X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=f3568ab40c3a9e861dd539d559a1896628d09209;hp=c25768245179cad1e2a1be819395d8ab4b65d7b5;hb=f8964235e69f58225dec378437b1789744cd22a9;hpb=7027ff61a34a12487712b382a061c654acc3a679 diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index c25768245..f3568ab40 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -22,6 +22,7 @@ #include #include #include +#include #include #include #include @@ -172,7 +173,7 @@ static int parse_argv(int argc, char *argv[]) { 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:j", options, NULL)) >= 0) { switch (c) { @@ -221,6 +222,11 @@ static int parse_argv(int argc, char *argv[]) { break; case ARG_UUID: + if (!id128_is_valid(optarg)) { + log_error("Invalid UUID: %s", optarg); + return -EINVAL; + } + arg_uuid = optarg; break; @@ -365,7 +371,7 @@ static int mount_all(const char *dest) { int r = 0; for (k = 0; k < ELEMENTSOF(mount_table); k++) { - char _cleanup_free_ *where = NULL; + _cleanup_free_ char *where = NULL; int t; where = strjoin(dest, "/", mount_table[k].where, NULL); @@ -492,7 +498,8 @@ 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); @@ -504,18 +511,24 @@ 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; } static int setup_boot_id(const char *dest) { - char _cleanup_free_ *from = NULL, *to = NULL; + _cleanup_free_ char *from = NULL, *to = NULL; sd_id128_t rnd; char as_uuid[37]; int r; @@ -569,7 +582,7 @@ static int copy_devnodes(const char *dest) { const char *d; int r = 0; - mode_t _cleanup_umask_ u; + _cleanup_umask_ mode_t u; assert(dest); @@ -577,7 +590,7 @@ static int copy_devnodes(const char *dest) { NULSTR_FOREACH(d, devnodes) { struct stat st; - char _cleanup_free_ *from = NULL, *to = NULL; + _cleanup_free_ char *from = NULL, *to = NULL; asprintf(&from, "/dev/%s", d); asprintf(&to, "%s/dev/%s", dest, d); @@ -633,9 +646,9 @@ static int setup_ptmx(const char *dest) { static int setup_dev_console(const char *dest, const char *console) { struct stat st; - char _cleanup_free_ *to = NULL; + _cleanup_free_ char *to = NULL; int r; - mode_t _cleanup_umask_ u; + _cleanup_umask_ mode_t u; assert(dest); assert(console); @@ -681,9 +694,9 @@ static int setup_dev_console(const char *dest, const char *console) { } static int setup_kmsg(const char *dest, int kmsg_socket) { - char _cleanup_free_ *from = NULL, *to = NULL; + _cleanup_free_ char *from = NULL, *to = NULL; int r, fd, k; - mode_t _cleanup_umask_ u; + _cleanup_umask_ mode_t u; union { struct cmsghdr cmsghdr; uint8_t buf[CMSG_SPACE(sizeof(int))]; @@ -765,7 +778,7 @@ static int setup_hostname(void) { static int setup_journal(const char *directory) { sd_id128_t machine_id; - char _cleanup_free_ *p = NULL, *b = NULL, *q = NULL, *d = NULL; + _cleanup_free_ char *p = NULL, *b = NULL, *q = NULL, *d = NULL; char *id; int r; @@ -914,22 +927,49 @@ static int setup_cgroup(const char *path) { return 0; } -static int drop_capabilities(void) { - return capability_bounding_set_drop(~arg_retain, false); -} +static int save_attributes(const char *cgroup, pid_t pid, const char *uuid, const char *directory) { + _cleanup_free_ char *path = NULL; + char buf[DECIMAL_STR_MAX(pid_t)]; + int r = 0, k; -static int is_os_tree(const char *path) { - int r; - char *p; - /* We use /bin/sh as flag file if something is an OS */ + assert(cgroup); + assert(pid >= 0); + assert(arg_directory); - if (asprintf(&p, "%s/bin/sh", path) < 0) - return -ENOMEM; +#ifdef HAVE_XATTR + assert_se(snprintf(buf, sizeof(buf), "%lu", (unsigned long) pid) < (int) sizeof(buf)); - r = access(p, F_OK); - free(p); + r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, cgroup, NULL, &path); + if (r < 0) { + log_error("Failed to get path: %s", strerror(-r)); + return r; + } + + r = setxattr(path, "trusted.init_pid", buf, strlen(buf), XATTR_CREATE); + if (r < 0) + log_warning("Failed to set %s attribute on %s: %m", "trusted.init_pid", path); - return r < 0 ? 0 : 1; + if (uuid) { + k = setxattr(path, "trusted.machine_id", uuid, strlen(uuid), XATTR_CREATE); + if (k < 0) { + log_warning("Failed to set %s attribute on %s: %m", "trusted.machine_id", path); + if (r == 0) + r = k; + } + } + + k = setxattr(path, "trusted.root_directory", directory, strlen(directory), XATTR_CREATE); + if (k < 0) { + log_warning("Failed to set %s attribute on %s: %m", "trusted.root_directory", path); + if (r == 0) + r = k; + } +#endif + return r; +} + +static int drop_capabilities(void) { + return capability_bounding_set_drop(~arg_retain, false); } static int process_pty(int master, pid_t pid, sigset_t *mask) { @@ -1177,7 +1217,7 @@ finish: int main(int argc, char *argv[]) { pid_t pid = 0; int r = EXIT_FAILURE, k; - _cleanup_free_ char *machine_root = NULL, *newcg = NULL; + _cleanup_free_ char *newcg = NULL; _cleanup_close_ int master = -1; int n_fd_passed; const char *console = NULL; @@ -1191,9 +1231,13 @@ 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; @@ -1205,7 +1249,7 @@ 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; } @@ -1240,8 +1284,8 @@ int main(int argc, char *argv[]) { goto finish; } - if (is_os_tree(arg_directory) <= 0) { - log_error("Directory %s doesn't look like an OS root directory. Refusing.", arg_directory); + if (path_is_os_tree(arg_directory) <= 0) { + log_error("Directory %s doesn't look like an OS root directory (/etc/os-release is missing). Refusing.", arg_directory); goto finish; } @@ -1257,20 +1301,14 @@ int main(int argc, char *argv[]) { fdset_close_others(fds); log_open(); - k = cg_get_machine_path(&machine_root); + k = cg_get_machine_path(arg_machine, &newcg); 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) { + k = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, newcg, true); + if (k <= 0 && k != -ENOENT) { log_error("Container already running."); free(newcg); @@ -1314,19 +1352,27 @@ 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) @@ -1360,6 +1406,7 @@ 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]); @@ -1416,6 +1463,8 @@ int main(int argc, char *argv[]) { if (setup_cgroup(newcg) < 0) goto child_fail; + close_pipe(pipefd2); + /* Mark everything as slave, so that we still * receive mounts from the real root, but don't * propagate mounts to the real root. */ @@ -1624,6 +1673,13 @@ 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]); + + save_attributes(newcg, pid, arg_uuid, arg_directory); + fdset_free(fds); fds = NULL; @@ -1633,16 +1689,16 @@ 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; }