X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=83be00231c20604d06f7b06097c789fb25375130;hb=9444b1f;hp=f57c75ffeeb8f11935086cd398ef34a750705e49;hpb=7fd1b19bc9e9f5574f2877936b8ac267c7706947;p=elogind.git diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index f57c75ffe..83be00231 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -39,10 +39,12 @@ #include #include #include +#include #include #include #include +#include #include "log.h" #include "util.h" @@ -59,6 +61,8 @@ #include "fdset.h" #include "build.h" #include "fileio.h" +#include "bus-internal.h" +#include "bus-message.h" #ifndef TTY_GID #define TTY_GID 5 @@ -73,9 +77,9 @@ typedef enum LinkJournal { static char *arg_directory = NULL; static char *arg_user = NULL; -static char **arg_controllers = NULL; -static char *arg_uuid = NULL; +static sd_id128_t arg_uuid = {}; static char *arg_machine = NULL; +static const char *arg_slice = NULL; static bool arg_private_network = false; static bool arg_read_only = false; static bool arg_boot = false; @@ -118,10 +122,9 @@ static int help(void) { " -D --directory=NAME Root directory for the container\n" " -b --boot Boot up full system (i.e. invoke init)\n" " -u --user=USER Run the command under specified user or uid\n" - " -C --controllers=LIST Put the container in specified comma-separated\n" - " cgroup hierarchies\n" " --uuid=UUID Set a specific machine UUID for the container\n" " -M --machine=NAME Set the machine name for the container\n" + " -S --slice=SLICE Place the container in the specified slice\n" " --private-network Disable network in container\n" " --read-only Mount the root directory read-only\n" " --capability=CAP In addition to the default, retain specified\n" @@ -154,7 +157,6 @@ static int parse_argv(int argc, char *argv[]) { { "version", no_argument, NULL, ARG_VERSION }, { "directory", required_argument, NULL, 'D' }, { "user", required_argument, NULL, 'u' }, - { "controllers", required_argument, NULL, 'C' }, { "private-network", no_argument, NULL, ARG_PRIVATE_NETWORK }, { "boot", no_argument, NULL, 'b' }, { "uuid", required_argument, NULL, ARG_UUID }, @@ -164,15 +166,16 @@ static int parse_argv(int argc, char *argv[]) { { "bind", required_argument, NULL, ARG_BIND }, { "bind-ro", required_argument, NULL, ARG_BIND_RO }, { "machine", required_argument, NULL, 'M' }, + { "slice", required_argument, NULL, 'S' }, { NULL, 0, NULL, 0 } }; - int c; + int c, r; assert(argc >= 0); assert(argv); - while ((c = getopt_long(argc, argv, "+hD:u:C:bj", options, NULL)) >= 0) { + while ((c = getopt_long(argc, argv, "+hD:u:C:bM:jS:", options, NULL)) >= 0) { switch (c) { @@ -203,15 +206,6 @@ static int parse_argv(int argc, char *argv[]) { break; - case 'C': - strv_free(arg_controllers); - arg_controllers = strv_split(optarg, ","); - if (!arg_controllers) - return log_oom(); - - cg_shorten_controllers(arg_controllers); - break; - case ARG_PRIVATE_NETWORK: arg_private_network = true; break; @@ -221,7 +215,15 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_UUID: - arg_uuid = optarg; + r = sd_id128_from_string(optarg, &arg_uuid); + if (r < 0) { + log_error("Invalid UUID: %s", optarg); + return r; + } + break; + + case 'S': + arg_slice = strdup(optarg); break; case 'M': @@ -291,7 +293,6 @@ static int parse_argv(int argc, char *argv[]) { _cleanup_free_ char *a = NULL, *b = NULL; char *e; char ***x; - int r; x = c == ARG_BIND ? &arg_bind : &arg_bind_ro; @@ -492,7 +493,8 @@ static int setup_timezone(const char *dest) { } static int setup_resolv_conf(const char *dest) { - char *where; + char _cleanup_free_ *where = NULL; + _cleanup_close_ int fd = -1; assert(dest); @@ -504,12 +506,18 @@ static int setup_resolv_conf(const char *dest) { if (!where) return log_oom(); + fd = open(where, O_WRONLY|O_CREAT|O_EXCL|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW, 0644); + /* We don't really care for the results of this really. If it * fails, it fails, but meh... */ - if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) >= 0) - mount("/etc/resolv.conf", where, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); - - free(where); + if (mount("/etc/resolv.conf", where, "bind", MS_BIND, NULL) < 0) + log_warning("Failed to bind mount /etc/resolv.conf: %m"); + else + if (mount("/etc/resolv.conf", where, "bind", + MS_BIND|MS_REMOUNT|MS_RDONLY, NULL) < 0) { + log_error("Failed to remount /etc/resolv.conf readonly: %m"); + return -errno; + } return 0; } @@ -895,25 +903,6 @@ static int setup_journal(const char *directory) { return 0; } -static int setup_cgroup(const char *path) { - char **c; - int r; - - r = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, path, 1); - if (r < 0) { - log_error("Failed to create cgroup: %s", strerror(-r)); - return r; - } - - STRV_FOREACH(c, arg_controllers) { - r = cg_create_and_attach(*c, path, 1); - if (r < 0) - log_warning("Failed to create cgroup in controller %s: %s", *c, strerror(-r)); - } - - return 0; -} - static int drop_capabilities(void) { return capability_bounding_set_drop(~arg_retain, false); } @@ -1160,10 +1149,55 @@ finish: return r; } +static int register_machine(void) { + _cleanup_bus_error_free_ sd_bus_error error = SD_BUS_ERROR_NULL; + _cleanup_bus_unref_ sd_bus *bus = NULL; + int r; + + r = sd_bus_open_system(&bus); + if (r < 0) { + log_error("Failed to open system bus: %s", strerror(-r)); + return r; + } + + r = sd_bus_call_method( + bus, + "org.freedesktop.login1", + "/org/freedesktop/login1", + "org.freedesktop.login1.Manager", + "CreateMachine", + &error, + NULL, + "sayssuss", + arg_machine, + SD_BUS_APPEND_ID128(arg_uuid), + "nspawn", + "container", + (uint32_t) 0, + strempty(arg_slice), + strempty(arg_directory)); + if (r < 0) { + log_error("Failed to register machine: %s", error.message); + return r; + } + + return 0; +} + +static bool audit_enabled(void) { + int fd; + + fd = socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT); + if (fd >= 0) { + close_nointr_nofail(fd); + return true; + } + return false; +} + int main(int argc, char *argv[]) { pid_t pid = 0; int r = EXIT_FAILURE, k; - _cleanup_free_ char *machine_root = NULL, *newcg = NULL; _cleanup_close_ int master = -1; int n_fd_passed; const char *console = NULL; @@ -1177,9 +1211,13 @@ int main(int argc, char *argv[]) { log_parse_environment(); log_open(); - r = parse_argv(argc, argv); - if (r <= 0) + k = parse_argv(argc, argv); + if (k < 0) goto finish; + else if (k == 0) { + r = EXIT_SUCCESS; + goto finish; + } if (arg_directory) { char *p; @@ -1191,7 +1229,7 @@ int main(int argc, char *argv[]) { arg_directory = get_current_dir_name(); if (!arg_directory) { - log_error("Failed to determine path"); + log_error("Failed to determine path, please use -D."); goto finish; } @@ -1204,7 +1242,7 @@ int main(int argc, char *argv[]) { goto finish; } - hostname_cleanup(arg_machine); + hostname_cleanup(arg_machine, false); if (isempty(arg_machine)) { log_error("Failed to determine machine name automatically, please use -M."); goto finish; @@ -1221,13 +1259,20 @@ int main(int argc, char *argv[]) { goto finish; } + if (arg_boot && audit_enabled()) { + log_warning("The kernel auditing subsystem is known to be incompatible with containers.\n" + "Please make sure to turn off auditing with 'audit=0' on the kernel command\n" + "line before using systemd-nspawn. Sleeping for 5s...\n"); + sleep(5); + } + if (path_equal(arg_directory, "/")) { log_error("Spawning container on root directory not supported."); goto finish; } if (path_is_os_tree(arg_directory) <= 0) { - log_error("Directory %s doesn't look like an OS root directory. Refusing.", arg_directory); + log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory); goto finish; } @@ -1243,28 +1288,6 @@ int main(int argc, char *argv[]) { fdset_close_others(fds); log_open(); - k = cg_get_machine_path(&machine_root); - if (k < 0) { - log_error("Failed to determine machine cgroup path: %s", strerror(-k)); - goto finish; - } - - newcg = strjoin(machine_root, "/", arg_machine, NULL); - if (!newcg) { - log_error("Failed to allocate cgroup path."); - goto finish; - } - - r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, newcg, false); - if (r <= 0 && r != -ENOENT) { - log_error("Container already running."); - - free(newcg); - newcg = NULL; - - goto finish; - } - master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); if (master < 0) { log_error("Failed to acquire pseudo tty: %m"); @@ -1300,19 +1323,27 @@ int main(int argc, char *argv[]) { goto finish; } + sd_notify(0, "READY=1"); + assert_se(sigemptyset(&mask) == 0); sigset_add_many(&mask, SIGCHLD, SIGWINCH, SIGTERM, SIGINT, -1); assert_se(sigprocmask(SIG_BLOCK, &mask, NULL) == 0); for (;;) { siginfo_t status; - int pipefd[2]; + int pipefd[2], pipefd2[2]; if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) { log_error("pipe2(): %m"); goto finish; } + if (pipe2(pipefd2, O_NONBLOCK|O_CLOEXEC) < 0) { + log_error("pipe2(): %m"); + close_pipe(pipefd); + goto finish; + } + pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL); if (pid < 0) { if (errno == EINVAL) @@ -1346,6 +1377,7 @@ int main(int argc, char *argv[]) { if (envp[n_env]) n_env ++; + /* Wait for the parent process to log our PID */ close_nointr_nofail(pipefd[1]); fd_wait_for_event(pipefd[0], POLLHUP, -1); close_nointr_nofail(pipefd[0]); @@ -1399,8 +1431,11 @@ int main(int argc, char *argv[]) { goto child_fail; } - if (setup_cgroup(newcg) < 0) - goto child_fail; + close_pipe(pipefd2); + + r = register_machine(); + if (r < 0) + goto finish; /* Mark everything as slave, so that we still * receive mounts from the real root, but don't @@ -1552,8 +1587,8 @@ int main(int argc, char *argv[]) { goto child_fail; } - if (arg_uuid) { - if (asprintf((char**)(envp + n_env++), "container_uuid=%s", arg_uuid) < 0) { + 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) { log_oom(); goto child_fail; } @@ -1610,6 +1645,11 @@ int main(int argc, char *argv[]) { close_nointr_nofail(pipefd[0]); close_nointr_nofail(pipefd[1]); + /* Wait for the child process to establish cgroup hierarchy */ + close_nointr_nofail(pipefd2[1]); + fd_wait_for_event(pipefd2[0], POLLHUP, -1); + close_nointr_nofail(pipefd2[0]); + fdset_free(fds); fds = NULL; @@ -1619,16 +1659,16 @@ int main(int argc, char *argv[]) { if (saved_attr_valid) tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); - r = wait_for_terminate(pid, &status); - if (r < 0) { + k = wait_for_terminate(pid, &status); + if (k < 0) { r = EXIT_FAILURE; break; } if (status.si_code == CLD_EXITED) { + r = status.si_status; if (status.si_status != 0) { log_error("Container failed with error code %i.", status.si_status); - r = status.si_status; break; } @@ -1662,12 +1702,11 @@ finish: close_pipe(kmsg_socket_pair); - if (newcg) - cg_kill_recursive_and_wait(SYSTEMD_CGROUP_CONTROLLER, newcg, true); + if (pid > 0) + kill(pid, SIGKILL); free(arg_directory); free(arg_machine); - strv_free(arg_controllers); fdset_free(fds);