X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=b91b0b8a91f4bccc9011028e20a101bb9506fbf7;hp=8b574214314a297ac7d5c68a764219c5354d21cc;hb=77b6e19458f37cfde127ec6aa9494c0ac45ad890;hpb=5674767ec2cf7d168fe9c30f78074231fbe3408c diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index 8b5742143..b91b0b8a9 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -42,6 +42,10 @@ #include #include +#ifdef HAVE_XATTR +#include +#endif + #include #include "log.h" @@ -60,6 +64,10 @@ #include "build.h" #include "fileio.h" +#ifndef TTY_GID +#define TTY_GID 5 +#endif + typedef enum LinkJournal { LINK_NO, LINK_AUTO, @@ -71,6 +79,7 @@ static char *arg_directory = NULL; static char *arg_user = NULL; static char **arg_controllers = NULL; static char *arg_uuid = NULL; +static char *arg_machine = NULL; static bool arg_private_network = false; static bool arg_read_only = false; static bool arg_boot = false; @@ -109,13 +118,14 @@ static int help(void) { printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n" "Spawn a minimal namespace container for debugging, testing and building.\n\n" " -h --help Show this help\n" - " --version Print version string\n" + " --version Print version string\n" " -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" " --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" @@ -157,6 +167,7 @@ static int parse_argv(int argc, char *argv[]) { { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, { "bind", required_argument, NULL, ARG_BIND }, { "bind-ro", required_argument, NULL, ARG_BIND_RO }, + { "machine", required_argument, NULL, 'M' }, { NULL, 0, NULL, 0 } }; @@ -165,7 +176,7 @@ 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) { @@ -190,22 +201,19 @@ static int parse_argv(int argc, char *argv[]) { case 'u': free(arg_user); - if (!(arg_user = strdup(optarg))) { - log_error("Failed to duplicate user name."); - return -ENOMEM; - } + arg_user = strdup(optarg); + if (!arg_user) + return log_oom(); break; case 'C': strv_free(arg_controllers); arg_controllers = strv_split(optarg, ","); - if (!arg_controllers) { - log_error("Failed to split controllers list."); - return -ENOMEM; - } - strv_uniq(arg_controllers); + if (!arg_controllers) + return log_oom(); + cg_shorten_controllers(arg_controllers); break; case ARG_PRIVATE_NETWORK: @@ -217,9 +225,27 @@ 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; + case 'M': + if (!hostname_is_valid(optarg)) { + log_error("Invalid machine name: %s", optarg); + return -EINVAL; + } + + free(arg_machine); + arg_machine = strdup(optarg); + if (!arg_machine) + return log_oom(); + + break; + case ARG_READ_ONLY: arg_read_only = true; break; @@ -335,7 +361,7 @@ static int mount_all(const char *dest) { { NULL, "/proc/sys", NULL, NULL, MS_BIND|MS_RDONLY|MS_REMOUNT, true }, /* Then, make it r/o */ { "sysfs", "/sys", "sysfs", NULL, MS_RDONLY|MS_NOSUID|MS_NOEXEC|MS_NODEV, true }, { "tmpfs", "/dev", "tmpfs", "mode=755", MS_NOSUID|MS_STRICTATIME, true }, - { "/dev/pts", "/dev/pts", NULL, NULL, MS_BIND, true }, + { "devpts", "/dev/pts", "devpts","newinstance,ptmxmode=0666,mode=620,gid=" STRINGIFY(TTY_GID), MS_NOSUID|MS_NOEXEC, true }, { "tmpfs", "/dev/shm", "tmpfs", "mode=1777", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, { "tmpfs", "/run", "tmpfs", "mode=755", MS_NOSUID|MS_NODEV|MS_STRICTATIME, true }, #ifdef HAVE_SELINUX @@ -348,7 +374,7 @@ 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); @@ -475,7 +501,8 @@ 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); @@ -487,18 +514,24 @@ 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; @@ -524,7 +557,7 @@ static int setup_boot_id(const char *dest) { SD_ID128_FORMAT_VAL(rnd)); char_array_0(as_uuid); - r = write_one_line_file(from, as_uuid); + r = write_string_file(from, as_uuid); if (r < 0) { log_error("Failed to write boot id: %s", strerror(-r)); return r; @@ -533,8 +566,8 @@ static int setup_boot_id(const char *dest) { if (mount(from, to, "bind", MS_BIND, NULL) < 0) { log_error("Failed to bind mount boot id: %m"); r = -errno; - } else - mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL); + } else if (mount(from, to, "bind", MS_BIND|MS_REMOUNT|MS_RDONLY, NULL)) + log_warning("Failed to make boot id read-only: %m"); unlink(from); return r; @@ -548,12 +581,11 @@ static int copy_devnodes(const char *dest) { "full\0" "random\0" "urandom\0" - "tty\0" - "ptmx\0"; + "tty\0"; const char *d; int r = 0; - mode_t _cleanup_umask_ u; + _cleanup_umask_ mode_t u; assert(dest); @@ -561,7 +593,7 @@ 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); @@ -600,11 +632,26 @@ static int copy_devnodes(const char *dest) { return r; } +static int setup_ptmx(const char *dest) { + _cleanup_free_ char *p = NULL; + + p = strappend(dest, "/dev/ptmx"); + if (!p) + return log_oom(); + + if (symlink("pts/ptmx", p) < 0) { + log_error("Failed to create /dev/ptmx symlink: %m"); + return -errno; + } + + return 0; +} + 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); @@ -650,14 +697,17 @@ 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))]; - } control; - struct msghdr mh; + } control = {}; + struct msghdr mh = { + .msg_control = &control, + .msg_controllen = sizeof(control), + }; struct cmsghdr *cmsg; assert(dest); @@ -698,12 +748,6 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { return -errno; } - zero(mh); - zero(control); - - mh.msg_control = &control; - mh.msg_controllen = sizeof(control); - cmsg = CMSG_FIRSTHDR(&mh); cmsg->cmsg_level = SOL_SOCKET; cmsg->cmsg_type = SCM_RIGHTS; @@ -728,30 +772,16 @@ static int setup_kmsg(const char *dest, int kmsg_socket) { } static int setup_hostname(void) { - char *hn; - int r = 0; - - hn = path_get_file_name(arg_directory); - if (hn) { - hn = strdup(hn); - if (!hn) - return -ENOMEM; - - hostname_cleanup(hn); - if (!isempty(hn)) - if (sethostname(hn, strlen(hn)) < 0) - r = -errno; - - free(hn); - } + if (sethostname(arg_machine, strlen(arg_machine)) < 0) + return -errno; - return r; + return 0; } 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; @@ -881,22 +911,70 @@ static int setup_journal(const char *directory) { return 0; } -static int drop_capabilities(void) { - return capability_bounding_set_drop(~arg_retain, false); +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 is_os_tree(const char *path) { - int r; - char *p; - /* We use /bin/sh as flag file if something is an OS */ +static int save_attributes(const char *cgroup, pid_t pid, const char *uuid, const char *directory) { +#ifdef HAVE_XATTR + _cleanup_free_ char *path = NULL; + char buf[DECIMAL_STR_MAX(pid_t)]; + int r = 0, k; - if (asprintf(&p, "%s/bin/sh", path) < 0) - return -ENOMEM; + assert(cgroup); + assert(pid >= 0); + assert(arg_directory); - r = access(p, F_OK); - free(p); + assert_se(snprintf(buf, sizeof(buf), "%lu", (unsigned long) pid) < (int) sizeof(buf)); + + 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); + + 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; + } + return r; +#else + return 0; +#endif +} - return r < 0 ? 0 : 1; +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) { @@ -1141,12 +1219,24 @@ finish: return r; } +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; - char *oldcg = NULL, *newcg = NULL; - char **controller = NULL; - int master = -1, n_fd_passed; + _cleanup_free_ char *newcg = NULL; + _cleanup_close_ int master = -1; + int n_fd_passed; const char *console = NULL; struct termios saved_attr, raw_attr; sigset_t mask; @@ -1158,9 +1248,13 @@ 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; @@ -1172,12 +1266,26 @@ 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; } path_kill_slashes(arg_directory); + if (!arg_machine) { + arg_machine = strdup(path_get_file_name(arg_directory)); + if (!arg_machine) { + log_oom(); + goto finish; + } + + hostname_cleanup(arg_machine, false); + if (isempty(arg_machine)) { + log_error("Failed to determine machine name automatically, please use -M."); + goto finish; + } + } + if (geteuid() != 0) { log_error("Need to be root."); goto finish; @@ -1188,13 +1296,20 @@ int main(int argc, char *argv[]) { goto finish; } + if (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 (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; } @@ -1210,27 +1325,20 @@ int main(int argc, char *argv[]) { fdset_close_others(fds); log_open(); - k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg); + k = cg_get_machine_path(arg_machine, &newcg); if (k < 0) { - log_error("Failed to determine current cgroup: %s", strerror(-k)); + log_error("Failed to determine machine cgroup path: %s", strerror(-k)); goto finish; } - if (asprintf(&newcg, "%s/nspawn-%lu", oldcg, (unsigned long) getpid()) < 0) { - log_error("Failed to allocate cgroup path."); - goto finish; - } + k = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, newcg, true); + if (k <= 0 && k != -ENOENT) { + log_error("Container already running."); - k = cg_create_and_attach(SYSTEMD_CGROUP_CONTROLLER, newcg, 0); - if (k < 0) { - log_error("Failed to create cgroup: %s", strerror(-k)); - goto finish; - } + free(newcg); + newcg = NULL; - STRV_FOREACH(controller, arg_controllers) { - k = cg_create_and_attach(*controller, newcg, 0); - if (k < 0) - log_warning("Failed to create cgroup in controller %s: %s", *controller, strerror(-k)); + goto finish; } master = posix_openpt(O_RDWR|O_NOCTTY|O_CLOEXEC|O_NDELAY); @@ -1264,23 +1372,31 @@ int main(int argc, char *argv[]) { } if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { - log_error("Failed to create kmsg socket pair"); + log_error("Failed to create kmsg socket pair."); 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) { + 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) @@ -1314,6 +1430,7 @@ 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]); @@ -1367,6 +1484,11 @@ int main(int argc, char *argv[]) { goto child_fail; } + 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. */ @@ -1393,6 +1515,9 @@ int main(int argc, char *argv[]) { if (copy_devnodes(arg_directory) < 0) goto child_fail; + if (setup_ptmx(arg_directory) < 0) + goto child_fail; + dev_setup(arg_directory); if (setup_dev_console(arg_directory, console) < 0) @@ -1529,7 +1654,7 @@ int main(int argc, char *argv[]) { } if ((asprintf((char **)(envp + n_env++), "LISTEN_FDS=%u", n_fd_passed) < 0) || - (asprintf((char **)(envp + n_env++), "LISTEN_PID=%lu", (unsigned long) getpid()) < 0)) { + (asprintf((char **)(envp + n_env++), "LISTEN_PID=%lu", (unsigned long) 1) < 0)) { log_oom(); goto child_fail; } @@ -1568,10 +1693,17 @@ int main(int argc, char *argv[]) { _exit(EXIT_FAILURE); } - log_info("Init process in the container running as PID %d", pid); + log_info("Init process in the container running as PID %lu.", (unsigned long) pid); 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; @@ -1581,16 +1713,16 @@ 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; } @@ -1622,21 +1754,14 @@ finish: if (saved_attr_valid) tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); - if (master >= 0) - close_nointr_nofail(master); - close_pipe(kmsg_socket_pair); - if (oldcg) - cg_attach(SYSTEMD_CGROUP_CONTROLLER, oldcg, 0); - if (newcg) cg_kill_recursive_and_wait(SYSTEMD_CGROUP_CONTROLLER, newcg, true); free(arg_directory); + free(arg_machine); strv_free(arg_controllers); - free(oldcg); - free(newcg); fdset_free(fds);