X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=7c47f6ecd2434356a3a1f6851fbdeaca2238f5bc;hp=097d7f779f0d781ba73e6662e48f35265fec4c8b;hb=7566e26721ee95d6fc864e9e6654fb61bd3cd603;hpb=0cb9fbcd44517ec90b2a678876194607beab5dec diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 097d7f779..7c47f6ecd 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -40,7 +40,6 @@ #include #include #include -#include #include #include #include @@ -86,11 +85,20 @@ #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, LINK_AUTO, @@ -98,6 +106,12 @@ 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 = {}; @@ -138,6 +152,7 @@ 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; @@ -149,6 +164,7 @@ 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) { @@ -192,11 +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; @@ -214,6 +232,7 @@ 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, @@ -223,6 +242,7 @@ static int parse_argv(int argc, char *argv[]) { ARG_NETWORK_VETH, ARG_NETWORK_BRIDGE, ARG_PERSONALITY, + ARG_VOLATILE, }; static const struct option options[] = { @@ -239,6 +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 }, @@ -254,6 +275,7 @@ static int parse_argv(int argc, char *argv[]) { { "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 }, {} }; @@ -461,6 +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; @@ -510,6 +568,25 @@ static int parse_argv(int argc, char *argv[]) { 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; @@ -536,6 +613,11 @@ static int parse_argv(int argc, char *argv[]) { 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; @@ -553,17 +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 }; @@ -628,11 +710,11 @@ 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; @@ -641,12 +723,14 @@ static int mount_binds(const char *dest, char **l, unsigned long flags) { 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) { @@ -656,11 +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)) @@ -679,8 +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; } } @@ -720,7 +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; @@ -739,7 +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; @@ -749,7 +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); @@ -763,11 +873,114 @@ 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 = {}; @@ -793,10 +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) { @@ -985,7 +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"); @@ -1136,10 +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) { @@ -1179,7 +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; @@ -1199,16 +1407,17 @@ 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 { _cleanup_bus_message_unref_ sd_bus_message *m = NULL; @@ -1218,7 +1427,7 @@ static int register_machine(pid_t pid) { "org.freedesktop.machine1", "/org/freedesktop/machine1", "org.freedesktop.machine1.Manager", - "CreateMachine"); + "CreateMachineWithNetwork"); if (r < 0) { log_error("Failed to create message: %s", strerror(-r)); return r; @@ -1226,13 +1435,14 @@ static int register_machine(pid_t pid) { r = sd_bus_message_append( m, - "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); if (r < 0) { log_error("Failed to append message arguments: %s", strerror(-r)); return r; @@ -1258,7 +1468,7 @@ static int register_machine(pid_t pid) { return r; } - r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 8, + r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 10, /* Allow the container to * access and create the API * device nodes, so that @@ -1277,7 +1487,18 @@ static int register_machine(pid_t pid) { * container to ever create * these device nodes. */ "/dev/pts/ptmx", "rw", - "char-pts", "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; @@ -1388,10 +1609,47 @@ static int reset_audit_loginuid(void) { return 0; } -static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { +#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; - int r; + struct ether_addr mac; + int r, i; if (!arg_private_network) return 0; @@ -1405,9 +1663,14 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { 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(&rtnl, 0); if (r < 0) { log_error("Failed to connect to netlink: %s", strerror(-r)); @@ -1432,13 +1695,7 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { return r; } - r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "veth"); - if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); - return r; - } - - r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA); + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "veth"); if (r < 0) { log_error("Failed to open netlink container: %s", strerror(-r)); return r; @@ -1456,6 +1713,12 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { return r; } + r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac); + if (r < 0) { + log_error("Failed to add netlink MAC address: %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)); @@ -1486,10 +1749,18 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) { 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 setup_bridge(const char veth_name[]) { +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; int r, bridge; @@ -1509,6 +1780,8 @@ static int setup_bridge(const char veth_name[]) { return -errno; } + *ifi = bridge; + r = sd_rtnl_open(&rtnl, 0); if (r < 0) { log_error("Failed to connect to netlink: %s", strerror(-r)); @@ -1521,6 +1794,12 @@ static int setup_bridge(const char veth_name[]) { return r; } + 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; + } + 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)); @@ -1691,13 +1970,7 @@ static int setup_macvlan(pid_t pid) { return r; } - r = sd_rtnl_message_append_string(m, IFLA_INFO_KIND, "macvlan"); - if (r < 0) { - log_error("Failed to append netlink kind: %s", strerror(-r)); - return r; - } - - r = sd_rtnl_message_open_container(m, IFLA_INFO_DATA); + 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; @@ -1731,22 +2004,25 @@ static int setup_macvlan(pid_t pid) { return 0; } -static int audit_still_doesnt_work_in_containers(void) { +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; - /* - 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. - */ - seccomp = seccomp_init(SCMP_ACT_ALLOW); if (!seccomp) return log_oom(); @@ -1757,6 +2033,26 @@ static int audit_still_doesnt_work_in_containers(void) { 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), @@ -1880,9 +2176,9 @@ static int setup_image(char **device_path, int *loop_nr) { static int dissect_image( int fd, - char **root_device, - char **home_device, - char **srv_device, + 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 @@ -1893,6 +2189,7 @@ static int dissect_image( _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; @@ -1982,6 +2279,7 @@ static int dissect_image( 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; @@ -2012,6 +2310,10 @@ static int dissect_image( 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; @@ -2029,6 +2331,8 @@ static int dissect_image( continue; home_nr = nr; + home_rw = !(flags & GPT_FLAG_READ_ONLY); + free(home); home = strdup(node); if (!home) @@ -2039,6 +2343,8 @@ static int dissect_image( continue; srv_nr = nr; + srv_rw = !(flags & GPT_FLAG_READ_ONLY); + free(srv); srv = strdup(node); if (!srv) @@ -2051,6 +2357,8 @@ static int dissect_image( continue; root_nr = nr; + root_rw = !(flags & GPT_FLAG_READ_ONLY); + free(root); root = strdup(node); if (!root) @@ -2064,6 +2372,9 @@ static int dissect_image( continue; secondary_root_nr = nr; + secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + + free(secondary_root); secondary_root = strdup(node); if (!secondary_root) @@ -2081,21 +2392,29 @@ static int dissect_image( 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; @@ -2105,7 +2424,7 @@ static int dissect_image( #endif } -static int mount_device(const char *what, const char *where, const char *directory) { +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; @@ -2114,6 +2433,9 @@ static int mount_device(const char *what, const char *where, const char *directo assert(what); assert(where); + if (arg_read_only) + rw = false; + if (directory) p = strappenda(where, directory); else @@ -2156,7 +2478,7 @@ static int mount_device(const char *what, const char *where, const char *directo return -ENOTSUP; } - if (mount(what, p, fstype, arg_read_only ? MS_NODEV|MS_RDONLY : 0, NULL) < 0) { + if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) { log_error("Failed to mount %s: %m", what); return -errno; } @@ -2168,13 +2490,17 @@ static int mount_device(const char *what, const char *where, const char *directo #endif } -static int mount_devices(const char *where, const char *root_device, const char *home_device, const char *srv_device) { +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); + 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; @@ -2182,7 +2508,7 @@ static int mount_devices(const char *where, const char *root_device, const char } if (home_device) { - r = mount_device(home_device, arg_directory, "/home"); + 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; @@ -2190,7 +2516,7 @@ static int mount_devices(const char *where, const char *root_device, const char } if (srv_device) { - r = mount_device(srv_device, arg_directory, "/srv"); + 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; @@ -2208,8 +2534,7 @@ static void loop_remove(int nr, int *image_fd) { if (image_fd && *image_fd >= 0) { ioctl(*image_fd, LOOP_CLR_FD); - close_nointr_nofail(*image_fd); - *image_fd = -1; + *image_fd = safe_close(*image_fd); } control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK); @@ -2244,9 +2569,9 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) { _exit(EXIT_FAILURE); if (pipe_fds[0] > 2) - close_nointr_nofail(pipe_fds[0]); + safe_close(pipe_fds[0]); if (pipe_fds[1] > 2) - close_nointr_nofail(pipe_fds[1]); + safe_close(pipe_fds[1]); nullfd = open("/dev/null", O_RDWR); if (nullfd < 0) @@ -2259,18 +2584,17 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) { _exit(EXIT_FAILURE); if (nullfd > 2) - close_nointr_nofail(nullfd); + safe_close(nullfd); reset_all_signal_handlers(); close_all_fds(NULL, 0); - execle("/usr/bin/getent", "getenv", database, key, NULL, &empty_env); - execle("/usr/getent", "getenv", database, key, NULL, &empty_env); + execle("/usr/bin/getent", "getent", database, key, NULL, &empty_env); + execle("/bin/getent", "getent", database, key, NULL, &empty_env); _exit(EXIT_FAILURE); } - close_nointr_nofail(pipe_fds[1]); - pipe_fds[1] = -1; + pipe_fds[1] = safe_close(pipe_fds[1]); *rpid = pid; @@ -2278,15 +2602,13 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) { } static int change_uid_gid(char **_home) { - - _cleanup_strv_free_ char **passwd = NULL; 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, l; + size_t sz = 0, l; uid_t uid; gid_t gid; pid_t pid; @@ -2453,7 +2775,7 @@ static int change_uid_gid(char **_home) { } r = mkdir_safe(home, 0755, uid, gid); - if (r < 0) { + if (r < 0 && r != -EEXIST) { log_error("Failed to make home directory: %s", strerror(-r)); return r; } @@ -2485,18 +2807,99 @@ static int change_uid_gid(char **_home) { 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_warning("Failed to wait for container: %s", strerror(-r)); + return 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); + + *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_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_pipe_ int kmsg_socket_pair[2] = { -1, -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; char veth_name[IFNAMSIZ]; bool secondary = false; + sigset_t mask, mask_chld; pid_t pid = 0; - sigset_t mask; log_parse_environment(); log_open(); @@ -2570,7 +2973,7 @@ int main(int argc, char *argv[]) { 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); + log_error("Directory %s doesn't look like an OS root directory (os-release file is missing). Refusing.", arg_directory); goto finish; } } else { @@ -2605,7 +3008,11 @@ int main(int argc, char *argv[]) { goto finish; } - r = dissect_image(image_fd, &root_device, &home_device, &srv_device, &secondary); + 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; } @@ -2623,7 +3030,8 @@ 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_image ? arg_image : 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"); @@ -2658,36 +3066,50 @@ 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 (;;) { - int parent_ready_fd = -1, child_ready_fd = -1; - siginfo_t status; - eventfd_t x; + ContainerStatus container_status; + _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL; + struct sigaction sa = { + .sa_handler = nop_handler, + .sa_flags = SA_NOCLDSTOP, + }; + + r = barrier_create(&barrier); + if (r < 0) { + log_error("Cannot initialize IPC barrier: %s", strerror(-r)); + goto finish; + } - parent_ready_fd = eventfd(0, EFD_CLOEXEC); - if (parent_ready_fd < 0) { - log_error("Failed to create event fd: %m"); + /* 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; } - child_ready_fd = eventfd(0, EFD_CLOEXEC); - if (child_ready_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; } @@ -2709,19 +3131,19 @@ int main(int argc, char *argv[]) { }; char **env_use; + 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(); @@ -2731,31 +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 @@ -2763,93 +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, home_device, srv_device) < 0) - goto child_fail; + 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 (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. */ - eventfd_write(child_ready_fd, 1); - close_nointr_nofail(child_ready_fd); - child_ready_fd = -1; + 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); @@ -2859,24 +3297,26 @@ int main(int argc, char *argv[]) { if (drop_capabilities() < 0) { log_error("drop_capabilities() failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } r = change_uid_gid(&home); if (r < 0) - goto child_fail; + _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); } } @@ -2884,13 +3324,13 @@ 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); } } @@ -2899,12 +3339,12 @@ int main(int argc, char *argv[]) { if (arg_personality != 0xffffffffLU) { if (personality(arg_personality) < 0) { log_error("personality() failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } } else if (secondary) { if (personality(PER_LINUX32) < 0) { log_error("personality() failed: %m"); - goto child_fail; + _exit(EXIT_FAILURE); } } @@ -2912,7 +3352,7 @@ int main(int argc, char *argv[]) { if (arg_selinux_context) if (setexeccon((security_context_t) arg_selinux_context) < 0) { log_error("setexeccon(\"%s\") failed: %m", arg_selinux_context); - goto child_fail; + _exit(EXIT_FAILURE); } #endif @@ -2922,7 +3362,7 @@ 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; @@ -2930,9 +3370,8 @@ int main(int argc, char *argv[]) { env_use = (char**) envp; /* Wait until the parent is ready with the setup, too... */ - eventfd_read(parent_ready_fd, &x); - close_nointr_nofail(parent_ready_fd); - parent_ready_fd = -1; + if (!barrier_place_and_sync(&barrier)) + _exit(EXIT_FAILURE); if (arg_boot) { char **a; @@ -2961,100 +3400,97 @@ 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; - /* Wait until the child reported that it is ready with - * all it needs to do with priviliges. After we got - * the notification we can make the process join its - * cgroup which might limit what it can do */ - eventfd_read(child_ready_fd, &x); - - r = register_machine(pid); - if (r < 0) - goto finish; - - r = move_network_interfaces(pid); - if (r < 0) - goto finish; + /* wait for child-setup to be done */ + if (barrier_place_and_sync(&barrier)) { + int ifi = 0; - r = setup_veth(pid, veth_name); - if (r < 0) - goto finish; - - r = setup_bridge(veth_name); - if (r < 0) - goto finish; - - r = setup_macvlan(pid); - if (r < 0) - goto finish; + r = move_network_interfaces(pid); + if (r < 0) + goto finish; - /* Notify the child that the parent is ready with all - * its setup, and thtat the child can now hand over - * control to the code to run inside the container. */ - eventfd_write(parent_ready_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; } } @@ -3073,6 +3509,7 @@ finish: strv_free(arg_network_macvlan); strv_free(arg_bind); strv_free(arg_bind_ro); + strv_free(arg_tmpfs); return r; }