg_volatile = VOLATILE_NO; static int help(void) { @@ -142,7 +173,8 @@ static int help(void) { " -h --help Show this help\n" " --version Print version string\n" " -q --quiet Do not show status information\n" - " -D --directory=NAME Root directory for the container\n" + " -D --directory=PATH Root directory for the container\n" + " -i --image=PATH File system device or image for the container\n" " -b --boot Boot up full system (i.e. invoke init)\n" " -u --user=USER Run the command under specified user or uid\n" " -M --machine=NAME Set the machine name for the container\n" @@ -152,8 +184,15 @@ static int help(void) { " --network-interface=INTERFACE\n" " Assign an existing network interface to the\n" " container\n" - " --network-veth Add a a virtual ethernet connection between host\n" + " --network-macvlan=INTERFACE\n" + " Create a macvlan network interface based on an\n" + " existing network interface to the container\n" + " --network-veth Add a virtual ethernet connection between host\n" " and container\n" + " --network-bridge=INTERFACE\n" + " Add a virtual ethernet connection between host\n" + " and container and add it to an existing bridge on\n" + " the host\n" " -Z --selinux-context=SECLABEL\n" " Set the SELinux security context to be used by\n" " processes in the container\n" @@ -169,11 +208,13 @@ static int help(void) { " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n" " the container\n" " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n" + " --tmpfs=PATH:[OPTIONS] Mount an empty tmpfs to the specified directory\n" " --setenv=NAME=VALUE Pass an environment variable to PID 1\n" " --share-system Share system namespaces with host\n" " --register=BOOLEAN Register container as machine\n" " --keep-unit Do not register a scope for the machine, reuse\n" - " the service unit nspawn is running in\n", + " the service unit nspawn is running in\n" + " --volatile[=MODE] Run the system in volatile mode\n", program_invocation_short_name); return 0; @@ -191,12 +232,17 @@ static int parse_argv(int argc, char *argv[]) { ARG_LINK_JOURNAL, ARG_BIND, ARG_BIND_RO, + ARG_TMPFS, ARG_SETENV, ARG_SHARE_SYSTEM, ARG_REGISTER, ARG_KEEP_UNIT, ARG_NETWORK_INTERFACE, + ARG_NETWORK_MACVLAN, ARG_NETWORK_VETH, + ARG_NETWORK_BRIDGE, + ARG_PERSONALITY, + ARG_VOLATILE, }; static const struct option options[] = { @@ -213,6 +259,7 @@ static int parse_argv(int argc, char *argv[]) { { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, { "bind", required_argument, NULL, ARG_BIND }, { "bind-ro", required_argument, NULL, ARG_BIND_RO }, + { "tmpfs", required_argument, NULL, ARG_TMPFS }, { "machine", required_argument, NULL, 'M' }, { "slice", required_argument, NULL, 'S' }, { "setenv", required_argument, NULL, ARG_SETENV }, @@ -223,7 +270,12 @@ static int parse_argv(int argc, char *argv[]) { { "register", required_argument, NULL, ARG_REGISTER }, { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT }, { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE }, - { "network-veth", no_argument, NULL, ARG_NETWORK_VETH }, + { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN }, + { "network-veth", no_argument, NULL, ARG_NETWORK_VETH }, + { "network-bridge", required_argument, NULL, ARG_NETWORK_BRIDGE }, + { "personality", required_argument, NULL, ARG_PERSONALITY }, + { "image", required_argument, NULL, 'i' }, + { "volatile", optional_argument, NULL, ARG_VOLATILE }, {} }; @@ -233,7 +285,7 @@ static int parse_argv(int argc, char *argv[]) { assert(argc >= 0); assert(argv); - while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:q", options, NULL)) >= 0) { + while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:", options, NULL)) >= 0) { switch (c) { @@ -255,6 +307,10 @@ static int parse_argv(int argc, char *argv[]) { break; + case 'i': + arg_image = optarg; + break; + case 'u': free(arg_user); arg_user = strdup(optarg); @@ -263,13 +319,25 @@ static int parse_argv(int argc, char *argv[]) { break; + case ARG_NETWORK_BRIDGE: + arg_network_bridge = optarg; + + /* fall through */ + case ARG_NETWORK_VETH: arg_network_veth = true; arg_private_network = true; break; case ARG_NETWORK_INTERFACE: - if (strv_push(&arg_network_interfaces, optarg) < 0) + if (strv_extend(&arg_network_interfaces, optarg) < 0) + return log_oom(); + + arg_private_network = true; + break; + + case ARG_NETWORK_MACVLAN: + if (strv_extend(&arg_network_macvlan, optarg) < 0) return log_oom(); /* fall through */ @@ -291,10 +359,7 @@ static int parse_argv(int argc, char *argv[]) { break; case 'S': - arg_slice = strdup(optarg); - if (!arg_slice) - return log_oom(); - + arg_slice = optarg; break; case 'M': @@ -418,6 +483,42 @@ static int parse_argv(int argc, char *argv[]) { break; } + case ARG_TMPFS: { + _cleanup_free_ char *a = NULL, *b = NULL; + char *e; + + e = strchr(optarg, ':'); + if (e) { + a = strndup(optarg, e - optarg); + b = strdup(e + 1); + } else { + a = strdup(optarg); + b = strdup("mode=0755"); + } + + if (!a || !b) + return log_oom(); + + if (!path_is_absolute(a)) { + log_error("Invalid tmpfs specification: %s", optarg); + return -EINVAL; + } + + r = strv_push(&arg_tmpfs, a); + if (r < 0) + return log_oom(); + + a = NULL; + + r = strv_push(&arg_tmpfs, b); + if (r < 0) + return log_oom(); + + b = NULL; + + break; + } + case ARG_SETENV: { char **n; @@ -457,6 +558,35 @@ static int parse_argv(int argc, char *argv[]) { arg_keep_unit = true; break; + case ARG_PERSONALITY: + + arg_personality = personality_from_string(optarg); + if (arg_personality == 0xffffffffLU) { + log_error("Unknown or unsupported personality '%s'.", optarg); + return -EINVAL; + } + + break; + + case ARG_VOLATILE: + + if (!optarg) + arg_volatile = VOLATILE_YES; + else { + r = parse_boolean(optarg); + if (r < 0) { + if (streq(optarg, "state")) + arg_volatile = VOLATILE_STATE; + else { + log_error("Failed to parse --volatile= argument: %s", optarg); + return r; + } + } else + arg_volatile = r ? VOLATILE_YES : VOLATILE_NO; + } + + break; + case '?': return -EINVAL; @@ -478,6 +608,16 @@ static int parse_argv(int argc, char *argv[]) { return -EINVAL; } + if (arg_directory && arg_image) { + log_error("--directory= and --image= may not be combined."); + return -EINVAL; + } + + if (arg_volatile != VOLATILE_NO && arg_read_only) { + log_error("Cannot combine --read-only with --volatile. Note that --volatile already implies a read-only base hierarchy."); + return -EINVAL; + } + arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus; return 1; @@ -495,17 +635,17 @@ static int mount_all(const char *dest) { } MountPoint; static const MountPoint mount_table[] = { - { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, - { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */ - { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ - { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, - { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true }, + { "proc", "/proc", "proc", NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, + { "/proc/sys", "/proc/sys", NULL, NULL, MS_BIND, true }, /* Bind mount first */ + { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ + { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, + { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true }, { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true }, - { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, - { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, + { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, + { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, #ifdef HAVE_SELINUX - { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */ - { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */ + { "/sys/fs/selinux", "/sys/fs/selinux", NULL, NULL, MS_BIND, false }, /* Bind mount first */ + { NULL, "/sys/fs/selinux", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, false }, /* Then, make it r/o */ #endif }; @@ -570,25 +710,27 @@ static int mount_all(const char *dest) { return r; } -static int mount_binds(const char *dest, char **l, unsigned long flags) { +static int mount_binds(const char *dest, char **l, bool ro) { char **x, **y; STRV_FOREACH_PAIR(x, y, l) { - char *where; + _cleanup_free_ char *where = NULL; struct stat source_st, dest_st; int r; if (stat(*x, &source_st) < 0) { - log_error("failed to stat %s: %m", *x); + log_error("Failed to stat %s: %m", *x); return -errno; } - where = strappenda(dest, *y); + where = strappend(dest, *y); + if (!where) + return log_oom(); + r = stat(where, &dest_st); if (r == 0) { if ((source_st.st_mode & S_IFMT) != (dest_st.st_mode & S_IFMT)) { - log_error("The file types of %s and %s do not match. Refusing bind mount", - *x, where); + log_error("The file types of %s and %s do not match. Refusing bind mount", *x, where); return -EINVAL; } } else if (errno == ENOENT) { @@ -598,11 +740,12 @@ static int mount_binds(const char *dest, char **l, unsigned long flags) { return r; } } else { - log_error("Failed to bind mount %s: %s", *x, strerror(errno)); + log_error("Failed to bind mount %s: %m", *x); return -errno; } + /* Create the mount point, but be conservative -- refuse to create block - * and char devices. */ + * and char devices. */ if (S_ISDIR(source_st.st_mode)) mkdir_label(where, 0755); else if (S_ISFIFO(source_st.st_mode)) @@ -621,8 +764,32 @@ static int mount_binds(const char *dest, char **l, unsigned long flags) { return -errno; } - if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) { - log_error("mount(%s) failed: %m", where); + if (ro) { + r = bind_remount_recursive(where, true); + if (r < 0) { + log_error("Read-Only bind mount failed: %s", strerror(-r)); + return r; + } + } + } + + return 0; +} + +static int mount_tmpfs(const char *dest) { + char **i, **o; + + STRV_FOREACH_PAIR(i, o, arg_tmpfs) { + _cleanup_free_ char *where = NULL; + + where = strappend(dest, *i); + if (!where) + return log_oom(); + + mkdir_label(where, 0755); + + if (mount("tmpfs", where, "tmpfs", MS_NODEV|MS_STRICTATIME, *o) < 0) { + log_error("tmpfs mount to %s failed: %m", where); return -errno; } } @@ -662,7 +829,6 @@ static int setup_timezone(const char *dest) { 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; @@ -681,7 +847,9 @@ static int setup_timezone(const char *dest) { if (!what) return log_oom(); + mkdir_parents(where, 0755); unlink(where); + if (symlink(what, where) < 0) { log_error("Failed to correct timezone of container: %m"); return 0; @@ -691,7 +859,7 @@ static int setup_timezone(const char *dest) { } static int setup_resolv_conf(const char *dest) { - char _cleanup_free_ *where = NULL; + _cleanup_free_ char *where = NULL; assert(dest); @@ -705,14 +873,117 @@ static int setup_resolv_conf(const char *dest) { /* We don't really care for the results of this really. If it * fails, it fails, but meh... */ - copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW); + mkdir_parents(where, 0755); + copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644); + + return 0; +} + +static int setup_volatile_state(const char *directory) { + const char *p; + int r; + + assert(directory); + + if (arg_volatile != 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) { + log_error("Failed to remount %s read-only: %s", directory, strerror(-r)); + return r; + } + + p = strappenda(directory, "/var"); + mkdir(p, 0755); + + if (mount("tmpfs", p, "tmpfs", MS_STRICTATIME, "mode=755") < 0) { + log_error("Failed to mount tmpfs to /var: %m"); + return -errno; + } + + return 0; +} + +static int setup_volatile(const char *directory) { + bool tmpfs_mounted = false, bind_mounted = false; + char template[] = "/tmp/nspawn-volatile-XXXXXX"; + const char *f, *t; + int r; + + assert(directory); + + if (arg_volatile != 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)) { + log_error("Failed to create temporary directory: %m"); + return -errno; + } + + if (mount("tmpfs", template, "tmpfs", MS_STRICTATIME, "mode=755") < 0) { + log_error("Failed to mount tmpfs for root directory: %m"); + r = -errno; + goto fail; + } + + tmpfs_mounted = true; + + f = strappenda(directory, "/usr"); + t = strappenda(template, "/usr"); + + mkdir(t, 0755); + if (mount(f, t, "bind", MS_BIND|MS_REC, NULL) < 0) { + log_error("Failed to create /usr bind mount: %m"); + r = -errno; + goto fail; + } + + bind_mounted = true; + + r = bind_remount_recursive(t, true); + if (r < 0) { + log_error("Failed to remount %s read-only: %s", t, strerror(-r)); + goto fail; + } + + if (mount(template, directory, NULL, MS_MOVE, NULL) < 0) { + log_error("Failed to move root mount: %m"); + r = -errno; + goto fail; + } + + rmdir(template); return 0; + +fail: + if (bind_mounted) + umount(t); + if (tmpfs_mounted) + umount(template); + rmdir(template); + return r; +} + +static char* id128_format_as_uuid(sd_id128_t id, char s[37]) { + + 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) { _cleanup_free_ char *from = NULL, *to = NULL; - sd_id128_t rnd; + sd_id128_t rnd = {}; char as_uuid[37]; int r; @@ -735,10 +1006,7 @@ static int setup_boot_id(const char *dest) { return r; } - snprintf(as_uuid, sizeof(as_uuid), - "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x", - SD_ID128_FORMAT_VAL(rnd)); - char_array_0(as_uuid); + id128_format_as_uuid(rnd, as_uuid); r = write_string_file(from, as_uuid); if (r < 0) { @@ -821,23 +1089,19 @@ static int setup_ptmx(const char *dest) { } static int setup_dev_console(const char *dest, const char *console) { + _cleanup_umask_ mode_t u; + const char *to; struct stat st; - _cleanup_free_ char *to = NULL; int r; - _cleanup_umask_ mode_t u; assert(dest); assert(console); u = umask(0000); - if (stat(console, &st) < 0) { - log_error("Failed to stat %s: %m", console); + if (stat("/dev/null", &st) < 0) { + log_error("Failed to stat /dev/null: %m"); return -errno; - - } else if (!S_ISCHR(st.st_mode)) { - log_error("/dev/console is not a char device"); - return -EIO; } r = chmod_and_chown(console, 0600, 0, 0); @@ -846,16 +1110,15 @@ static int setup_dev_console(const char *dest, const char *console) { return r; } - if (asprintf(&to, "%s/dev/console", dest) < 0) - return log_oom(); - /* We need to bind mount the right tty to /dev/console since * ptys can only exist on pts file systems. To have something - * to bind mount things on we create a device node first, that - * has the right major/minor (note that the major minor - * doesn't actually matter here, since we mount it over - * anyway). */ + * to bind mount things on we create a device node first, and + * use /dev/null for that since we the cgroups device policy + * allows us to create that freely, while we cannot create + * /dev/console. (Note that the major minor doesn't actually + * matter here, since we mount it over anyway). */ + to = strappenda(dest, "/dev/console"); if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) { log_error("mknod() for /dev/console failed: %m"); return -errno; @@ -932,7 +1195,7 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { /* 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_DONTWAIT|MSG_NOSIGNAL); - close_nointr_nofail(fd); + safe_close(fd); if (k < 0) { log_error("Failed to send FIFO fd: %m"); @@ -1083,10 +1346,8 @@ static int setup_journal(const char *directory) { } else if (access(p, F_OK) < 0) return 0; - if (dir_is_empty(q) == 0) { - log_error("%s not empty.", q); - return -ENOTEMPTY; - } + if (dir_is_empty(q) == 0) + log_warning("%s is not empty, proceeding anyway.", q); r = mkdir_p(q, 0755); if (r < 0) { @@ -1126,7 +1387,7 @@ static int drop_capabilities(void) { return capability_bounding_set_drop(~arg_retain, false); } -static int register_machine(pid_t pid) { +static int register_machine(pid_t pid, int local_ifindex) { _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; _cleanup_bus_unref_ sd_bus *bus = NULL; int r; @@ -1146,33 +1407,110 @@ static int register_machine(pid_t pid) { "org.freedesktop.machine1", "/org/freedesktop/machine1", "org.freedesktop.machine1.Manager", - "RegisterMachine", + "RegisterMachineWithNetwork", &error, NULL, - "sayssus", + "sayssusai", arg_machine, SD_BUS_MESSAGE_APPEND_ID128(arg_uuid), "nspawn", "container", (uint32_t) pid, - strempty(arg_directory)); + strempty(arg_directory), + local_ifindex > 0 ? 1 : 0, local_ifindex); } else { - r = sd_bus_call_method( + _cleanup_bus_message_unref_ sd_bus_message *m = NULL; + + r = sd_bus_message_new_method_call( bus, + &m, "org.freedesktop.machine1", "/org/freedesktop/machine1", "org.freedesktop.machine1.Manager", - "CreateMachine", - &error, - NULL, - "sayssusa(sv)", + "CreateMachineWithNetwork"); + if (r < 0) { + log_error("Failed to create message: %s", strerror(-r)); + return 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), - !isempty(arg_slice), "Slice", "s", arg_slice); + local_ifindex > 0 ? 1 : 0, local_ifindex); + if (r < 0) { + log_error("Failed to append message arguments: %s", strerror(-r)); + return r; + } + + r = sd_bus_message_open_container(m, 'a', "(sv)"); + if (r < 0) { + log_error("Failed to open container: %s", strerror(-r)); + return r; + } + + if (!isempty(arg_slice)) { + r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice); + if (r < 0) { + log_error("Failed to append slice: %s", strerror(-r)); + return r; + } + } + + r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict"); + if (r < 0) { + log_error("Failed to add device policy: %s", strerror(-r)); + return r; + } + + r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 10, + /* 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", + /* 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 the container + * access to all kdbus + * devices. Again, the + * container cannot create + * these nodes, only use + * them. We use a pretty + * open match here, so that + * the kernel API can still + * change. */ + "char-kdbus", "rw", + "char-kdbus/*", "rw"); + if (r < 0) { + log_error("Failed to add device whitelist: %s", strerror(-r)); + return r; + } + + r = sd_bus_message_close_container(m); + if (r < 0) { + log_error("Failed to close container: %s", strerror(-r)); + return r; + } + + r = sd_bus_call(bus, m, 0, &error, NULL); } if (r < 0) { @@ -1246,7 +1584,7 @@ static int reset_audit_loginuid(void) { return 0; r = read_one_line_file("/proc/self/loginuid", &p); - if (r == -EEXIST) + if (r == -ENOENT) return 0; if (r < 0) { log_error("Failed to read /proc/self/loginuid: %s", strerror(-r)); @@ -1271,11 +1609,47 @@ static int reset_audit_loginuid(void) { return 0; } -static int setup_veth(int netns_fd) { +#define HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2) + +static int get_mac(struct ether_addr *mac) { + int r; + + uint8_t result[8]; + size_t l, sz; + uint8_t *v; + + l = strlen(arg_machine); + sz = sizeof(sd_id128_t) + l; + 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 */ + memcpy(v + sizeof(sd_id128_t), arg_machine, l); + + /* 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_rtnl_message_unref_ sd_rtnl_message *m = NULL; _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; - char iface_name[IFNAMSIZ] = "ve-"; - int r; + struct ether_addr mac; + int r, i; if (!arg_private_network) return 0; @@ -1283,23 +1657,35 @@ static int setup_veth(int netns_fd) { if (!arg_network_veth) return 0; - strncpy(iface_name+3, arg_machine, sizeof(iface_name) - 3); + /* Use two different interface name prefixes depending whether + * we are in bridge mode or not. */ + if (arg_network_bridge) + memcpy(iface_name, "vb-", 3); + else + memcpy(iface_name, "ve-", 3); + strncpy(iface_name+3, arg_machine, IFNAMSIZ - 3); + + r = get_mac(&mac); + if (r < 0) { + log_error("Failed to generate predictable MAC address for host0"); + return r; + } - r = sd_rtnl_open(0, &rtnl); + r = sd_rtnl_open(&rtnl, 0); if (r < 0) { log_error("Failed to connect to netlink: %s", strerror(-r)); return r; } - r = sd_rtnl_message_new_link(RTM_NEWLINK, 0, &m); + r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, 0); if (r < 0) { log_error("Failed to allocate netlink message: %s", strerror(-r)); return r; } - r = sd_rtnl_message_append_string(m, IFLA_IFNAME, "host0"); + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, iface_name); if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); + log_error("Failed to add netlink interface name: %s", strerror(-r)); return r; } @@ -1309,31 +1695,31 @@ static int setup_veth(int netns_fd) { return r; } - r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "veth"); + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "veth"); if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); + log_error("Failed to open netlink container: %s", strerror(-r)); return r; } - r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA); + r = sd_rtnl_message_open_container(m, VETH_INFO_PEER); if (r < 0) { log_error("Failed to open netlink container: %s", strerror(-r)); return r; } - r = sd_rtnl_message_open_container(m, VETH_INFO_PEER); + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, "host0"); if (r < 0) { - log_error("z Failed to open netlink container: %s", strerror(-r)); + log_error("Failed to add netlink interface name: %s", strerror(-r)); return r; } - r = sd_rtnl_message_append_string(m, IFLA_IFNAME, iface_name); + r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac); if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); + log_error("Failed to add netlink MAC address: %s", strerror(-r)); return r; } - r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_FD, netns_fd); + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); if (r < 0) { log_error("Failed to add netlink namespace field: %s", strerror(-r)); return r; @@ -1363,141 +1749,1157 @@ static int setup_veth(int netns_fd) { return r; } + i = (int) if_nametoindex(iface_name); + if (i <= 0) { + log_error("Failed to resolve interface %s: %m", iface_name); + return -errno; + } + + *ifi = i; + return 0; } -static int move_network_interfaces(pid_t pid) { - _cleanup_udev_unref_ struct udev *udev = NULL; +static int setup_bridge(const char veth_name[], int *ifi) { + _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; - char **i; - int r; + int r, bridge; if (!arg_private_network) return 0; - if (strv_isempty(arg_network_interfaces)) + if (!arg_network_veth) + return 0; + + if (!arg_network_bridge) return 0; - r = sd_rtnl_open(0, &rtnl); + bridge = (int) if_nametoindex(arg_network_bridge); + if (bridge <= 0) { + log_error("Failed to resolve interface %s: %m", arg_network_bridge); + return -errno; + } + + *ifi = bridge; + + r = sd_rtnl_open(&rtnl, 0); if (r < 0) { log_error("Failed to connect to netlink: %s", strerror(-r)); return r; } - udev = udev_new(); - if (!udev) { - log_error("Failed to connect to udev."); - return -ENOMEM; + r = sd_rtnl_message_new_link(rtnl, &m, RTM_SETLINK, 0); + if (r < 0) { + log_error("Failed to allocate netlink message: %s", strerror(-r)); + return r; } - STRV_FOREACH(i, arg_network_interfaces) { - _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; - _cleanup_udev_device_unref_ struct udev_device *d = NULL; - char ifi_str[2 + DECIMAL_STR_MAX(int)]; - int ifi; - - ifi = (int) if_nametoindex(*i); - if (ifi <= 0) { - log_error("Failed to resolve interface %s: %m", *i); - return -errno; - } + r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP); + if (r < 0) { + log_error("Failed to set IFF_UP flag: %s", strerror(-r)); + return r; + } - sprintf(ifi_str, "n%i", ifi); - d = udev_device_new_from_device_id(udev, ifi_str); - if (!d) { - log_error("Failed to get udev device for interface %s: %m", *i); - return -errno; - } + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, veth_name); + if (r < 0) { + log_error("Failed to add netlink interface name field: %s", strerror(-r)); + return r; + } - if (udev_device_get_is_initialized(d) <= 0) { - log_error("Network interface %s is not initialized yet.", *i); - return -EBUSY; - } + r = sd_rtnl_message_append_u32(m, IFLA_MASTER, bridge); + if (r < 0) { + log_error("Failed to add netlink master field: %s", strerror(-r)); + return r; + } - r = sd_rtnl_message_new_link(RTM_NEWLINK, ifi, &m); - if (r < 0) { - log_error("Failed to allocate netlink message: %s", strerror(-r)); + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) { + log_error("Failed to add veth interface to bridge: %s", strerror(-r)); + return r; + } + + 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) { + log_error("Failed to resolve interface %s: %m", name); + return -errno; + } + + sprintf(ifi_str, "n%i", ifi); + d = udev_device_new_from_device_id(udev, ifi_str); + if (!d) { + log_error("Failed to get udev device for interface %s: %m", name); + return -errno; + } + + if (udev_device_get_is_initialized(d) <= 0) { + log_error("Network interface %s is not initialized yet.", name); + return -EBUSY; + } + + return ifi; +} + +static int move_network_interfaces(pid_t pid) { + _cleanup_udev_unref_ struct udev *udev = NULL; + _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + char **i; + int r; + + if (!arg_private_network) + return 0; + + if (strv_isempty(arg_network_interfaces)) + return 0; + + r = sd_rtnl_open(&rtnl, 0); + if (r < 0) { + log_error("Failed to connect to netlink: %s", strerror(-r)); + return r; + } + + udev = udev_new(); + if (!udev) { + log_error("Failed to connect to udev."); + return -ENOMEM; + } + + STRV_FOREACH(i, arg_network_interfaces) { + _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL; + int ifi; + + ifi = parse_interface(udev, *i); + if (ifi < 0) + return ifi; + + r = sd_rtnl_message_new_link(rtnl, &m, RTM_NEWLINK, ifi); + if (r < 0) { + log_error("Failed to allocate netlink message: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); + if (r < 0) { + log_error("Failed to append namespace PID to netlink message: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) { + log_error("Failed to move interface %s to namespace: %s", *i, strerror(-r)); + return r; + } + } + + return 0; +} + +static int setup_macvlan(pid_t pid) { + _cleanup_udev_unref_ struct udev *udev = NULL; + _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + char **i; + int r; + + if (!arg_private_network) + return 0; + + if (strv_isempty(arg_network_macvlan)) + return 0; + + r = sd_rtnl_open(&rtnl, 0); + if (r < 0) { + log_error("Failed to connect to netlink: %s", strerror(-r)); + return r; + } + + udev = udev_new(); + if (!udev) { + log_error("Failed to connect to udev."); + return -ENOMEM; + } + + STRV_FOREACH(i, arg_network_macvlan) { + _cleanup_rtnl_message_unref_ sd_rtnl_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) { + log_error("Failed to allocate netlink message: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi); + if (r < 0) { + log_error("Failed to add netlink interface index: %s", strerror(-r)); + return r; + } + + n = strappend("mv-", *i); + if (!n) + return log_oom(); + + strshorten(n, IFNAMSIZ-1); + + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n); + if (r < 0) { + log_error("Failed to add netlink interface name: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); + if (r < 0) { + log_error("Failed to add netlink namespace field: %s", strerror(-r)); return r; } - r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); - if (r < 0) { - log_error("Failed to append namespace PID to netlink message: %s", strerror(-r)); - return r; + r = sd_rtnl_message_open_container(m, IFLA_LINKINFO); + if (r < 0) { + log_error("Failed to open netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "macvlan"); + if (r < 0) { + log_error("Failed to open netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_append_u32(m, IFLA_MACVLAN_MODE, MACVLAN_MODE_BRIDGE); + if (r < 0) { + log_error("Failed to append macvlan mode: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_close_container(m); + if (r < 0) { + log_error("Failed to close netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_message_close_container(m); + if (r < 0) { + log_error("Failed to close netlink container: %s", strerror(-r)); + return r; + } + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) { + log_error("Failed to add new macvlan interfaces: %s", strerror(-r)); + return r; + } + } + + return 0; +} + +static int setup_seccomp(void) { + +#ifdef HAVE_SECCOMP + static const int blacklist[] = { + SCMP_SYS(kexec_load), + SCMP_SYS(open_by_handle_at), + SCMP_SYS(init_module), + SCMP_SYS(finit_module), + SCMP_SYS(delete_module), + SCMP_SYS(iopl), + SCMP_SYS(ioperm), + SCMP_SYS(swapon), + SCMP_SYS(swapoff), + }; + + scmp_filter_ctx seccomp; + unsigned i; + int r; + + seccomp = seccomp_init(SCMP_ACT_ALLOW); + if (!seccomp) + return log_oom(); + + r = seccomp_add_secondary_archs(seccomp); + if (r < 0) { + log_error("Failed to add secondary archs to seccomp filter: %s", strerror(-r)); + goto finish; + } + + for (i = 0; i < ELEMENTSOF(blacklist); i++) { + r = seccomp_rule_add(seccomp, SCMP_ACT_ERRNO(EPERM), blacklist[i], 0); + if (r == -EFAULT) + continue; /* unknown syscall */ + if (r < 0) { + log_error("Failed to block syscall: %s", strerror(-r)); + goto finish; + } + } + + /* + Audit is broken in containers, much of the userspace audit + hookup will fail if running inside a container. We don't + care and just turn off creation of audit sockets. + + This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail + with EAFNOSUPPORT which audit userspace uses as indication + that audit is disabled in the kernel. + */ + + r = seccomp_rule_add( + seccomp, + SCMP_ACT_ERRNO(EAFNOSUPPORT), + SCMP_SYS(socket), + 2, + SCMP_A0(SCMP_CMP_EQ, AF_NETLINK), + SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT)); + if (r < 0) { + log_error("Failed to add audit seccomp rule: %s", strerror(-r)); + goto finish; + } + + r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0); + if (r < 0) { + log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r)); + goto finish; + } + + r = seccomp_load(seccomp); + if (r < 0) + log_error("Failed to install seccomp audit filter: %s", strerror(-r)); + +finish: + seccomp_release(seccomp); + return r; +#else + return 0; +#endif + +} + +static int setup_image(char **device_path, int *loop_nr) { + struct loop_info64 info = { + .lo_flags = LO_FLAGS_AUTOCLEAR|LO_FLAGS_PARTSCAN + }; + _cleanup_close_ int fd = -1, control = -1, loop = -1; + _cleanup_free_ char* loopdev = NULL; + struct stat st; + int r, nr; + + assert(device_path); + assert(loop_nr); + + fd = open(arg_image, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY); + if (fd < 0) { + log_error("Failed to open %s: %m", arg_image); + return -errno; + } + + if (fstat(fd, &st) < 0) { + log_error("Failed to stat %s: %m", arg_image); + return -errno; + } + + if (S_ISBLK(st.st_mode)) { + char *p; + + p = strdup(arg_image); + if (!p) + return log_oom(); + + *device_path = p; + + *loop_nr = -1; + + r = fd; + fd = -1; + + return r; + } + + if (!S_ISREG(st.st_mode)) { + log_error("%s is not a regular file or block device: %m", arg_image); + return -EINVAL; + } + + control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); + if (control < 0) { + log_error("Failed to open /dev/loop-control: %m"); + return -errno; + } + + nr = ioctl(control, LOOP_CTL_GET_FREE); + if (nr < 0) { + log_error("Failed to allocate loop device: %m"); + return -errno; + } + + if (asprintf(&loopdev, "/dev/loop%i", nr) < 0) + return log_oom(); + + loop = open(loopdev, O_CLOEXEC|(arg_read_only ? O_RDONLY : O_RDWR)|O_NONBLOCK|O_NOCTTY); + if (loop < 0) { + log_error("Failed to open loop device %s: %m", loopdev); + return -errno; + } + + if (ioctl(loop, LOOP_SET_FD, fd) < 0) { + log_error("Failed to set loopback file descriptor on %s: %m", loopdev); + return -errno; + } + + if (arg_read_only) + info.lo_flags |= LO_FLAGS_READ_ONLY; + + if (ioctl(loop, LOOP_SET_STATUS64, &info) < 0) { + log_error("Failed to set loopback settings on %s: %m", loopdev); + return -errno; + } + + *device_path = loopdev; + loopdev = NULL; + + *loop_nr = nr; + + r = loop; + loop = -1; + + return r; +} + +static int dissect_image( + int fd, + char **root_device, bool *root_device_rw, + char **home_device, bool *home_device_rw, + char **srv_device, bool *srv_device_rw, + bool *secondary) { + +#ifdef HAVE_BLKID + int home_nr = -1, root_nr = -1, secondary_root_nr = -1, srv_nr = -1; + _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = 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; + _cleanup_udev_unref_ struct udev *udev = NULL; + struct udev_list_entry *first, *item; + bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true; + const char *pttype = NULL; + blkid_partlist pl; + struct stat st; + int r; + + assert(fd >= 0); + assert(root_device); + assert(home_device); + assert(srv_device); + assert(secondary); + + b = blkid_new_probe(); + if (!b) + return log_oom(); + + errno = 0; + r = blkid_probe_set_device(b, fd, 0, 0); + if (r != 0) { + if (errno == 0) + return log_oom(); + + log_error("Failed to set device on blkid probe: %m"); + return -errno; + } + + blkid_probe_enable_partitions(b, 1); + blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS); + + errno = 0; + r = blkid_do_safeprobe(b); + if (r == -2 || r == 1) { + log_error("Failed to identify any partition table on %s.\n" + "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + return -EINVAL; + } else if (r != 0) { + if (errno == 0) + errno = EIO; + log_error("Failed to probe: %m"); + return -errno; + } + + blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL); + if (!streq_ptr(pttype, "gpt")) { + log_error("Image %s does not carry a GUID Partition Table.\n" + "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + return -EINVAL; + } + + errno = 0; + pl = blkid_probe_get_partitions(b); + if (!pl) { + if (errno == 0) + return log_oom(); + + log_error("Failed to list partitions of %s", arg_image); + return -errno; + } + + udev = udev_new(); + if (!udev) + return log_oom(); + + if (fstat(fd, &st) < 0) { + log_error("Failed to stat block device: %m"); + return -errno; + } + + d = udev_device_new_from_devnum(udev, 'b', st.st_rdev); + if (!d) + return log_oom(); + + e = udev_enumerate_new(udev); + if (!e) + return log_oom(); + + r = udev_enumerate_add_match_parent(e, d); + if (r < 0) + return log_oom(); + + r = udev_enumerate_scan_devices(e); + if (r < 0) { + log_error("Failed to scan for partition devices of %s: %s", arg_image, strerror(-r)); + return r; + } + + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) { + _cleanup_udev_device_unref_ struct udev_device *q; + const char *stype, *node; + unsigned long long flags; + sd_id128_t type_id; + blkid_partition pp; + dev_t qn; + int nr; + + errno = 0; + q = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item)); + if (!q) { + if (!errno) + errno = ENOMEM; + + log_error("Failed to get partition device of %s: %m", arg_image); + return -errno; + } + + qn = udev_device_get_devnum(q); + if (major(qn) == 0) + continue; + + if (st.st_rdev == qn) + continue; + + node = udev_device_get_devnode(q); + if (!node) + continue; + + pp = blkid_partlist_devno_to_partition(pl, qn); + if (!pp) + continue; + + flags = blkid_partition_get_flags(pp); + if (flags & GPT_FLAG_NO_AUTO) + continue; + + nr = blkid_partition_get_partno(pp); + if (nr < 0) + continue; + + stype = blkid_partition_get_type_string(pp); + if (!stype) + continue; + + if (sd_id128_from_string(stype, &type_id) < 0) + continue; + + if (sd_id128_equal(type_id, GPT_HOME)) { + + if (home && nr >= home_nr) + continue; + + home_nr = nr; + home_rw = !(flags & GPT_FLAG_READ_ONLY); + + free(home); + home = strdup(node); + if (!home) + return log_oom(); + } else if (sd_id128_equal(type_id, GPT_SRV)) { + + if (srv && nr >= srv_nr) + continue; + + srv_nr = nr; + srv_rw = !(flags & GPT_FLAG_READ_ONLY); + + free(srv); + srv = strdup(node); + if (!srv) + return log_oom(); + } +#ifdef GPT_ROOT_NATIVE + else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { + + if (root && nr >= root_nr) + continue; + + root_nr = nr; + root_rw = !(flags & GPT_FLAG_READ_ONLY); + + free(root); + root = strdup(node); + if (!root) + return log_oom(); + } +#endif +#ifdef GPT_ROOT_SECONDARY + else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY)) { + + if (secondary_root && nr >= secondary_root_nr) + continue; + + secondary_root_nr = nr; + secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + + + free(secondary_root); + secondary_root = strdup(node); + if (!secondary_root) + return log_oom(); + } +#endif + } + + if (!root && !secondary_root) { + log_error("Failed to identify root partition in disk image %s.\n" + "Note that the disk image needs to follow http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/ to be supported by systemd-nspawn.", arg_image); + return -EINVAL; + } + + if (root) { + *root_device = root; + root = NULL; + + *root_device_rw = root_rw; + *secondary = false; + } else if (secondary_root) { + *root_device = secondary_root; + secondary_root = NULL; + + *root_device_rw = secondary_root_rw; + *secondary = true; + } + + if (home) { + *home_device = home; + home = NULL; + + *home_device_rw = home_rw; + } + + if (srv) { + *srv_device = srv; + srv = NULL; + + *srv_device_rw = srv_rw; + } + + return 0; +#else + log_error("--image= is not supported, compiled without blkid support."); + return -ENOTSUP; +#endif +} + +static int mount_device(const char *what, const char *where, const char *directory, bool rw) { +#ifdef HAVE_BLKID + _cleanup_blkid_free_probe_ blkid_probe b = NULL; + const char *fstype, *p; + int r; + + assert(what); + assert(where); + + if (arg_read_only) + rw = false; + + if (directory) + p = strappenda(where, directory); + else + p = where; + + errno = 0; + b = blkid_new_probe_from_filename(what); + if (!b) { + if (errno == 0) + return log_oom(); + log_error("Failed to allocate prober for %s: %m", what); + return -errno; + } + + blkid_probe_enable_superblocks(b, 1); + blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE); + + errno = 0; + r = blkid_do_safeprobe(b); + if (r == -1 || r == 1) { + log_error("Cannot determine file system type of %s", what); + return -EINVAL; + } else if (r != 0) { + if (errno == 0) + errno = EIO; + log_error("Failed to probe %s: %m", what); + return -errno; + } + + errno = 0; + if (blkid_probe_lookup_value(b, "TYPE", &fstype, NULL) < 0) { + if (errno == 0) + errno = EINVAL; + log_error("Failed to determine file system type of %s", what); + return -errno; + } + + if (streq(fstype, "crypto_LUKS")) { + log_error("nspawn currently does not support LUKS disk images."); + return -ENOTSUP; + } + + if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) { + log_error("Failed to mount %s: %m", what); + return -errno; + } + + return 0; +#else + log_error("--image= is not supported, compiled without blkid support."); + return -ENOTSUP; +#endif +} + +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) { + int r; + + assert(where); + + if (root_device) { + r = mount_device(root_device, arg_directory, NULL, root_device_rw); + if (r < 0) { + log_error("Failed to mount root directory: %s", strerror(-r)); + return r; + } + } + + if (home_device) { + r = mount_device(home_device, arg_directory, "/home", home_device_rw); + if (r < 0) { + log_error("Failed to mount home directory: %s", strerror(-r)); + return r; + } + } + + if (srv_device) { + r = mount_device(srv_device, arg_directory, "/srv", srv_device_rw); + if (r < 0) { + log_error("Failed to mount server data directory: %s", strerror(-r)); + return r; + } + } + + return 0; +} + +static void loop_remove(int nr, int *image_fd) { + _cleanup_close_ int control = -1; + + if (nr < 0) + return; + + if (image_fd && *image_fd >= 0) { + ioctl(*image_fd, LOOP_CLR_FD); + *image_fd = safe_close(*image_fd); + } + + control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); + if (control < 0) + return; + + ioctl(control, LOOP_CTL_REMOVE, 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) { + log_error("Failed to allocate pipe: %m"); + return -errno; + } + + pid = fork(); + if (pid < 0) { + log_error("Failed to fork getent child: %m"); + return -errno; + } 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); + + reset_all_signal_handlers(); + 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], *w, *x, *state, *u, *g, *h; + _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 */ + + if (setgroups(0, NULL) < 0) { + log_error("setgroups() failed: %m"); + return -errno; + } + + if (setresgid(0, 0, 0) < 0) { + log_error("setregid() failed: %m"); + return -errno; + } + + if (setresuid(0, 0, 0) < 0) { + log_error("setreuid() failed: %m"); + return -errno; + } + + *_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("Failed to read from getent: %m"); + return -errno; + } + + truncate_nl(line); + + wait_for_terminate_and_warn("getent passwd", pid); + + 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; } - r = sd_rtnl_call(rtnl, m, 0, NULL); + log_error("Failed to read from getent: %m"); + return -errno; + } + + truncate_nl(line); + + wait_for_terminate_and_warn("getent initgroups", pid); + + /* Skip over the username and subsequent separator whitespace */ + x = line; + x += strcspn(x, WHITESPACE); + x += strspn(x, WHITESPACE); + + FOREACH_WORD(w, l, x, state) { + char c[l+1]; + + memcpy(c, w, 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 move interface %s to namespace: %s", *i, strerror(-r)); - return r; + log_error("Failed to parse group data from getent."); + return -EIO; } } - return 0; -} + r = mkdir_parents(home, 0775); + if (r < 0) { + log_error("Failed to make home root directory: %s", strerror(-r)); + return r; + } -static int audit_still_doesnt_work_in_containers(void) { + r = mkdir_safe(home, 0755, uid, gid); + if (r < 0 && r != -EEXIST) { + log_error("Failed to make home directory: %s", strerror(-r)); + return r; + } -#ifdef HAVE_SECCOMP - scmp_filter_ctx seccomp; - int r; + fchown(STDIN_FILENO, uid, gid); + fchown(STDOUT_FILENO, uid, gid); + fchown(STDERR_FILENO, uid, gid); - /* - Audit is broken in containers, much of the userspace audit - hookup will fail if running inside a container. We don't - care and just turn off creation of audit sockets. + if (setgroups(n_uids, uids) < 0) { + log_error("Failed to set auxiliary groups: %m"); + return -errno; + } - This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail - with EAFNOSUPPORT which audit userspace uses as indication - that audit is disabled in the kernel. - */ + if (setresgid(gid, gid, gid) < 0) { + log_error("setregid() failed: %m"); + return -errno; + } - seccomp = seccomp_init(SCMP_ACT_ALLOW); - if (!seccomp) - return log_oom(); + if (setresuid(uid, uid, uid) < 0) { + log_error("setreuid() failed: %m"); + return -errno; + } - r = seccomp_rule_add_exact( - seccomp, - SCMP_ACT_ERRNO(EAFNOSUPPORT), - SCMP_SYS(socket), - 2, - SCMP_A0(SCMP_CMP_EQ, AF_NETLINK), - SCMP_A2(SCMP_CMP_EQ, NETLINK_AUDIT)); - if (r < 0) { - log_error("Failed to add audit seccomp rule: %s", strerror(-r)); - goto finish; + if (_home) { + *_home = home; + home = NULL; } - r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0); + return 0; +} + +/* + * Return values: + * < 0 : wait_for_terminate() failed to get the state of the + * container, the container was terminated by a signal, or + * failed for an unknown reason. No change is made to the + * container argument. + * > 0 : The program executed in the container terminated with an + * error. The exit code of the program executed in the + * container is returned. No change is made to the container + * argument. + * 0 : The container is being rebooted, has been shut down or exited + * successfully. The container argument has been set to either + * CONTAINER_TERMINATED or CONTAINER_REBOOTED. + * + * That is, success is indicated by a return value of zero, and an + * error is indicated by a non-zero value. + */ +static int wait_for_container(pid_t pid, ContainerStatus *container) { + int r; + siginfo_t status; + + r = wait_for_terminate(pid, &status); if (r < 0) { - log_error("Failed to unset NO_NEW_PRIVS: %s", strerror(-r)); - goto finish; + log_warning("Failed to wait for container: %s", strerror(-r)); + return r; } - r = seccomp_load(seccomp); - if (r < 0) - log_error("Failed to install seccomp audit filter: %s", strerror(-r)); + switch (status.si_code) { + case CLD_EXITED: + r = status.si_status; + if (r == 0) { + if (!arg_quiet) + log_debug("Container %s exited successfully.", + arg_machine); -finish: - seccomp_release(seccomp); - return r; -#else - return 0; -#endif + *container = CONTAINER_TERMINATED; + } else { + log_error("Container %s failed with error code %i.", + arg_machine, status.si_status); + } + break; + + case CLD_KILLED: + if (status.si_status == SIGINT) { + if (!arg_quiet) + log_info("Container %s has been shut down.", + arg_machine); + + *container = CONTAINER_TERMINATED; + r = 0; + break; + } else if (status.si_status == SIGHUP) { + if (!arg_quiet) + log_info("Container %s is being rebooted.", + arg_machine); + *container = CONTAINER_REBOOTED; + r = 0; + break; + } + /* CLD_KILLED fallthrough */ + + case CLD_DUMPED: + log_error("Container %s terminated by signal %s.", + arg_machine, signal_to_string(status.si_status)); + r = -1; + break; + + default: + log_error("Container %s failed due to unknown reason.", + arg_machine); + r = -1; + break; + } + + return r; } +static void nop_handler(int sig) {} + int main(int argc, char *argv[]) { - _cleanup_close_ int master = -1, kdbus_fd = -1, sync_fd = -1, netns_fd = -1; - _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 }; - _cleanup_free_ char *kdbus_domain = NULL; + _cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL; + bool root_device_rw = true, home_device_rw = true, srv_device_rw = true; + _cleanup_close_ int master = -1, kdbus_fd = -1, image_fd = -1; + _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }; _cleanup_fdset_free_ FDSet *fds = NULL; + int r = EXIT_FAILURE, k, n_fd_passed, loop_nr = -1; const char *console = NULL; - int r = EXIT_FAILURE, k; - int n_fd_passed; + char veth_name[IFNAMSIZ]; + bool secondary = false; + sigset_t mask, mask_chld; pid_t pid = 0; - sigset_t mask; log_parse_environment(); log_open(); @@ -1510,24 +2912,25 @@ int main(int argc, char *argv[]) { goto finish; } - if (arg_directory) { - char *p; + if (!arg_image) { + if (arg_directory) { + char *p; - p = path_make_absolute_cwd(arg_directory); - free(arg_directory); - arg_directory = p; - } else - arg_directory = get_current_dir_name(); + p = path_make_absolute_cwd(arg_directory); + free(arg_directory); + arg_directory = p; + } else + arg_directory = get_current_dir_name(); - if (!arg_directory) { - log_error("Failed to determine path, please use -D."); - goto finish; + if (!arg_directory) { + log_error("Failed to determine path, please use -D."); + goto finish; + } + path_kill_slashes(arg_directory); } - path_kill_slashes(arg_directory); - if (!arg_machine) { - arg_machine = strdup(basename(arg_directory)); + arg_machine = strdup(basename(arg_image ? arg_image : arg_directory)); if (!arg_machine) { log_oom(); goto finish; @@ -1550,39 +2953,69 @@ int main(int argc, char *argv[]) { goto finish; } - if (path_equal(arg_directory, "/")) { - log_error("Spawning container on root directory not supported."); - goto finish; + log_close(); + n_fd_passed = sd_listen_fds(false); + if (n_fd_passed > 0) { + k = fdset_new_listen_fds(&fds, false); + if (k < 0) { + log_error("Failed to collect file descriptors: %s", strerror(-k)); + goto finish; + } } + fdset_close_others(fds); + log_open(); - if (arg_boot) { - if (path_is_os_tree(arg_directory) <= 0) { - log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory); + if (arg_directory) { + if (path_equal(arg_directory, "/")) { + log_error("Spawning container on root directory not supported."); goto finish; } + + if (arg_boot) { + if (path_is_os_tree(arg_directory) <= 0) { + log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory); + goto finish; + } + } else { + const char *p; + + p = strappenda(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); + goto finish; + + } + } } else { - const char *p; + char template[] = "/tmp/nspawn-root-XXXXXX"; - p = strappenda(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); + if (!mkdtemp(template)) { + log_error("Failed to create temporary directory: %m"); + r = -errno; goto finish; + } + arg_directory = strdup(template); + if (!arg_directory) { + r = log_oom(); + goto finish; } - } - log_close(); - n_fd_passed = sd_listen_fds(false); - if (n_fd_passed > 0) { - k = fdset_new_listen_fds(&fds, false); - if (k < 0) { - log_error("Failed to collect file descriptors: %s", strerror(-k)); + image_fd = setup_image(&device_path, &loop_nr); + if (image_fd < 0) { + r = image_fd; goto finish; } + + r = dissect_image(image_fd, + &root_device, &root_device_rw, + &home_device, &home_device_rw, + &srv_device, &srv_device_rw, + &secondary); + if (r < 0) + goto finish; } - fdset_close_others(fds); - log_open(); master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); if (master < 0) { @@ -1597,21 +3030,14 @@ int main(int argc, char *argv[]) { } if (!arg_quiet) - log_info("Spawning container %s on %s. Press ^] three times within 1s to abort execution.", arg_machine, arg_directory); + log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.", + arg_machine, arg_image ? arg_image : arg_directory); if (unlockpt(master) < 0) { log_error("Failed to unlock tty: %m"); goto finish; } - if (arg_network_veth) { - netns_fd = open("/proc/self/ns/net", O_RDWR|O_CLOEXEC); - if (netns_fd < 0) { - log_error("Failed to open network namespace fd: %m"); - goto finish; - } - } - if (access("/dev/kdbus/control", F_OK) >= 0) { if (arg_share_system) { @@ -1640,36 +3066,56 @@ int main(int argc, char *argv[]) { sd_notify(0, "READY=1"); assert_se(sigemptyset(&mask) == 0); + assert_se(sigemptyset(&mask_chld) == 0); + sigaddset(&mask_chld, SIGCHLD); sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); for (;;) { - siginfo_t status; + ContainerStatus container_status; + _barrier_destroy_ Barrier barrier = { }; + struct sigaction sa = { + .sa_handler = nop_handler, + .sa_flags = SA_NOCLDSTOP, + }; + + r = barrier_init(&barrier); + if (r < 0) { + log_error("Cannot initialize IPC barrier: %s", strerror(-r)); + goto finish; + } + + /* Child can be killed before execv(), so handle SIGCHLD + * in order to interrupt parent's blocking calls and + * give it a chance to call wait() and terminate. */ + r = sigprocmask(SIG_UNBLOCK, &mask_chld, NULL); + if (r < 0) { + log_error("Failed to change the signal mask: %m"); + goto finish; + } - sync_fd = eventfd(0, EFD_CLOEXEC); - if (sync_fd < 0) { - log_error("Failed to create event fd: %m"); + r = sigaction(SIGCHLD, &sa, NULL); + if (r < 0) { + log_error("Failed to install SIGCHLD handler: %m"); goto finish; } - pid = syscall(__NR_clone, - SIGCHLD|CLONE_NEWNS| - (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| - (arg_private_network ? CLONE_NEWNET : 0), NULL); + pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWNS| + (arg_share_system ? 0 : CLONE_NEWIPC|CLONE_NEWPID|CLONE_NEWUTS)| + (arg_private_network ? CLONE_NEWNET : 0), NULL); if (pid < 0) { if (errno == EINVAL) log_error("clone() failed, do you have namespace support enabled in your kernel? (You need UTS, IPC, PID and NET namespacing built in): %m"); else log_error("clone() failed: %m"); + r = pid; goto finish; } if (pid == 0) { /* child */ - const char *home = NULL; - uid_t uid = (uid_t) -1; - gid_t gid = (gid_t) -1; + _cleanup_free_ char *home = NULL; unsigned n_env = 2; const char *envp[] = { "PATH=" DEFAULT_PATH_SPLIT_USR, @@ -1684,21 +3130,20 @@ int main(int argc, char *argv[]) { NULL }; char **env_use; - eventfd_t x; + + barrier_set_role(&barrier, BARRIER_CHILD); envp[n_env] = strv_find_prefix(environ, "TERM="); if (envp[n_env]) n_env ++; - close_nointr_nofail(master); - master = -1; + master = safe_close(master); close_nointr(STDIN_FILENO); close_nointr(STDOUT_FILENO); close_nointr(STDERR_FILENO); - close_nointr_nofail(kmsg_socket_pair[0]); - kmsg_socket_pair[0] = -1; + kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]); reset_all_signal_handlers(); @@ -1708,31 +3153,31 @@ int main(int argc, char *argv[]) { k = open_terminal(console, O_RDWR); if (k != STDIN_FILENO) { if (k >= 0) { - close_nointr_nofail(k); + safe_close(k); k = -EINVAL; } log_error("Failed to open console: %s", strerror(-k)); - goto child_fail; + _exit(EXIT_FAILURE); } if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) { log_error("Failed to duplicate console: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } if (setsid() < 0) { log_error("setsid() failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } if (reset_audit_loginuid() < 0) - goto child_fail; + _exit(EXIT_FAILURE); if (prctl(PR_SET_PDEATHSIG, SIGKILL) < 0) { log_error("PR_SET_PDEATHSIG failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } /* Mark everything as slave, so that we still @@ -1740,91 +3185,109 @@ int main(int argc, char *argv[]) { * propagate mounts to the real root. */ if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0) { log_error("MS_SLAVE|MS_REC failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } + if (mount_devices(arg_directory, + root_device, root_device_rw, + home_device, home_device_rw, + srv_device, srv_device_rw) < 0) + _exit(EXIT_FAILURE); + /* Turn directory into bind mount */ if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REC, NULL) < 0) { - log_error("Failed to make bind mount."); - goto child_fail; + log_error("Failed to make bind mount: %m"); + _exit(EXIT_FAILURE); } - if (arg_read_only) - if (mount(arg_directory, arg_directory, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY|MS_REC, NULL) < 0) { - log_error("Failed to make read-only."); - goto child_fail; + r = setup_volatile(arg_directory); + if (r < 0) + _exit(EXIT_FAILURE); + + if (setup_volatile_state(arg_directory) < 0) + _exit(EXIT_FAILURE); + + r = base_filesystem_create(arg_directory); + if (r < 0) + _exit(EXIT_FAILURE); + + if (arg_read_only) { + k = bind_remount_recursive(arg_directory, true); + if (k < 0) { + log_error("Failed to make tree read-only: %s", strerror(-k)); + _exit(EXIT_FAILURE); } + } if (mount_all(arg_directory) < 0) - goto child_fail; + _exit(EXIT_FAILURE); if (copy_devnodes(arg_directory) < 0) - goto child_fail; + _exit(EXIT_FAILURE); if (setup_ptmx(arg_directory) < 0) - goto child_fail; + _exit(EXIT_FAILURE); dev_setup(arg_directory); - if (setup_veth(netns_fd) < 0) - goto child_fail; - - if (netns_fd >= 0) { - close_nointr_nofail(netns_fd); - netns_fd = -1; - } - - if (audit_still_doesnt_work_in_containers() < 0) - goto child_fail; + if (setup_seccomp() < 0) + _exit(EXIT_FAILURE); if (setup_dev_console(arg_directory, console) < 0) - goto child_fail; + _exit(EXIT_FAILURE); if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0) - goto child_fail; + _exit(EXIT_FAILURE); - close_nointr_nofail(kmsg_socket_pair[1]); - kmsg_socket_pair[1] = -1; + kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); if (setup_boot_id(arg_directory) < 0) - goto child_fail; + _exit(EXIT_FAILURE); if (setup_timezone(arg_directory) < 0) - goto child_fail; + _exit(EXIT_FAILURE); if (setup_resolv_conf(arg_directory) < 0) - goto child_fail; + _exit(EXIT_FAILURE); if (setup_journal(arg_directory) < 0) - goto child_fail; + _exit(EXIT_FAILURE); - if (mount_binds(arg_directory, arg_bind, 0) < 0) - goto child_fail; + if (mount_binds(arg_directory, arg_bind, false) < 0) + _exit(EXIT_FAILURE); - if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0) - goto child_fail; + if (mount_binds(arg_directory, arg_bind_ro, true) < 0) + _exit(EXIT_FAILURE); + + if (mount_tmpfs(arg_directory) < 0) + _exit(EXIT_FAILURE); if (setup_kdbus(arg_directory, kdbus_domain) < 0) - goto child_fail; + _exit(EXIT_FAILURE); + + /* Tell the parent that we are ready, and that + * it can cgroupify us to that we lack access + * to certain devices and resources. */ + barrier_place(&barrier); if (chdir(arg_directory) < 0) { log_error("chdir(%s) failed: %m", arg_directory); - goto child_fail; + _exit(EXIT_FAILURE); } if (mount(arg_directory, "/", NULL, MS_MOVE, NULL) < 0) { log_error("mount(MS_MOVE) failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } if (chroot(".") < 0) { log_error("chroot() failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } if (chdir("/") < 0) { log_error("chdir() failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } umask(0022); @@ -1834,76 +3297,26 @@ int main(int argc, char *argv[]) { if (drop_capabilities() < 0) { log_error("drop_capabilities() failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } - if (arg_user) { - - /* Note that this resolves user names - * inside the container, and hence - * accesses the NSS modules from the - * container and not the host. This is - * a bit weird... */ - - if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) { - log_error("get_user_creds() failed: %m"); - goto child_fail; - } - - if (mkdir_parents_label(home, 0775) < 0) { - log_error("mkdir_parents_label() failed: %m"); - goto child_fail; - } - - if (mkdir_safe_label(home, 0775, uid, gid) < 0) { - log_error("mkdir_safe_label() failed: %m"); - goto child_fail; - } - - if (initgroups((const char*)arg_user, gid) < 0) { - log_error("initgroups() failed: %m"); - goto child_fail; - } - - if (setresgid(gid, gid, gid) < 0) { - log_error("setregid() failed: %m"); - goto child_fail; - } - - if (setresuid(uid, uid, uid) < 0) { - log_error("setreuid() failed: %m"); - goto child_fail; - } - } else { - /* Reset everything fully to 0, just in case */ - - if (setgroups(0, NULL) < 0) { - log_error("setgroups() failed: %m"); - goto child_fail; - } - - if (setresgid(0, 0, 0) < 0) { - log_error("setregid() failed: %m"); - goto child_fail; - } - - if (setresuid(0, 0, 0) < 0) { - log_error("setreuid() failed: %m"); - goto child_fail; - } - } + r = change_uid_gid(&home); + if (r < 0) + _exit(EXIT_FAILURE); if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) || (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) || (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) { log_oom(); - goto child_fail; + _exit(EXIT_FAILURE); } if (!sd_id128_equal(arg_uuid, SD_ID128_NULL)) { - if (asprintf((char**)(envp + n_env++), "container_uuid=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(arg_uuid)) < 0) { + char as_uuid[37]; + + if (asprintf((char**)(envp + n_env++), "container_uuid=%s", id128_format_as_uuid(arg_uuid, as_uuid)) < 0) { log_oom(); - goto child_fail; + _exit(EXIT_FAILURE); } } @@ -1911,21 +3324,37 @@ int main(int argc, char *argv[]) { k = fdset_cloexec(fds, false); if (k < 0) { log_error("Failed to unset O_CLOEXEC for file descriptors."); - goto child_fail; + _exit(EXIT_FAILURE); } if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) || (asprintf((char **)(envp + n_env++), "LISTEN_PID=1") < 0)) { log_oom(); - goto child_fail; + _exit(EXIT_FAILURE); } } setup_hostname(); - eventfd_read(sync_fd, &x); - close_nointr_nofail(sync_fd); - sync_fd = -1; + if (arg_personality != 0xffffffffLU) { + if (personality(arg_personality) < 0) { + log_error("personality() failed: %m"); + _exit(EXIT_FAILURE); + } + } else if (secondary) { + if (personality(PER_LINUX32) < 0) { + log_error("personality() failed: %m"); + _exit(EXIT_FAILURE); + } + } + +#ifdef HAVE_SELINUX + if (arg_selinux_context) + if (setexeccon((security_context_t) arg_selinux_context) < 0) { + log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context); + _exit(EXIT_FAILURE); + } +#endif if (!strv_isempty(arg_setenv)) { char **n; @@ -1933,18 +3362,17 @@ int main(int argc, char *argv[]) { n = strv_env_merge(2, envp, arg_setenv); if (!n) { log_oom(); - goto child_fail; + _exit(EXIT_FAILURE); } env_use = n; } else env_use = (char**) envp; -#ifdef HAVE_SELINUX - if (arg_selinux_context) - if (setexeccon(arg_selinux_context) < 0) - log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context); -#endif + /* Wait until the parent is ready with the setup, too... */ + if (!barrier_place_and_sync(&barrier)) + _exit(EXIT_FAILURE); + if (arg_boot) { char **a; size_t l; @@ -1972,93 +3400,116 @@ int main(int argc, char *argv[]) { } log_error("execv() failed: %m"); - - child_fail: _exit(EXIT_FAILURE); } + barrier_set_role(&barrier, BARRIER_PARENT); fdset_free(fds); fds = NULL; - r = register_machine(pid); - if (r < 0) - goto finish; + /* wait for child-setup to be done */ + if (barrier_place_and_sync(&barrier)) { + int ifi = 0; - r = move_network_interfaces(pid); - if (r < 0) - goto finish; + r = move_network_interfaces(pid); + if (r < 0) + goto finish; - eventfd_write(sync_fd, 1); - close_nointr_nofail(sync_fd); - sync_fd = -1; + r = setup_veth(pid, veth_name, &ifi); + if (r < 0) + goto finish; - k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3); - if (k < 0) { - r = EXIT_FAILURE; - break; - } + r = setup_bridge(veth_name, &ifi); + if (r < 0) + goto finish; - if (!arg_quiet) - putc('\n', stdout); + r = setup_macvlan(pid); + if (r < 0) + goto finish; - /* Kill if it is not dead yet anyway */ - terminate_machine(pid); + r = register_machine(pid, ifi); + if (r < 0) + goto finish; - /* Redundant, but better safe than sorry */ - kill(pid, SIGKILL); + /* Block SIGCHLD here, before notifying child. + * process_pty() will handle it with the other signals. */ + r = sigprocmask(SIG_BLOCK, &mask_chld, NULL); + if (r < 0) + goto finish; - k = wait_for_terminate(pid, &status); - pid = 0; + /* Reset signal to default */ + r = default_signals(SIGCHLD, -1); + if (r < 0) + goto finish; - if (k < 0) { - r = EXIT_FAILURE; - break; - } + /* Notify the child that the parent is ready with all + * its setup, and that the child can now hand over + * control to the code to run inside the container. */ + barrier_place(&barrier); - if (status.si_code == CLD_EXITED) { - r = status.si_status; - if (status.si_status != 0) { - log_error("Container %s failed with error code %i.", arg_machine, status.si_status); + k = process_pty(master, &mask, arg_boot ? pid : 0, SIGRTMIN+3); + if (k < 0) { + r = EXIT_FAILURE; break; } if (!arg_quiet) - log_debug("Container %s exited successfully.", arg_machine); - break; - } else if (status.si_code == CLD_KILLED && - status.si_status == SIGINT) { + putc('\n', stdout); - if (!arg_quiet) - log_info("Container %s has been shut down.", arg_machine); - r = 0; - break; - } else if (status.si_code == CLD_KILLED && - status.si_status == SIGHUP) { + /* Kill if it is not dead yet anyway */ + terminate_machine(pid); + } - if (!arg_quiet) - log_info("Container %s is being rebooted.", arg_machine); - continue; - } else if (status.si_code == CLD_KILLED || - status.si_code == CLD_DUMPED) { + /* Normally redundant, but better safe than sorry */ + kill(pid, SIGKILL); - log_error("Container %s terminated by signal %s.", arg_machine, signal_to_string(status.si_status)); + r = wait_for_container(pid, &container_status); + pid = 0; + + if (r < 0) { + /* We failed to wait for the container, or the + * container exited abnormally */ r = EXIT_FAILURE; break; - } else { - log_error("Container %s failed due to unknown reason.", arg_machine); - r = EXIT_FAILURE; + } else if (r > 0 || container_status == CONTAINER_TERMINATED) + /* The container exited with a non-zero + * status, or with zero status and no reboot + * was requested. */ + break; + + /* CONTAINER_REBOOTED, loop again */ + + if (arg_keep_unit) { + /* Special handling if we are running as a + * service: instead of simply restarting the + * machine we want to restart the entire + * service, so let's inform systemd about this + * with the special exit code 133. The service + * file uses RestartForceExitStatus=133 so + * that this results in a full nspawn + * restart. This is necessary since we might + * have cgroup parameters set we want to have + * flushed out. */ + r = 133; break; } } finish: + loop_remove(loop_nr, &image_fd); + if (pid > 0) kill(pid, SIGKILL); free(arg_directory); free(arg_machine); - free(arg_setenv); - free(arg_network_interfaces); + free(arg_user); + strv_free(arg_setenv); + strv_free(arg_network_interfaces); + strv_free(arg_network_macvlan); + strv_free(arg_bind); + strv_free(arg_bind_ro); + strv_free(arg_tmpfs); return r; }