X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=9fe040c81b6342005e89d72ef98eabdf0edc8c8c;hp=a13c1fcd246dad7ba3f141d6f3069fd3154618bd;hb=ada4799ac5ad2e6c0fe11dc5c096faca8a85876b;hpb=60e1651a31c9c0ed1caef1a63f5e3a87156b0b1e diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index a13c1fcd2..9fe040c81 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -31,7 +31,6 @@ #include #include #include -#include #include #include #include @@ -90,13 +89,26 @@ #include "base-filesystem.h" #include "barrier.h" #include "event-util.h" +#include "capability.h" #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 @@ -171,6 +183,7 @@ 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 void help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" @@ -196,12 +209,14 @@ 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" + " -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" + " -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" @@ -224,8 +239,8 @@ 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) { @@ -268,7 +283,6 @@ static int parse_argv(int argc, char *argv[]) { ARG_KEEP_UNIT, ARG_NETWORK_INTERFACE, ARG_NETWORK_MACVLAN, - ARG_NETWORK_VETH, ARG_NETWORK_BRIDGE, ARG_PERSONALITY, ARG_VOLATILE, @@ -303,11 +317,12 @@ 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-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' }, {} }; @@ -317,7 +332,7 @@ 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) { @@ -368,7 +383,7 @@ 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; @@ -642,6 +657,65 @@ 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 '?': return -EINVAL; @@ -697,6 +771,11 @@ 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; @@ -868,6 +947,112 @@ static int mount_binds(const char *dest, char **l, bool ro) { return 0; } +static int mount_cgroup_hierarchy(const char *dest, const char *controller, const char *hierarchy, bool read_only) { + char *to; + int r; + + to = strappenda(dest, "/sys/fs/cgroup/", hierarchy); + + r = path_is_mount_point(to, false); + if (r < 0) + return log_error_errno(r, "Failed to determine if %s is mounted already: %m", to); + if (r > 0) + return 0; + + mkdir_p(to, 0755); + + if (mount("cgroup", to, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV|(read_only ? MS_RDONLY : 0), controller) < 0) + return log_error_errno(errno, "Failed to mount to %s: %m", to); + + return 1; +} + +static int mount_cgroup(const char *dest) { + _cleanup_set_free_free_ Set *controllers = NULL; + _cleanup_free_ char *own_cgroup_path = NULL; + const char *cgroup_root, *systemd_root, *systemd_own; + int r; + + controllers = set_new(&string_hash_ops); + if (!controllers) + return log_oom(); + + r = cg_kernel_controllers(controllers); + if (r < 0) + return log_error_errno(r, "Failed to determine cgroup controllers: %m"); + + r = cg_pid_get_path(NULL, 0, &own_cgroup_path); + if (r < 0) + return log_error_errno(r, "Failed to determine our own cgroup path: %m"); + + cgroup_root = strappenda(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"); + + for (;;) { + _cleanup_free_ char *controller = NULL, *origin = NULL, *combined = NULL; + + controller = set_steal_first(controllers); + if (!controller) + break; + + origin = strappend("/sys/fs/cgroup/", controller); + if (!origin) + return log_oom(); + + r = readlink_malloc(origin, &combined); + if (r == -EINVAL) { + /* Not a symbolic link, but directly a single cgroup hierarchy */ + + r = mount_cgroup_hierarchy(dest, controller, controller, true); + if (r < 0) + return r; + + } else if (r < 0) + return log_error_errno(r, "Failed to read link %s: %m", origin); + else { + _cleanup_free_ char *target = NULL; + + target = strjoin(dest, "/sys/fs/cgroup/", controller, NULL); + if (!target) + return log_oom(); + + /* A symbolic link, a combination of controllers in one hierarchy */ + + if (!filename_is_valid(combined)) { + log_warning("Ignoring invalid combined hierarchy %s.", combined); + continue; + } + + r = mount_cgroup_hierarchy(dest, combined, combined, true); + if (r < 0) + return r; + + if (symlink(combined, target) < 0) + return log_error_errno(errno, "Failed to create symlink for combined hiearchy: %m"); + } + } + + r = mount_cgroup_hierarchy(dest, "name=systemd", "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); + 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"); + 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"); + + if (mount(NULL, cgroup_root, NULL, MS_REMOUNT|MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME|MS_RDONLY, "mode=755") < 0) + return log_error_errno(errno, "Failed to remount %s read-only: %m", cgroup_root); + + return 0; +} + static int mount_tmpfs(const char *dest) { char **i, **o; @@ -985,7 +1170,7 @@ 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); @@ -1184,7 +1369,7 @@ 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", dest); + return log_error_errno(errno, "mknod(%s) failed: %m", to); } } @@ -1242,8 +1427,8 @@ 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))]; @@ -1294,7 +1479,7 @@ 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) @@ -1305,6 +1490,198 @@ 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) @@ -2082,6 +2459,27 @@ finish: } +static int setup_propagate(const char *root) { + const char *p, *q; + + (void) mkdir_p("/run/systemd/nspawn/", 0755); + (void) mkdir_p("/run/systemd/nspawn/propagate", 0600); + p = strappenda("/run/systemd/nspawn/propagate/", arg_machine); + (void) mkdir_p(p, 0600); + + q = strappenda(root, "/run/systemd/nspawn/incoming"); + mkdir_parents(q, 0755); + mkdir_p(q, 0600); + + if (mount(p, q, NULL, MS_BIND, NULL) < 0) + return log_error_errno(errno, "Failed to install propagation bind mount."); + + if (mount(NULL, q, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY, NULL) < 0) + return log_error_errno(errno, "Failed to make propagation mount read-only"); + + return 0; +} + static int setup_image(char **device_path, int *loop_nr) { struct loop_info64 info = { .lo_flags = LO_FLAGS_AUTOCLEAR|LO_FLAGS_PARTSCAN @@ -2159,6 +2557,12 @@ 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 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, @@ -2186,6 +2590,7 @@ static int dissect_image( blkid_partlist pl; struct stat st; int r; + bool is_gpt, is_mbr; assert(fd >= 0); assert(root_device); @@ -2214,8 +2619,9 @@ 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) @@ -2225,9 +2631,14 @@ 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; } @@ -2267,9 +2678,8 @@ static int dissect_image( 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; @@ -2300,81 +2710,117 @@ static int dissect_image( continue; flags = blkid_partition_get_flags(pp); - if (flags & GPT_FLAG_NO_AUTO) + if (is_gpt && (flags & GPT_FLAG_NO_AUTO)) + continue; + if (is_mbr && (flags != 0x80)) /* Bootable flag */ continue; nr = blkid_partition_get_partno(pp); if (nr < 0) continue; - stype = blkid_partition_get_type_string(pp); - if (!stype) - continue; - - if (sd_id128_from_string(stype, &type_id) < 0) - continue; + if (is_gpt) { + sd_id128_t type_id; + const char *stype; - 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); + + r = free_and_strdup(&home, node); + if (r < 0) + return log_oom(); - srv_nr = nr; - srv_rw = !(flags & GPT_FLAG_READ_ONLY); + } else if (sd_id128_equal(type_id, GPT_SRV)) { - free(srv); - srv = strdup(node); - if (!srv) - return log_oom(); - } + 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; - if (secondary_root && nr >= secondary_root_nr) + 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 (is_mbr) { + int type; + + type = blkid_partition_get_type(pp); + if (type != 0x83) /* Linux partition */ continue; - secondary_root_nr = nr; - secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY); + /* Note that there's a certain, intended + * asymmetry here: while for GPT we simply + * take the first valid partition and ignore + * all others of the same type, for MBR we + * fail if there are multiple suitable + * partitions. This is because the GPT + * partition types are defined by us, and + * hence we can define their lookup semantics, + * while for the MBR logic we reuse existing + * definitions, and simply don't want to make + * out the situation. */ + + if (root) { + log_error("Identified multiple bootable Linux 0x83 partitions on\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); + return -EINVAL; + } + root_nr = nr; - free(secondary_root); - secondary_root = strdup(node); - if (!secondary_root) + 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); + log_error("Failed to identify root partition in disk image\n" + " %s\n" + PARTITION_TABLE_BLURB, arg_image); return -EINVAL; } @@ -2861,15 +3307,33 @@ static int on_orderly_shutdown(sd_event_source *s, const struct signalfd_siginfo } static int determine_names(void) { + int r; if (!arg_image && !arg_directory) { - if (arg_machine) - arg_directory = strappend("/var/lib/container/", arg_machine); - else + if (arg_machine) { + _cleanup_(image_unrefp) Image *i = NULL; + + r = image_find(arg_machine, &i); + if (r < 0) + return log_error_errno(r, "Failed to find image for machine '%s': %m", arg_machine); + else if (r == 0) { + log_error("No image for machine '%s': %m", arg_machine); + return -ENOENT; + } + + if (i->type == IMAGE_GPT) + r = set_sanitized_path(&arg_image, i->path); + else + r = set_sanitized_path(&arg_directory, i->path); + if (r < 0) + return log_error_errno(r, "Invalid image directory: %m"); + + arg_read_only = arg_read_only || i->read_only; + } else arg_directory = get_current_dir_name(); - if (!arg_directory) { - log_error("Failed to determine path, please use -D."); + if (!arg_directory && !arg_machine) { + log_error("Failed to determine path, please use -D or -i."); return -EINVAL; } } @@ -2910,18 +3374,18 @@ static int determine_names(void) { int main(int argc, char *argv[]) { - _cleanup_free_ char *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL; + _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; - const char *console = NULL; char veth_name[IFNAMSIZ]; bool secondary = false, remove_subvol = false; 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; log_parse_environment(); log_open(); @@ -2967,20 +3431,8 @@ 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) { + _cleanup_release_lock_file_ LockFile original_lock = LOCK_FILE_INIT; char *np; /* If the specified path is a mount point we @@ -3003,6 +3455,12 @@ 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); @@ -3014,6 +3472,31 @@ 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 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); + } + } } if (arg_boot) { @@ -3040,6 +3523,16 @@ 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; @@ -3073,9 +3566,9 @@ int main(int argc, char *argv[]) { goto finish; } - console = ptsname(master); - if (!console) { - r = log_error_errno(errno, "Failed to determine tty name: %m"); + r = ptsname_malloc(master, &console); + if (r < 0) { + r = log_error_errno(r, "Failed to determine tty name: %m"); goto finish; } @@ -3088,15 +3581,6 @@ int main(int argc, char *argv[]) { 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; - } - - sd_notify(false, - "READY=1\n" - "STATUS=Container running."); - assert_se(sigemptyset(&mask) == 0); sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); @@ -3105,6 +3589,7 @@ 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 = { @@ -3118,6 +3603,16 @@ 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. */ @@ -3176,6 +3671,7 @@ int main(int argc, char *argv[]) { 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(); @@ -3260,6 +3756,9 @@ int main(int argc, char *argv[]) { dev_setup(arg_directory); + if (setup_propagate(arg_directory) < 0) + _exit(EXIT_FAILURE); + if (setup_seccomp() < 0) _exit(EXIT_FAILURE); @@ -3268,9 +3767,17 @@ 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); + if (setup_boot_id(arg_directory) < 0) _exit(EXIT_FAILURE); @@ -3292,10 +3799,12 @@ int main(int argc, char *argv[]) { if (mount_tmpfs(arg_directory) < 0) _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. */ - (void)barrier_place(&barrier); + /* Wait until we are cgroup-ified, so that we + * can mount the right cgroup path writable */ + (void) barrier_sync_next(&barrier); + + if (mount_cgroup(arg_directory) < 0) + _exit(EXIT_FAILURE); if (chdir(arg_directory) < 0) { log_error_errno(errno, "chdir(%s) failed: %m", arg_directory); @@ -3434,10 +3943,13 @@ int main(int argc, char *argv[]) { fdset_free(fds); fds = NULL; - /* wait for child-setup to be done */ - if (barrier_place_and_sync(&barrier)) { - _cleanup_event_unref_ sd_event *event = NULL; - _cleanup_(pty_forward_freep) PTYForward *forward = NULL; + kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]); + rtnl_socket_pair[1] = safe_close(rtnl_socket_pair[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)) { int ifi = 0; r = move_network_interfaces(pid); @@ -3474,46 +3986,71 @@ int main(int argc, char *argv[]) { /* 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); + (void) barrier_place(&barrier); - r = sd_event_new(&event); - if (r < 0) { - log_error_errno(r, "Failed to get default event source: %m"); - goto finish; - } + /* And wait that the child is completely ready now. */ + if (barrier_place_and_sync(&barrier)) { + _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; - 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); - } + sd_notifyf(false, + "READY=1\n" + "STATUS=Container running.\n" + "X_NSPAWN_LEADER_PID=" PID_FMT, pid); - /* simply exit on sigchld */ - sd_event_add_signal(event, NULL, SIGCHLD, NULL, NULL); + r = sd_event_new(&event); + if (r < 0) { + log_error_errno(r, "Failed to get default event source: %m"); + goto finish; + } - r = pty_forward_new(event, master, &forward); - if (r < 0) { - log_error_errno(r, "Failed to create PTY forwarder: %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); + } - r = sd_event_loop(event); - if (r < 0) { - log_error_errno(r, "Failed to run event loop: %m"); - goto finish; - } + /* 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; + + (void) expose_ports(rtnl, &exposed); + } + + rtnl_socket_pair[0] = safe_close(rtnl_socket_pair[0]); + + r = pty_forward_new(event, master, true, &forward); + if (r < 0) { + log_error_errno(r, "Failed to create PTY forwarder: %m"); + goto finish; + } + + 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); - forward = pty_forward_free(forward); + forward = pty_forward_free(forward); - if (!arg_quiet) - putc('\n', stdout); + if (!arg_quiet && last_char != '\n') + putc('\n', stdout); - /* Kill if it is not dead yet anyway */ - terminate_machine(pid); + /* Kill if it is not dead yet anyway */ + terminate_machine(pid); + } } /* Normally redundant, but better safe than sorry */ @@ -3551,6 +4088,8 @@ int main(int argc, char *argv[]) { r = 0; break; } + + flush_ports(&exposed); } finish: @@ -3571,6 +4110,13 @@ finish: log_warning_errno(k, "Cannot remove subvolume '%s', ignoring: %m", arg_directory); } + if (arg_machine) { + const char *p; + + p = strappenda("/run/systemd/nspawn/propagate/", arg_machine); + (void) rm_rf(p, false, true, false); + } + free(arg_directory); free(arg_template); free(arg_image); @@ -3583,5 +4129,13 @@ finish: 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; }