X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=a9b9a3e062f7a6d0abd78660ab5c3e22bd3996b6;hp=e5a24dda70c0460ea6fb7931de869f87c3cd9e27;hb=4aab5d0cbd979b2cccb88534f118bceaa86466d8;hpb=080e78329a742dd95bdee321fee81c305e073491 diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index e5a24dda7..a9b9a3e06 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -43,6 +43,8 @@ #include #include #include +#include +#include #ifdef HAVE_SELINUX #include @@ -93,11 +95,22 @@ #include "cap-list.h" #include "btrfs-util.h" #include "machine-image.h" +#include "list.h" +#include "in-addr-util.h" +#include "fw-util.h" +#include "local-addresses.h" #ifdef HAVE_SECCOMP #include "seccomp-util.h" #endif +typedef struct ExposePort { + int protocol; + uint16_t host_port; + uint16_t container_port; + LIST_FIELDS(struct ExposePort, ports); +} ExposePort; + typedef enum ContainerStatus { CONTAINER_TERMINATED, CONTAINER_REBOOTED @@ -167,11 +180,16 @@ static bool arg_register = true; static bool arg_keep_unit = false; static char **arg_network_interfaces = NULL; static char **arg_network_macvlan = NULL; +static char **arg_network_ipvlan = NULL; static bool arg_network_veth = false; static const char *arg_network_bridge = NULL; static unsigned long arg_personality = 0xffffffffLU; static char *arg_image = NULL; static Volatile arg_volatile = VOLATILE_NO; +static ExposePort *arg_expose_ports = NULL; +static char **arg_property = NULL; +static uid_t arg_uid_shift = UID_INVALID, arg_uid_range = 0x10000U; +static bool arg_userns = false; static void help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" @@ -190,6 +208,7 @@ static void help(void) { " -M --machine=NAME Set the machine name for the container\n" " --uuid=UUID Set a specific machine UUID for the container\n" " -S --slice=SLICE Place the container in the specified slice\n" + " --property=NAME=VALUE Set scope unit property\n" " --private-network Disable network in container\n" " --network-interface=INTERFACE\n" " Assign an existing network interface to the\n" @@ -197,12 +216,19 @@ static void help(void) { " --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" + " --network-ipvlan=INTERFACE\n" + " Create a ipvlan network interface based on an\n" + " existing network interface to the container\n" + " -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" + " --private-users[=UIDBASE[:NUIDS]]\n" + " Run within user namespace\n" + " -p --port=[PROTOCOL:]HOSTPORT[:CONTAINERPORT]\n" + " Expose a container IP port on the host\n" " -Z --selinux-context=SECLABEL\n" " Set the SELinux security context to be used by\n" " processes in the container\n" @@ -225,8 +251,8 @@ static void help(void) { " --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" - " --volatile[=MODE] Run the system in volatile mode\n", - program_invocation_short_name); + " --volatile[=MODE] Run the system in volatile mode\n" + , program_invocation_short_name); } static int set_sanitized_path(char **b, const char *path) { @@ -269,11 +295,13 @@ static int parse_argv(int argc, char *argv[]) { ARG_KEEP_UNIT, ARG_NETWORK_INTERFACE, ARG_NETWORK_MACVLAN, - ARG_NETWORK_VETH, + ARG_NETWORK_IPVLAN, ARG_NETWORK_BRIDGE, ARG_PERSONALITY, ARG_VOLATILE, ARG_TEMPLATE, + ARG_PROPERTY, + ARG_PRIVATE_USERS, }; static const struct option options[] = { @@ -304,11 +332,15 @@ static int parse_argv(int argc, char *argv[]) { { "keep-unit", no_argument, NULL, ARG_KEEP_UNIT }, { "network-interface", required_argument, NULL, ARG_NETWORK_INTERFACE }, { "network-macvlan", required_argument, NULL, ARG_NETWORK_MACVLAN }, - { "network-veth", no_argument, NULL, ARG_NETWORK_VETH }, + { "network-ipvlan", required_argument, NULL, ARG_NETWORK_IPVLAN }, + { "network-veth", no_argument, NULL, 'n' }, { "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 }, + { "port", required_argument, NULL, 'p' }, + { "property", required_argument, NULL, ARG_PROPERTY }, + { "private-users", optional_argument, NULL, ARG_PRIVATE_USERS }, {} }; @@ -318,7 +350,7 @@ 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:qi:x", options, NULL)) >= 0) + while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:qi:xp:n", options, NULL)) >= 0) switch (c) { @@ -369,7 +401,7 @@ static int parse_argv(int argc, char *argv[]) { /* fall through */ - case ARG_NETWORK_VETH: + case 'n': arg_network_veth = true; arg_private_network = true; break; @@ -385,6 +417,13 @@ static int parse_argv(int argc, char *argv[]) { if (strv_extend(&arg_network_macvlan, optarg) < 0) return log_oom(); + arg_private_network = true; + break; + + case ARG_NETWORK_IPVLAN: + if (strv_extend(&arg_network_ipvlan, optarg) < 0) + return log_oom(); + /* fall through */ case ARG_PRIVATE_NETWORK: @@ -643,6 +682,100 @@ static int parse_argv(int argc, char *argv[]) { break; + case 'p': { + const char *split, *e; + uint16_t container_port, host_port; + int protocol; + ExposePort *p; + + if ((e = startswith(optarg, "tcp:"))) + protocol = IPPROTO_TCP; + else if ((e = startswith(optarg, "udp:"))) + protocol = IPPROTO_UDP; + else { + e = optarg; + protocol = IPPROTO_TCP; + } + + split = strchr(e, ':'); + if (split) { + char v[split - e + 1]; + + memcpy(v, e, split - e); + v[split - e] = 0; + + r = safe_atou16(v, &host_port); + if (r < 0 || host_port <= 0) { + log_error("Failed to parse host port: %s", optarg); + return -EINVAL; + } + + r = safe_atou16(split + 1, &container_port); + } else { + r = safe_atou16(e, &container_port); + host_port = container_port; + } + + if (r < 0 || container_port <= 0) { + log_error("Failed to parse host port: %s", optarg); + return -EINVAL; + } + + LIST_FOREACH(ports, p, arg_expose_ports) { + if (p->protocol == protocol && p->host_port == host_port) { + log_error("Duplicate port specification: %s", optarg); + return -EINVAL; + } + } + + p = new(ExposePort, 1); + if (!p) + return log_oom(); + + p->protocol = protocol; + p->host_port = host_port; + p->container_port = container_port; + + LIST_PREPEND(ports, arg_expose_ports, p); + + break; + } + + case ARG_PROPERTY: + if (strv_extend(&arg_property, optarg) < 0) + return log_oom(); + + break; + + case ARG_PRIVATE_USERS: + if (optarg) { + _cleanup_free_ char *buffer = NULL; + const char *range, *shift; + + range = strchr(optarg, ':'); + if (range) { + buffer = strndup(optarg, range - optarg); + if (!buffer) + return log_oom(); + shift = buffer; + + range++; + if (safe_atou32(range, &arg_uid_range) < 0 || arg_uid_range <= 0) { + log_error("Failed to parse UID range: %s", range); + return -EINVAL; + } + } else + shift = optarg; + + if (parse_uid(shift, &arg_uid_shift) < 0) { + log_error("Failed to parse UID: %s", optarg); + return -EINVAL; + } + } + + arg_userns = true; + break; + case '?': return -EINVAL; @@ -698,6 +831,11 @@ static int parse_argv(int argc, char *argv[]) { return -EINVAL; } + if (arg_expose_ports && !arg_private_network) { + log_error("Cannot use --port= without private networking."); + return -EINVAL; + } + arg_retain = (arg_retain | plus | (arg_private_network ? 1ULL << CAP_NET_ADMIN : 0)) & ~minus; return 1; @@ -723,6 +861,7 @@ static int mount_all(const char *dest) { { "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", "/tmp", "tmpfs", "mode=1777", 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 */ @@ -733,10 +872,7 @@ static int mount_all(const char *dest) { int r = 0; for (k = 0; k < ELEMENTSOF(mount_table); k++) { - _cleanup_free_ char *where = NULL; -#ifdef HAVE_SELINUX - _cleanup_free_ char *options = NULL; -#endif + _cleanup_free_ char *where = NULL, *options = NULL; const char *o; int t; @@ -783,6 +919,19 @@ static int mount_all(const char *dest) { #endif o = mount_table[k].options; + if (arg_userns && arg_uid_shift != UID_INVALID && streq_ptr(mount_table[k].type, "tmpfs")) { + char *uid_options = NULL; + + if (o) + asprintf(&uid_options, "%s,uid=" UID_FMT ",gid=" UID_FMT, o, arg_uid_shift, arg_uid_shift); + else + asprintf(&uid_options, "uid=" UID_FMT ",gid=" UID_FMT, arg_uid_shift, arg_uid_shift); + if (!uid_options) + return log_oom(); + + free(options); + o = options = uid_options; + } if (mount(mount_table[k].what, where, @@ -820,8 +969,12 @@ static int mount_binds(const char *dest, char **l, bool ro) { 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); + if (S_ISDIR(source_st.st_mode) && !S_ISDIR(dest_st.st_mode)) { + log_error("Cannot bind mount directory %s on file %s.", *x, where); + return -EINVAL; + } + if (!S_ISDIR(source_st.st_mode) && S_ISDIR(dest_st.st_mode)) { + log_error("Cannot bind mount file %s on directory %s.", *x, where); return -EINVAL; } } else if (errno == ENOENT) { @@ -833,27 +986,18 @@ static int mount_binds(const char *dest, char **l, bool ro) { return -errno; } - /* Create the mount point, but be conservative -- refuse to create block - * and char devices. */ + /* Create the mount point. Any non-directory file can be + * mounted on any non-directory file (regular, fifo, socket, + * char, block). + */ if (S_ISDIR(source_st.st_mode)) { r = mkdir_label(where, 0755); if (r < 0 && errno != EEXIST) return log_error_errno(r, "Failed to create mount point %s: %m", where); - } else if (S_ISFIFO(source_st.st_mode)) { - r = mkfifo(where, 0644); - if (r < 0 && errno != EEXIST) - return log_error_errno(errno, "Failed to create mount point %s: %m", where); - } else if (S_ISSOCK(source_st.st_mode)) { - r = mknod(where, 0644 | S_IFSOCK, 0); - if (r < 0 && errno != EEXIST) - return log_error_errno(errno, "Failed to create mount point %s: %m", where); - } else if (S_ISREG(source_st.st_mode)) { + } else { r = touch(where); if (r < 0) return log_error_errno(r, "Failed to create mount point %s: %m", where); - } else { - log_error("Refusing to create mountpoint for file: %s", *x); - return -ENOTSUP; } if (mount(*x, where, "bind", MS_BIND, NULL) < 0) @@ -873,7 +1017,7 @@ static int mount_cgroup_hierarchy(const char *dest, const char *controller, cons char *to; int r; - to = strappenda(dest, "/sys/fs/cgroup/", hierarchy); + to = strjoina(dest, "/sys/fs/cgroup/", hierarchy); r = path_is_mount_point(to, false); if (r < 0) @@ -883,9 +1027,17 @@ static int mount_cgroup_hierarchy(const char *dest, const char *controller, cons mkdir_p(to, 0755); - if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV|(read_only ? MS_RDONLY : 0), controller) < 0) + /* The superblock mount options of the mount point need to be + * identical to the hosts', and hence writable... */ + if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, controller) < 0) return log_error_errno(errno, "Failed to mount to %s: %m", to); + /* ... hence let's only make the bind mount read-only, not the + * superblock. */ + if (read_only) { + if (mount(NULL, to, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0) + return log_error_errno(errno, "Failed to remount %s read-only: %m", to); + } return 1; } @@ -907,7 +1059,7 @@ static int mount_cgroup(const char *dest) { if (r < 0) return log_error_errno(r, "Failed to determine our own cgroup path: %m"); - cgroup_root = strappenda(dest, "/sys/fs/cgroup"); + cgroup_root = strjoina(dest, "/sys/fs/cgroup"); if (mount("tmpfs", cgroup_root, "tmpfs", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME, "mode=755") < 0) return log_error_errno(errno, "Failed to mount tmpfs to /sys/fs/cgroup: %m"); @@ -951,21 +1103,21 @@ static int mount_cgroup(const char *dest) { return r; if (symlink(combined, target) < 0) - return log_error_errno(errno, "Failed to create symlink for combined hiearchy: %m"); + return log_error_errno(errno, "Failed to create symlink for combined hierarchy: %m"); } } - r = mount_cgroup_hierarchy(dest, "name=systemd", "systemd", false); + r = mount_cgroup_hierarchy(dest, "name=systemd,xattr", "systemd", false); if (r < 0) return r; /* Make our own cgroup a (writable) bind mount */ - systemd_own = strappenda(dest, "/sys/fs/cgroup/systemd", own_cgroup_path); + systemd_own = strjoina(dest, "/sys/fs/cgroup/systemd", own_cgroup_path); if (mount(systemd_own, systemd_own, NULL, MS_BIND, NULL) < 0) return log_error_errno(errno, "Failed to turn %s into a bind mount: %m", own_cgroup_path); /* And then remount the systemd cgroup root read-only */ - systemd_root = strappenda(dest, "/sys/fs/cgroup/systemd"); + systemd_root = strjoina(dest, "/sys/fs/cgroup/systemd"); if (mount(NULL, systemd_root, NULL, MS_BIND|MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_RDONLY, NULL) < 0) return log_error_errno(errno, "Failed to mount cgroup root read-only: %m"); @@ -1092,7 +1244,7 @@ static int setup_resolv_conf(const char *dest) { return 0; } - r = copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644); + r = copy_file("/etc/resolv.conf", where, O_TRUNC|O_NOFOLLOW, 0644, 0); if (r < 0) { log_warning_errno(r, "Failed to copy /etc/resolv.conf to %s: %m", where); @@ -1118,7 +1270,7 @@ static int setup_volatile_state(const char *directory) { if (r < 0) return log_error_errno(r, "Failed to remount %s read-only: %m", directory); - p = strappenda(directory, "/var"); + p = strjoina(directory, "/var"); r = mkdir(p, 0755); if (r < 0 && errno != EEXIST) return log_error_errno(errno, "Failed to create %s: %m", directory); @@ -1154,8 +1306,8 @@ static int setup_volatile(const char *directory) { tmpfs_mounted = true; - f = strappenda(directory, "/usr"); - t = strappenda(template, "/usr"); + f = strjoina(directory, "/usr"); + t = strjoina(template, "/usr"); r = mkdir(t, 0755); if (r < 0 && errno != EEXIST) { @@ -1292,6 +1444,10 @@ static int copy_devnodes(const char *dest) { if (mknod(to, st.st_mode, st.st_rdev) < 0) return log_error_errno(errno, "mknod(%s) failed: %m", to); + + if (arg_userns && arg_uid_shift != UID_INVALID) + if (lchown(to, arg_uid_shift, arg_uid_shift) < 0) + return log_error_errno(errno, "chown() of device node %s failed: %m", to); } } @@ -1308,6 +1464,10 @@ static int setup_ptmx(const char *dest) { if (symlink("pts/ptmx", p) < 0) return log_error_errno(errno, "Failed to create /dev/ptmx symlink: %m"); + if (arg_userns && arg_uid_shift != UID_INVALID) + if (lchown(p, arg_uid_shift, arg_uid_shift) < 0) + return log_error_errno(errno, "lchown() of symlink %s failed: %m", p); + return 0; } @@ -1337,7 +1497,7 @@ static int setup_dev_console(const char *dest, const char *console) { * /dev/console. (Note that the major minor doesn't actually * matter here, since we mount it over anyway). */ - to = strappenda(dest, "/dev/console"); + to = strjoina(dest, "/dev/console"); if (mknod(to, (st.st_mode & ~07777) | 0600, st.st_rdev) < 0) return log_error_errno(errno, "mknod() for /dev/console failed: %m"); @@ -1349,8 +1509,8 @@ static int setup_dev_console(const char *dest, const char *console) { static int setup_kmsg(const char *dest, int kmsg_socket) { _cleanup_free_ char *from = NULL, *to = NULL; - int r, fd, k; _cleanup_umask_ mode_t u; + int r, fd, k; union { struct cmsghdr cmsghdr; uint8_t buf[CMSG_SPACE(sizeof(int))]; @@ -1401,7 +1561,7 @@ 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); + k = sendmsg(kmsg_socket, &mh, MSG_NOSIGNAL); safe_close(fd); if (k < 0) @@ -1412,6 +1572,198 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { return 0; } +static int send_rtnl(int send_fd) { + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(int))]; + } control = {}; + struct msghdr mh = { + .msg_control = &control, + .msg_controllen = sizeof(control), + }; + struct cmsghdr *cmsg; + _cleanup_close_ int fd = -1; + ssize_t k; + + assert(send_fd >= 0); + + if (!arg_expose_ports) + return 0; + + fd = socket(PF_NETLINK, SOCK_RAW|SOCK_CLOEXEC|SOCK_NONBLOCK, NETLINK_ROUTE); + if (fd < 0) + return log_error_errno(errno, "failed to allocate container netlink: %m"); + + cmsg = CMSG_FIRSTHDR(&mh); + cmsg->cmsg_level = SOL_SOCKET; + cmsg->cmsg_type = SCM_RIGHTS; + cmsg->cmsg_len = CMSG_LEN(sizeof(int)); + memcpy(CMSG_DATA(cmsg), &fd, sizeof(int)); + + mh.msg_controllen = cmsg->cmsg_len; + + /* Store away the fd in the socket, so that it stays open as + * long as we run the child */ + k = sendmsg(send_fd, &mh, MSG_NOSIGNAL); + if (k < 0) + return log_error_errno(errno, "Failed to send netlink fd: %m"); + + return 0; +} + +static int flush_ports(union in_addr_union *exposed) { + ExposePort *p; + int r, af = AF_INET; + + assert(exposed); + + if (!arg_expose_ports) + return 0; + + if (in_addr_is_null(af, exposed)) + return 0; + + log_debug("Lost IP address."); + + LIST_FOREACH(ports, p, arg_expose_ports) { + r = fw_add_local_dnat(false, + af, + p->protocol, + NULL, + NULL, 0, + NULL, 0, + p->host_port, + exposed, + p->container_port, + NULL); + if (r < 0) + log_warning_errno(r, "Failed to modify firewall: %m"); + } + + *exposed = IN_ADDR_NULL; + return 0; +} + +static int expose_ports(sd_rtnl *rtnl, union in_addr_union *exposed) { + _cleanup_free_ struct local_address *addresses = NULL; + _cleanup_free_ char *pretty = NULL; + union in_addr_union new_exposed; + ExposePort *p; + bool add; + int af = AF_INET, r; + + assert(exposed); + + /* Invoked each time an address is added or removed inside the + * container */ + + if (!arg_expose_ports) + return 0; + + r = local_addresses(rtnl, 0, af, &addresses); + if (r < 0) + return log_error_errno(r, "Failed to enumerate local addresses: %m"); + + add = r > 0 && + addresses[0].family == af && + addresses[0].scope < RT_SCOPE_LINK; + + if (!add) + return flush_ports(exposed); + + new_exposed = addresses[0].address; + if (in_addr_equal(af, exposed, &new_exposed)) + return 0; + + in_addr_to_string(af, &new_exposed, &pretty); + log_debug("New container IP is %s.", strna(pretty)); + + LIST_FOREACH(ports, p, arg_expose_ports) { + + r = fw_add_local_dnat(true, + af, + p->protocol, + NULL, + NULL, 0, + NULL, 0, + p->host_port, + &new_exposed, + p->container_port, + in_addr_is_null(af, exposed) ? NULL : exposed); + if (r < 0) + log_warning_errno(r, "Failed to modify firewall: %m"); + } + + *exposed = new_exposed; + return 0; +} + +static int on_address_change(sd_rtnl *rtnl, sd_rtnl_message *m, void *userdata) { + union in_addr_union *exposed = userdata; + + assert(rtnl); + assert(m); + assert(exposed); + + expose_ports(rtnl, exposed); + return 0; +} + +static int watch_rtnl(sd_event *event, int recv_fd, union in_addr_union *exposed, sd_rtnl **ret) { + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(int))]; + } control = {}; + struct msghdr mh = { + .msg_control = &control, + .msg_controllen = sizeof(control), + }; + struct cmsghdr *cmsg; + _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + int fd, r; + ssize_t k; + + assert(event); + assert(recv_fd >= 0); + assert(ret); + + if (!arg_expose_ports) + return 0; + + k = recvmsg(recv_fd, &mh, MSG_NOSIGNAL); + if (k < 0) + return log_error_errno(errno, "Failed to recv netlink fd: %m"); + + cmsg = CMSG_FIRSTHDR(&mh); + assert(cmsg->cmsg_level == SOL_SOCKET); + assert(cmsg->cmsg_type == SCM_RIGHTS); + assert(cmsg->cmsg_len == CMSG_LEN(sizeof(int))); + memcpy(&fd, CMSG_DATA(cmsg), sizeof(int)); + + r = sd_rtnl_open_fd(&rtnl, fd, 1, RTNLGRP_IPV4_IFADDR); + if (r < 0) { + safe_close(fd); + return log_error_errno(r, "Failed to create rtnl object: %m"); + } + + r = sd_rtnl_add_match(rtnl, RTM_NEWADDR, on_address_change, exposed); + if (r < 0) + return log_error_errno(r, "Failed to subscribe to RTM_NEWADDR messages: %m"); + + r = sd_rtnl_add_match(rtnl, RTM_DELADDR, on_address_change, exposed); + if (r < 0) + return log_error_errno(r, "Failed to subscribe to RTM_DELADDR messages: %m"); + + r = sd_rtnl_attach_event(rtnl, event, 0); + if (r < 0) + return log_error_errno(r, "Failed to add to even loop: %m"); + + *ret = rtnl; + rtnl = NULL; + + return 0; +} + static int setup_hostname(void) { if (arg_share_system) @@ -1608,6 +1960,7 @@ static int register_machine(pid_t pid, int local_ifindex) { local_ifindex > 0 ? 1 : 0, local_ifindex); } else { _cleanup_bus_message_unref_ sd_bus_message *m = NULL; + char **i; r = sd_bus_message_new_method_call( bus, @@ -1617,7 +1970,7 @@ static int register_machine(pid_t pid, int local_ifindex) { "org.freedesktop.machine1.Manager", "CreateMachineWithNetwork"); if (r < 0) - return log_error_errno(r, "Failed to create message: %m"); + return bus_log_create_error(r); r = sd_bus_message_append( m, @@ -1630,21 +1983,21 @@ static int register_machine(pid_t pid, int local_ifindex) { strempty(arg_directory), local_ifindex > 0 ? 1 : 0, local_ifindex); if (r < 0) - return log_error_errno(r, "Failed to append message arguments: %m"); + return bus_log_create_error(r); r = sd_bus_message_open_container(m, 'a', "(sv)"); if (r < 0) - return log_error_errno(r, "Failed to open container: %m"); + return bus_log_create_error(r); if (!isempty(arg_slice)) { r = sd_bus_message_append(m, "(sv)", "Slice", "s", arg_slice); if (r < 0) - return log_error_errno(r, "Failed to append slice: %m"); + return bus_log_create_error(r); } r = sd_bus_message_append(m, "(sv)", "DevicePolicy", "s", "strict"); if (r < 0) - return log_error_errno(r, "Failed to add device policy: %m"); + return bus_log_create_error(r); r = sd_bus_message_append(m, "(sv)", "DeviceAllow", "a(ss)", 9, /* Allow the container to @@ -1670,9 +2023,23 @@ static int register_machine(pid_t pid, int local_ifindex) { if (r < 0) return log_error_errno(r, "Failed to add device whitelist: %m"); + STRV_FOREACH(i, arg_property) { + r = sd_bus_message_open_container(m, 'r', "sv"); + if (r < 0) + return bus_log_create_error(r); + + r = bus_append_unit_property_assignment(m, *i); + if (r < 0) + return r; + + r = sd_bus_message_close_container(m); + if (r < 0) + return bus_log_create_error(r); + } + r = sd_bus_message_close_container(m); if (r < 0) - return log_error_errno(r, "Failed to close container: %m"); + return bus_log_create_error(r); r = sd_bus_call(bus, m, 0, &error, NULL); } @@ -2109,21 +2476,105 @@ static int setup_macvlan(pid_t pid) { return 0; } +static int setup_ipvlan(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_ipvlan)) + return 0; + + r = sd_rtnl_open(&rtnl, 0); + if (r < 0) + return log_error_errno(r, "Failed to connect to netlink: %m"); + + udev = udev_new(); + if (!udev) { + log_error("Failed to connect to udev."); + return -ENOMEM; + } + + STRV_FOREACH(i, arg_network_ipvlan) { + _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) + return log_error_errno(r, "Failed to allocate netlink message: %m"); + + r = sd_rtnl_message_append_u32(m, IFLA_LINK, ifi); + if (r < 0) + return log_error_errno(r, "Failed to add netlink interface index: %m"); + + n = strappend("iv-", *i); + if (!n) + return log_oom(); + + strshorten(n, IFNAMSIZ-1); + + r = sd_rtnl_message_append_string(m, IFLA_IFNAME, n); + if (r < 0) + return log_error_errno(r, "Failed to add netlink interface name: %m"); + + r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid); + if (r < 0) + return log_error_errno(r, "Failed to add netlink namespace field: %m"); + + r = sd_rtnl_message_open_container(m, IFLA_LINKINFO); + if (r < 0) + return log_error_errno(r, "Failed to open netlink container: %m"); + + r = sd_rtnl_message_open_container_union(m, IFLA_INFO_DATA, "ipvlan"); + if (r < 0) + return log_error_errno(r, "Failed to open netlink container: %m"); + + r = sd_rtnl_message_append_u16(m, IFLA_IPVLAN_MODE, IPVLAN_MODE_L2); + if (r < 0) + return log_error_errno(r, "Failed to add ipvlan mode: %m"); + + r = sd_rtnl_message_close_container(m); + if (r < 0) + return log_error_errno(r, "Failed to close netlink container: %m"); + + r = sd_rtnl_message_close_container(m); + if (r < 0) + return log_error_errno(r, "Failed to close netlink container: %m"); + + r = sd_rtnl_call(rtnl, m, 0, NULL); + if (r < 0) + return log_error_errno(r, "Failed to add new ipvlan interfaces: %m"); + } + + 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), }; + static const int kmod_blacklist[] = { + SCMP_SYS(init_module), + SCMP_SYS(finit_module), + SCMP_SYS(delete_module), + }; + scmp_filter_ctx seccomp; unsigned i; int r; @@ -2148,8 +2599,22 @@ static int setup_seccomp(void) { } } - /* - Audit is broken in containers, much of the userspace audit + /* If the CAP_SYS_MODULE capability is not requested then + * we'll block the kmod syscalls too */ + if (!(arg_retain & (1ULL << CAP_SYS_MODULE))) { + for (i = 0; i < ELEMENTSOF(kmod_blacklist); i++) { + r = seccomp_rule_add(seccomp, SCMP_ACT_ERRNO(EPERM), kmod_blacklist[i], 0); + if (r == -EFAULT) + continue; /* unknown syscall */ + if (r < 0) { + log_error_errno(r, "Failed to block syscall: %m"); + 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. @@ -2194,10 +2659,10 @@ static int setup_propagate(const char *root) { (void) mkdir_p("/run/systemd/nspawn/", 0755); (void) mkdir_p("/run/systemd/nspawn/propagate", 0600); - p = strappenda("/run/systemd/nspawn/propagate/", arg_machine); + p = strjoina("/run/systemd/nspawn/propagate/", arg_machine); (void) mkdir_p(p, 0600); - q = strappenda(root, "/run/systemd/nspawn/incoming"); + q = strjoina(root, "/run/systemd/nspawn/incoming"); mkdir_parents(q, 0755); mkdir_p(q, 0600); @@ -2287,6 +2752,13 @@ static int setup_image(char **device_path, int *loop_nr) { return r; } +#define PARTITION_TABLE_BLURB \ + "Note that the disk image needs to either contain only a single MBR partition of\n" \ + "type 0x83 that is marked bootable, or a single GPT partition of type " \ + "0FC63DAF-8483-4772-8E79-3D69D8477DE4 or follow\n" \ + " http://www.freedesktop.org/wiki/Specifications/DiscoverablePartitionsSpec/\n" \ + "to be bootable with systemd-nspawn." + static int dissect_image( int fd, char **root_device, bool *root_device_rw, @@ -2302,17 +2774,18 @@ static int dissect_image( #ifdef GPT_ROOT_SECONDARY int secondary_root_nr = -1; #endif - - _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL; + _cleanup_free_ char *home = NULL, *root = NULL, *secondary_root = NULL, *srv = NULL, *generic = NULL; _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL; _cleanup_udev_device_unref_ struct udev_device *d = NULL; _cleanup_blkid_free_probe_ blkid_probe b = NULL; _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; + bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true, generic_rw = true; + bool is_gpt, is_mbr, multiple_generic = false; const char *pttype = NULL; blkid_partlist pl; struct stat st; + unsigned i; int r; assert(fd >= 0); @@ -2342,8 +2815,9 @@ static int dissect_image( 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); + log_error("Failed to identify any partition table on\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); return -EINVAL; } else if (r != 0) { if (errno == 0) @@ -2353,9 +2827,14 @@ static int dissect_image( } 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); + + is_gpt = streq_ptr(pttype, "gpt"); + is_mbr = streq_ptr(pttype, "dos"); + + if (!is_gpt && !is_mbr) { + log_error("No GPT or MBR partition table discovered on\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); return -EINVAL; } @@ -2380,24 +2859,83 @@ static int dissect_image( if (!d) return log_oom(); - e = udev_enumerate_new(udev); - if (!e) - return log_oom(); + for (i = 0;; i++) { + int n, m; - r = udev_enumerate_add_match_parent(e, d); - if (r < 0) - return log_oom(); + if (i >= 10) { + log_error("Kernel partitions never appeared."); + return -ENXIO; + } - r = udev_enumerate_scan_devices(e); - if (r < 0) - return log_error_errno(r, "Failed to scan for partition devices of %s: %m", arg_image); + 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) + return log_error_errno(r, "Failed to scan for partition devices of %s: %m", arg_image); + + /* Count the partitions enumerated by the kernel */ + n = 0; + first = udev_enumerate_get_list_entry(e); + udev_list_entry_foreach(item, first) + n++; + + /* Count the partitions enumerated by blkid */ + m = blkid_partlist_numof_partitions(pl); + if (n == m + 1) + break; + if (n > m + 1) { + log_error("blkid and kernel partition list do not match."); + return -EIO; + } + if (n < m + 1) { + unsigned j; + + /* The kernel has probed fewer partitions than + * blkid? Maybe the kernel prober is still + * running or it got EBUSY because udev + * already opened the device. Let's reprobe + * the device, which is a synchronous call + * that waits until probing is complete. */ + + for (j = 0; j < 20; j++) { + + r = ioctl(fd, BLKRRPART, 0); + if (r < 0) + r = -errno; + if (r >= 0 || r != -EBUSY) + break; + + /* If something else has the device + * open, such as an udev rule, the + * ioctl will return EBUSY. Since + * there's no way to wait until it + * isn't busy anymore, let's just wait + * a bit, and try again. + * + * This is really something they + * should fix in the kernel! */ + + usleep(50 * USEC_PER_MSEC); + } + + if (r < 0) + return log_error_errno(r, "Failed to reread partition table: %m"); + } + + e = udev_enumerate_unref(e); + } 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; + const char *node; unsigned long long flags; - sd_id128_t type_id; blkid_partition pp; dev_t qn; int nr; @@ -2428,82 +2966,110 @@ static int dissect_image( 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 (is_gpt) { + sd_id128_t type_id; + const char *stype; - if (sd_id128_from_string(stype, &type_id) < 0) - continue; + if (flags & GPT_FLAG_NO_AUTO) + continue; - if (sd_id128_equal(type_id, GPT_HOME)) { + stype = blkid_partition_get_type_string(pp); + if (!stype) + continue; - if (home && nr >= home_nr) + if (sd_id128_from_string(stype, &type_id) < 0) continue; - home_nr = nr; - home_rw = !(flags & GPT_FLAG_READ_ONLY); + if (sd_id128_equal(type_id, GPT_HOME)) { - free(home); - home = strdup(node); - if (!home) - return log_oom(); - } else if (sd_id128_equal(type_id, GPT_SRV)) { + if (home && nr >= home_nr) + continue; - if (srv && nr >= srv_nr) - continue; + home_nr = nr; + home_rw = !(flags & GPT_FLAG_READ_ONLY); - srv_nr = nr; - srv_rw = !(flags & GPT_FLAG_READ_ONLY); + r = free_and_strdup(&home, node); + if (r < 0) + return log_oom(); - free(srv); - srv = strdup(node); - if (!srv) - 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); + + r = free_and_strdup(&srv, node); + if (r < 0) + return log_oom(); + } #ifdef GPT_ROOT_NATIVE - else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { + else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE)) { - if (root && nr >= root_nr) - continue; + if (root && nr >= root_nr) + continue; - root_nr = nr; - root_rw = !(flags & GPT_FLAG_READ_ONLY); + root_nr = nr; + root_rw = !(flags & GPT_FLAG_READ_ONLY); - free(root); - root = strdup(node); - if (!root) - return log_oom(); - } + r = free_and_strdup(&root, node); + if (r < 0) + return log_oom(); + } #endif #ifdef GPT_ROOT_SECONDARY - else if (sd_id128_equal(type_id, 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); + + r = free_and_strdup(&secondary_root, node); + if (r < 0) + return log_oom(); + } +#endif + else if (sd_id128_equal(type_id, GPT_LINUX_GENERIC)) { - if (secondary_root && nr >= secondary_root_nr) + if (generic) + multiple_generic = true; + else { + generic_rw = !(flags & GPT_FLAG_READ_ONLY); + + r = free_and_strdup(&generic, node); + if (r < 0) + return log_oom(); + } + } + + } else if (is_mbr) { + int type; + + if (flags != 0x80) /* Bootable flag */ continue; - secondary_root_nr = nr; - secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + type = blkid_partition_get_type(pp); + if (type != 0x83) /* Linux partition */ + continue; + if (generic) + multiple_generic = true; + else { + generic_rw = true; - free(secondary_root); - secondary_root = strdup(node); - if (!secondary_root) - return log_oom(); + r = free_and_strdup(&root, node); + if (r < 0) + 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) { @@ -2518,6 +3084,31 @@ static int dissect_image( *root_device_rw = secondary_root_rw; *secondary = true; + } else if (generic) { + + /* There were no partitions with precise meanings + * around, but we found generic partitions. In this + * case, if there's only one, we can go ahead and boot + * it, otherwise we bail out, because we really cannot + * make any sense of it. */ + + if (multiple_generic) { + log_error("Identified multiple bootable Linux partitions on\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); + return -EINVAL; + } + + *root_device = generic; + generic = NULL; + + *root_device_rw = generic_rw; + *secondary = false; + } else { + log_error("Failed to identify root partition in disk image\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); + return -EINVAL; } if (home) { @@ -2554,7 +3145,7 @@ static int mount_device(const char *what, const char *where, const char *directo rw = false; if (directory) - p = strappenda(where, directory); + p = strjoina(where, directory); else p = where; @@ -2645,7 +3236,7 @@ static void loop_remove(int nr, int *image_fd) { if (image_fd && *image_fd >= 0) { r = ioctl(*image_fd, LOOP_CLR_FD); if (r < 0) - log_warning_errno(errno, "Failed to close loop image: %m"); + log_debug_errno(errno, "Failed to close loop image: %m"); *image_fd = safe_close(*image_fd); } @@ -2657,7 +3248,7 @@ static void loop_remove(int nr, int *image_fd) { r = ioctl(control, LOOP_CTL_REMOVE, nr); if (r < 0) - log_warning_errno(errno, "Failed to remove loop %d: %m", nr); + log_debug_errno(errno, "Failed to remove loop %d: %m", nr); } static int spawn_getent(const char *database, const char *key, pid_t *rpid) { @@ -3003,7 +3594,7 @@ static int determine_names(void) { return -ENOENT; } - if (i->type == IMAGE_GPT) + if (i->type == IMAGE_RAW) r = set_sanitized_path(&arg_image, i->path); else r = set_sanitized_path(&arg_directory, i->path); @@ -3054,12 +3645,43 @@ static int determine_names(void) { return 0; } +static int determine_uid_shift(void) { + int r; + + if (!arg_userns) + return 0; + + if (arg_uid_shift == UID_INVALID) { + struct stat st; + + r = stat(arg_directory, &st); + if (r < 0) + return log_error_errno(errno, "Failed to determine UID base of %s: %m", arg_directory); + + arg_uid_shift = st.st_uid & UINT32_C(0xffff0000); + + if (arg_uid_shift != (st.st_gid & UINT32_C(0xffff0000))) { + log_error("UID and GID base of %s don't match.", arg_directory); + return -EINVAL; + } + + arg_uid_range = UINT32_C(0x10000); + } + + if (arg_uid_shift > (uid_t) -1 - arg_uid_range) { + log_error("UID base too high for UID range."); + return -EINVAL; + } + + log_info("Using user namespaces with base " UID_FMT " and range " UID_FMT ".", arg_uid_shift, arg_uid_range); + return 0; +} + int main(int argc, char *argv[]) { _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL, *console = NULL; bool root_device_rw = true, home_device_rw = true, srv_device_rw = true; _cleanup_close_ int master = -1, image_fd = -1; - _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }; _cleanup_fdset_free_ FDSet *fds = NULL; int r, n_fd_passed, loop_nr = -1; char veth_name[IFNAMSIZ]; @@ -3067,6 +3689,9 @@ int main(int argc, char *argv[]) { sigset_t mask, mask_chld; pid_t pid = 0; int ret = EXIT_SUCCESS; + union in_addr_union exposed = {}; + _cleanup_release_lock_file_ LockFile tree_global_lock = LOCK_FILE_INIT, tree_local_lock = LOCK_FILE_INIT; + bool interactive; log_parse_environment(); log_open(); @@ -3112,20 +3737,7 @@ int main(int argc, char *argv[]) { goto finish; } - if (arg_template) { - r = btrfs_subvol_snapshot(arg_template, arg_directory, arg_read_only, true); - if (r == -EEXIST) { - if (!arg_quiet) - log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template); - } else if (r < 0) { - log_error_errno(r, "Couldn't create snapshort %s from %s: %m", arg_directory, arg_template); - goto finish; - } else { - if (!arg_quiet) - log_info("Populated %s from template %s.", arg_directory, arg_template); - } - - } else if (arg_ephemeral) { + if (arg_ephemeral) { char *np; /* If the specified path is a mount point we @@ -3148,6 +3760,12 @@ int main(int argc, char *argv[]) { goto finish; } + r = image_path_lock(np, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock); + if (r < 0) { + log_error_errno(r, "Failed to lock %s: %m", np); + goto finish; + } + r = btrfs_subvol_snapshot(arg_directory, np, arg_read_only, true); if (r < 0) { free(np); @@ -3159,6 +3777,31 @@ int main(int argc, char *argv[]) { arg_directory = np; remove_subvol = true; + + } else { + r = image_path_lock(arg_directory, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock); + if (r == -EBUSY) { + log_error_errno(r, "Directory tree %s is currently busy.", arg_directory); + goto finish; + } + if (r < 0) { + log_error_errno(r, "Failed to lock %s: %m", arg_directory); + return r; + } + + if (arg_template) { + r = btrfs_subvol_snapshot(arg_template, arg_directory, arg_read_only, true); + if (r == -EEXIST) { + if (!arg_quiet) + log_info("Directory %s already exists, not populating from template %s.", arg_directory, arg_template); + } else if (r < 0) { + log_error_errno(r, "Couldn't create snapshot %s from %s: %m", arg_directory, arg_template); + goto finish; + } else { + if (!arg_quiet) + log_info("Populated %s from template %s.", arg_directory, arg_template); + } + } } if (arg_boot) { @@ -3170,7 +3813,7 @@ int main(int argc, char *argv[]) { } else { const char *p; - p = strappenda(arg_directory, + p = strjoina(arg_directory, argc > optind && path_is_absolute(argv[optind]) ? argv[optind] : "/usr/bin/"); if (access(p, F_OK) < 0) { log_error("Directory %s lacks the binary to execute or doesn't look like a binary tree. Refusing.", arg_directory); @@ -3185,6 +3828,16 @@ int main(int argc, char *argv[]) { assert(arg_image); assert(!arg_template); + r = image_path_lock(arg_image, (arg_read_only ? LOCK_SH : LOCK_EX) | LOCK_NB, &tree_global_lock, &tree_local_lock); + if (r == -EBUSY) { + r = log_error_errno(r, "Disk image %s is currently busy.", arg_image); + goto finish; + } + if (r < 0) { + r = log_error_errno(r, "Failed to create image lock: %m"); + goto finish; + } + if (!mkdtemp(template)) { log_error_errno(errno, "Failed to create temporary directory: %m"); r = -errno; @@ -3212,6 +3865,12 @@ int main(int argc, char *argv[]) { goto finish; } + r = determine_uid_shift(); + if (r < 0) + goto finish; + + interactive = isatty(STDIN_FILENO) > 0 && isatty(STDOUT_FILENO) > 0; + master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); if (master < 0) { r = log_error_errno(errno, "Failed to acquire pseudo tty: %m"); @@ -3224,19 +3883,14 @@ int main(int argc, char *argv[]) { goto finish; } - if (!arg_quiet) - log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.", - arg_machine, arg_image ?: arg_directory); - if (unlockpt(master) < 0) { r = log_error_errno(errno, "Failed to unlock tty: %m"); goto finish; } - if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { - r = log_error_errno(errno, "Failed to create kmsg socket pair: %m"); - goto finish; - } + if (!arg_quiet) + log_info("Spawning container %s on %s.\nPress ^] three times within 1s to kill container.", + arg_machine, arg_image ?: arg_directory); assert_se(sigemptyset(&mask) == 0); sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); @@ -3246,6 +3900,7 @@ int main(int argc, char *argv[]) { assert_se(sigaddset(&mask_chld, SIGCHLD) == 0); for (;;) { + _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 }, rtnl_socket_pair[2] = { -1, -1 }; ContainerStatus container_status; _cleanup_(barrier_destroy) Barrier barrier = BARRIER_NULL; struct sigaction sa = { @@ -3259,6 +3914,16 @@ int main(int argc, char *argv[]) { goto finish; } + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { + r = log_error_errno(errno, "Failed to create kmsg socket pair: %m"); + goto finish; + } + + if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, rtnl_socket_pair) < 0) { + r = log_error_errno(errno, "Failed to create rtnl socket pair: %m"); + 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. */ @@ -3312,30 +3977,33 @@ int main(int argc, char *argv[]) { master = safe_close(master); - close_nointr(STDIN_FILENO); - close_nointr(STDOUT_FILENO); - close_nointr(STDERR_FILENO); - kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]); + rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]); reset_all_signal_handlers(); reset_signal_mask(); - r = open_terminal(console, O_RDWR); - if (r != STDIN_FILENO) { - if (r >= 0) { - safe_close(r); - r = -EINVAL; - } + if (interactive) { + close_nointr(STDIN_FILENO); + close_nointr(STDOUT_FILENO); + close_nointr(STDERR_FILENO); - log_error_errno(r, "Failed to open console: %m"); - _exit(EXIT_FAILURE); - } + r = open_terminal(console, O_RDWR); + if (r != STDIN_FILENO) { + if (r >= 0) { + safe_close(r); + r = -EINVAL; + } - if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || - dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) { - log_error_errno(errno, "Failed to duplicate console: %m"); - _exit(EXIT_FAILURE); + log_error_errno(r, "Failed to open console: %m"); + _exit(EXIT_FAILURE); + } + + if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || + dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) { + log_error_errno(errno, "Failed to duplicate console: %m"); + _exit(EXIT_FAILURE); + } } if (setsid() < 0) { @@ -3351,6 +4019,9 @@ int main(int argc, char *argv[]) { _exit(EXIT_FAILURE); } + if (arg_private_network) + loopback_setup(); + /* Mark everything as slave, so that we still * receive mounts from the real root, but don't * propagate mounts to the real root. */ @@ -3412,13 +4083,16 @@ int main(int argc, char *argv[]) { if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0) _exit(EXIT_FAILURE); - kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); + if (send_rtnl(rtnl_socket_pair[1]) < 0) + _exit(EXIT_FAILURE); + rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]); + /* Tell the parent that we are ready, and that * it can cgroupify us to that we lack access * to certain devices and resources. */ - (void) barrier_place(&barrier); + (void) barrier_place(&barrier); /* #1 */ if (setup_boot_id(arg_directory) < 0) _exit(EXIT_FAILURE); @@ -3443,7 +4117,7 @@ int main(int argc, char *argv[]) { /* Wait until we are cgroup-ified, so that we * can mount the right cgroup path writable */ - (void) barrier_sync_next(&barrier); + (void) barrier_place_and_sync(&barrier); /* #2 */ if (mount_cgroup(arg_directory) < 0) _exit(EXIT_FAILURE); @@ -3468,16 +4142,50 @@ int main(int argc, char *argv[]) { _exit(EXIT_FAILURE); } - umask(0022); + if (arg_userns) { + if (unshare(CLONE_NEWUSER) < 0) { + log_error_errno(errno, "unshare(CLONE_NEWUSER) failed: %m"); + _exit(EXIT_FAILURE); + } - if (arg_private_network) - loopback_setup(); + /* Tell the parent, that it now can + * write the UID map. */ + (void) barrier_place(&barrier); /* #3 */ + + /* Wait until the parent wrote the UID + * map */ + (void) barrier_place_and_sync(&barrier); /* #4 */ + } + + umask(0022); if (drop_capabilities() < 0) { log_error_errno(errno, "drop_capabilities() failed: %m"); _exit(EXIT_FAILURE); } + setup_hostname(); + + if (arg_personality != 0xffffffffLU) { + if (personality(arg_personality) < 0) { + log_error_errno(errno, "personality() failed: %m"); + _exit(EXIT_FAILURE); + } + } else if (secondary) { + if (personality(PER_LINUX32) < 0) { + log_error_errno(errno, "personality() failed: %m"); + _exit(EXIT_FAILURE); + } + } + +#ifdef HAVE_SELINUX + if (arg_selinux_context) + if (setexeccon((security_context_t) arg_selinux_context) < 0) { + log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context); + _exit(EXIT_FAILURE); + } +#endif + r = change_uid_gid(&home); if (r < 0) _exit(EXIT_FAILURE); @@ -3512,28 +4220,6 @@ int main(int argc, char *argv[]) { } } - setup_hostname(); - - if (arg_personality != 0xffffffffLU) { - if (personality(arg_personality) < 0) { - log_error_errno(errno, "personality() failed: %m"); - _exit(EXIT_FAILURE); - } - } else if (secondary) { - if (personality(PER_LINUX32) < 0) { - log_error_errno(errno, "personality() failed: %m"); - _exit(EXIT_FAILURE); - } - } - -#ifdef HAVE_SELINUX - if (arg_selinux_context) - if (setexeccon((security_context_t) arg_selinux_context) < 0) { - log_error_errno(errno, "setexeccon(\"%s\") failed: %m", arg_selinux_context); - _exit(EXIT_FAILURE); - } -#endif - if (!strv_isempty(arg_setenv)) { char **n; @@ -3547,9 +4233,10 @@ int main(int argc, char *argv[]) { } else env_use = (char**) envp; - /* Wait until the parent is ready with the setup, too... */ - if (!barrier_place_and_sync(&barrier)) - _exit(EXIT_FAILURE); + /* Let the parent know that we are ready and + * wait until the parent is ready with the + * setup, too... */ + (void) barrier_place_and_sync(&barrier); /* #5 */ if (arg_boot) { char **a; @@ -3585,13 +4272,15 @@ int main(int argc, char *argv[]) { fdset_free(fds); fds = NULL; + kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); + rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[1]); + + (void) barrier_place(&barrier); /* #1 */ + /* Wait for the most basic Child-setup to be done, * before we add hardware to it, and place it in a * cgroup. */ - if (barrier_sync_next(&barrier)) { - _cleanup_event_unref_ sd_event *event = NULL; - _cleanup_(pty_forward_freep) PTYForward *forward = NULL; - char last_char = 0; + if (barrier_sync(&barrier)) { /* #1 */ int ifi = 0; r = move_network_interfaces(pid); @@ -3610,10 +4299,43 @@ int main(int argc, char *argv[]) { if (r < 0) goto finish; + r = setup_ipvlan(pid); + if (r < 0) + goto finish; + r = register_machine(pid, ifi); if (r < 0) goto finish; + /* 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. */ + (void) barrier_place(&barrier); /* #2 */ + + if (arg_userns) { + char uid_map[strlen("/proc//uid_map") + DECIMAL_STR_MAX(uid_t) + 1], line[DECIMAL_STR_MAX(uid_t)*3+3+1]; + + (void) barrier_place_and_sync(&barrier); /* #3 */ + + xsprintf(uid_map, "/proc/" PID_FMT "/uid_map", pid); + xsprintf(line, UID_FMT " " UID_FMT " " UID_FMT "\n", 0, arg_uid_shift, arg_uid_range); + r = write_string_file(uid_map, line); + if (r < 0) { + log_error_errno(r, "Failed to write UID map: %m"); + goto finish; + } + + /* We always assign the same UID and GID ranges */ + xsprintf(uid_map, "/proc/" PID_FMT "/gid_map", pid); + r = write_string_file(uid_map, line); + if (r < 0) { + log_error_errno(r, "Failed to write GID map: %m"); + goto finish; + } + + (void) barrier_place(&barrier); /* #4 */ + } + /* Block SIGCHLD here, before notifying child. * process_pty() will handle it with the other signals. */ r = sigprocmask(SIG_BLOCK, &mask_chld, NULL); @@ -3625,58 +4347,69 @@ int main(int argc, char *argv[]) { if (r < 0) goto finish; - /* 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. */ - (void) barrier_place(&barrier); + /* Let the child know that we are ready and wait that the child is completely ready now. */ + if (barrier_place_and_sync(&barrier)) { /* #5 */ + _cleanup_event_unref_ sd_event *event = NULL; + _cleanup_(pty_forward_freep) PTYForward *forward = NULL; + _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL; + char last_char = 0; - /* And wait that the child is completely ready now. */ - (void) barrier_place_and_sync(&barrier); + sd_notifyf(false, + "READY=1\n" + "STATUS=Container running.\n" + "X_NSPAWN_LEADER_PID=" PID_FMT, pid); - sd_notify(false, - "READY=1\n" - "STATUS=Container running."); + r = sd_event_new(&event); + if (r < 0) { + log_error_errno(r, "Failed to get default event source: %m"); + goto finish; + } - r = sd_event_new(&event); - if (r < 0) { - log_error_errno(r, "Failed to get default event source: %m"); - goto finish; - } + if (arg_boot) { + /* Try to kill the init system on SIGINT or SIGTERM */ + sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, UINT32_TO_PTR(pid)); + sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, UINT32_TO_PTR(pid)); + } else { + /* Immediately exit */ + sd_event_add_signal(event, NULL, SIGINT, NULL, NULL); + sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL); + } - if (arg_boot) { - /* Try to kill the init system on SIGINT or SIGTERM */ - sd_event_add_signal(event, NULL, SIGINT, on_orderly_shutdown, UINT32_TO_PTR(pid)); - sd_event_add_signal(event, NULL, SIGTERM, on_orderly_shutdown, UINT32_TO_PTR(pid)); - } else { - /* Immediately exit */ - sd_event_add_signal(event, NULL, SIGINT, NULL, NULL); - sd_event_add_signal(event, NULL, SIGTERM, NULL, NULL); - } + /* simply exit on sigchld */ + sd_event_add_signal(event, NULL, SIGCHLD, NULL, NULL); - /* simply exit on sigchld */ - sd_event_add_signal(event, NULL, SIGCHLD, NULL, NULL); + if (arg_expose_ports) { + r = watch_rtnl(event, rtnl_socket_pair[0], &exposed, &rtnl); + if (r < 0) + goto finish; - r = pty_forward_new(event, master, true, &forward); - if (r < 0) { - log_error_errno(r, "Failed to create PTY forwarder: %m"); - goto finish; - } + (void) expose_ports(rtnl, &exposed); + } - r = sd_event_loop(event); - if (r < 0) { - log_error_errno(r, "Failed to run event loop: %m"); - goto finish; - } + rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]); - pty_forward_get_last_char(forward, &last_char); + r = pty_forward_new(event, master, true, !interactive, &forward); + if (r < 0) { + log_error_errno(r, "Failed to create PTY forwarder: %m"); + goto finish; + } - forward = pty_forward_free(forward); + r = sd_event_loop(event); + if (r < 0) { + log_error_errno(r, "Failed to run event loop: %m"); + goto finish; + } + + pty_forward_get_last_char(forward, &last_char); - if (!arg_quiet && last_char != '\n') - putc('\n', stdout); + forward = pty_forward_free(forward); - /* Kill if it is not dead yet anyway */ - terminate_machine(pid); + if (!arg_quiet && last_char != '\n') + putc('\n', stdout); + + /* Kill if it is not dead yet anyway */ + terminate_machine(pid); + } } /* Normally redundant, but better safe than sorry */ @@ -3714,6 +4447,8 @@ int main(int argc, char *argv[]) { r = 0; break; } + + flush_ports(&exposed); } finish: @@ -3737,7 +4472,7 @@ finish: if (arg_machine) { const char *p; - p = strappenda("/run/systemd/nspawn/propagate", arg_machine); + p = strjoina("/run/systemd/nspawn/propagate/", arg_machine); (void) rm_rf(p, false, true, false); } @@ -3749,9 +4484,18 @@ finish: strv_free(arg_setenv); strv_free(arg_network_interfaces); strv_free(arg_network_macvlan); + strv_free(arg_network_ipvlan); strv_free(arg_bind); strv_free(arg_bind_ro); strv_free(arg_tmpfs); + flush_ports(&exposed); + + while (arg_expose_ports) { + ExposePort *p = arg_expose_ports; + LIST_REMOVE(ports, arg_expose_ports, p); + free(p); + } + return r < 0 ? EXIT_FAILURE : ret; }