X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?a=blobdiff_plain;f=src%2Fnspawn%2Fnspawn.c;h=8b574214314a297ac7d5c68a764219c5354d21cc;hb=37d3ab1b7e114f0fb6dfb2e7273569b42794b76a;hp=d181e3b338564c76de9caa57832fdb991005fb85;hpb=d40361453be0525a0455d549b2b863931b069358;p=elogind.git diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c index d181e3b33..8b5742143 100644 --- a/src/nspawn/nspawn.c +++ b/src/nspawn/nspawn.c @@ -33,6 +33,7 @@ #include #include #include +#include #include #include #include @@ -55,6 +56,9 @@ #include "loopback-setup.h" #include "sd-id128.h" #include "dev-setup.h" +#include "fdset.h" +#include "build.h" +#include "fileio.h" typedef enum LinkJournal { LINK_NO, @@ -94,23 +98,33 @@ static uint64_t arg_retain = (1ULL << CAP_SYS_PTRACE) | (1ULL << CAP_SYS_TTY_CONFIG) | (1ULL << CAP_SYS_RESOURCE) | - (1ULL << CAP_SYS_BOOT); + (1ULL << CAP_SYS_BOOT) | + (1ULL << CAP_AUDIT_WRITE) | + (1ULL << CAP_AUDIT_CONTROL); +static char **arg_bind = NULL; +static char **arg_bind_ro = NULL; 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" - " -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 cgroup hierarchies\n" - " --uuid=UUID Set a specific machine UUID 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 capability\n" - " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n" - " -j Equivalent to --link-journal=host\n", + " -h --help Show this help\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" + " --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" + " capability\n" + " --link-journal=MODE Link up guest journal, one of no, auto, guest, host\n" + " -j Equivalent to --link-journal=host\n" + " --bind=PATH[:PATH] Bind mount a file or directory from the host into\n" + " the container\n" + " --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n", program_invocation_short_name); return 0; @@ -119,15 +133,19 @@ static int help(void) { static int parse_argv(int argc, char *argv[]) { enum { - ARG_PRIVATE_NETWORK = 0x100, + ARG_VERSION = 0x100, + ARG_PRIVATE_NETWORK, ARG_UUID, ARG_READ_ONLY, ARG_CAPABILITY, - ARG_LINK_JOURNAL + ARG_LINK_JOURNAL, + ARG_BIND, + ARG_BIND_RO }; static const struct option options[] = { { "help", no_argument, NULL, 'h' }, + { "version", no_argument, NULL, ARG_VERSION }, { "directory", required_argument, NULL, 'D' }, { "user", required_argument, NULL, 'u' }, { "controllers", required_argument, NULL, 'C' }, @@ -137,6 +155,8 @@ static int parse_argv(int argc, char *argv[]) { { "read-only", no_argument, NULL, ARG_READ_ONLY }, { "capability", required_argument, NULL, ARG_CAPABILITY }, { "link-journal", required_argument, NULL, ARG_LINK_JOURNAL }, + { "bind", required_argument, NULL, ARG_BIND }, + { "bind-ro", required_argument, NULL, ARG_BIND_RO }, { NULL, 0, NULL, 0 } }; @@ -153,6 +173,11 @@ static int parse_argv(int argc, char *argv[]) { help(); return 0; + case ARG_VERSION: + puts(PACKAGE_STRING); + puts(SYSTEMD_FEATURES); + return 0; + case 'D': free(arg_directory); arg_directory = canonicalize_file_name(optarg); @@ -244,6 +269,43 @@ static int parse_argv(int argc, char *argv[]) { break; + case ARG_BIND: + case ARG_BIND_RO: { + _cleanup_free_ char *a = NULL, *b = NULL; + char *e; + char ***x; + int r; + + x = c == ARG_BIND ? &arg_bind : &arg_bind_ro; + + e = strchr(optarg, ':'); + if (e) { + a = strndup(optarg, e - optarg); + b = strdup(e + 1); + } else { + a = strdup(optarg); + b = strdup(optarg); + } + + if (!a || !b) + return log_oom(); + + if (!path_is_absolute(a) || !path_is_absolute(b)) { + log_error("Invalid bind mount specification: %s", optarg); + return -EINVAL; + } + + r = strv_extend(x, a); + if (r < 0) + return r; + + r = strv_extend(x, b); + if (r < 0) + return r; + + break; + } + case '?': return -EINVAL; @@ -289,14 +351,9 @@ static int mount_all(const char *dest) { char _cleanup_free_ *where = NULL; int t; - if (asprintf(&where, "%s/%s", dest, mount_table[k].where) < 0) { - log_oom(); - - if (r == 0) - r = -ENOMEM; - - break; - } + where = strjoin(dest, "/", mount_table[k].where, NULL); + if (!where) + return log_oom(); t = path_is_mount_point(where, true); if (t < 0) { @@ -312,7 +369,7 @@ static int mount_all(const char *dest) { if (mount_table[k].what && t > 0) continue; - mkdir_p_label(where, 0755); + mkdir_p(where, 0755); if (mount(mount_table[k].what, where, @@ -331,6 +388,32 @@ static int mount_all(const char *dest) { return r; } +static int mount_binds(const char *dest, char **l, unsigned long flags) { + char **x, **y; + + STRV_FOREACH_PAIR(x, y, l) { + _cleanup_free_ char *where = NULL; + + where = strjoin(dest, "/", *y, NULL); + if (!where) + return log_oom(); + + mkdir_p_label(where, 0755); + + if (mount(*x, where, "bind", MS_BIND, NULL) < 0) { + log_error("mount(%s) failed: %m", where); + return -errno; + } + + if (flags && mount(NULL, where, NULL, MS_REMOUNT|MS_BIND|flags, NULL) < 0) { + log_error("mount(%s) failed: %m", where); + return -errno; + } + } + + return 0; +} + static int setup_timezone(const char *dest) { _cleanup_free_ char *where = NULL, *p = NULL, *q = NULL, *check = NULL, *what = NULL; char *z, *y; @@ -404,6 +487,8 @@ static int setup_resolv_conf(const char *dest) { if (!where) return log_oom(); + /* 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); @@ -666,58 +751,58 @@ static int setup_hostname(void) { static int setup_journal(const char *directory) { sd_id128_t machine_id; - char *p = NULL, *b = NULL, *l, *q = NULL, *d = NULL; + char _cleanup_free_ *p = NULL, *b = NULL, *q = NULL, *d = NULL; + char *id; int r; if (arg_link_journal == LINK_NO) return 0; p = strappend(directory, "/etc/machine-id"); - if (!p) { - r = log_oom(); - goto finish; - } + if (!p) + return log_oom(); r = read_one_line_file(p, &b); - if (r == -ENOENT && arg_link_journal == LINK_AUTO) { - r = 0; - goto finish; - } else if (r < 0) { - log_error("Failed to read machine ID: %s", strerror(-r)); + if (r == -ENOENT && arg_link_journal == LINK_AUTO) + return 0; + else if (r < 0) { + log_error("Failed to read machine ID from %s: %s", p, strerror(-r)); return r; } - l = strstrip(b); - if (isempty(l) && arg_link_journal == LINK_AUTO) { - r = 0; - goto finish; - } + id = strstrip(b); + if (isempty(id) && arg_link_journal == LINK_AUTO) + return 0; - /* Verify validaty */ - r = sd_id128_from_string(l, &machine_id); + /* Verify validity */ + r = sd_id128_from_string(id, &machine_id); if (r < 0) { - log_error("Failed to parse machine ID: %s", strerror(-r)); - goto finish; + log_error("Failed to parse machine ID from %s: %s", p, strerror(-r)); + return r; } free(p); - p = strappend("/var/log/journal/", l); - q = strjoin(directory, "/var/log/journal/", l, NULL); - if (!p || !q) { - r = log_oom(); - goto finish; + p = strappend("/var/log/journal/", id); + q = strjoin(directory, "/var/log/journal/", id, NULL); + if (!p || !q) + return log_oom(); + + if (path_is_mount_point(p, false) > 0) { + if (arg_link_journal != LINK_AUTO) { + log_error("%s: already a mount point, refusing to use for journal", p); + return -EEXIST; + } + + return 0; } - if (path_is_mount_point(p, false) > 0 || - path_is_mount_point(q, false) > 0) { + if (path_is_mount_point(q, false) > 0) { if (arg_link_journal != LINK_AUTO) { - log_error("Journal already a mount point, refusing."); - r = -EEXIST; - goto finish; + log_error("%s: already a mount point, refusing to use for journal", q); + return -EEXIST; } - r = 0; - goto finish; + return 0; } r = readlink_and_make_absolute(p, &d); @@ -726,89 +811,74 @@ static int setup_journal(const char *directory) { arg_link_journal == LINK_AUTO) && path_equal(d, q)) { - mkdir_p(q, 0755); - - r = 0; - goto finish; + r = mkdir_p(q, 0755); + if (r < 0) + log_warning("failed to create directory %s: %m", q); + return 0; } if (unlink(p) < 0) { log_error("Failed to remove symlink %s: %m", p); - r = -errno; - goto finish; + return -errno; } } else if (r == -EINVAL) { if (arg_link_journal == LINK_GUEST && rmdir(p) < 0) { - if (errno == ENOTDIR) - log_error("%s already exists and is neither symlink nor directory.", p); - else { + if (errno == ENOTDIR) { + log_error("%s already exists and is neither a symlink nor a directory", p); + return r; + } else { log_error("Failed to remove %s: %m", p); - r = -errno; + return -errno; } - - goto finish; } } else if (r != -ENOENT) { log_error("readlink(%s) failed: %m", p); - goto finish; + return r; } if (arg_link_journal == LINK_GUEST) { if (symlink(q, p) < 0) { log_error("Failed to symlink %s to %s: %m", q, p); - r = -errno; - goto finish; + return -errno; } - mkdir_p(q, 0755); - - r = 0; - goto finish; + r = mkdir_p(q, 0755); + if (r < 0) + log_warning("failed to create directory %s: %m", q); + return 0; } if (arg_link_journal == LINK_HOST) { r = mkdir_p(p, 0755); if (r < 0) { log_error("Failed to create %s: %m", p); - goto finish; + return r; } - } else if (access(p, F_OK) < 0) { - r = 0; - goto finish; - } + } else if (access(p, F_OK) < 0) + return 0; if (dir_is_empty(q) == 0) { log_error("%s not empty.", q); - r = -ENOTEMPTY; - goto finish; + return -ENOTEMPTY; } r = mkdir_p(q, 0755); if (r < 0) { log_error("Failed to create %s: %m", q); - goto finish; + return r; } if (mount(p, q, "bind", MS_BIND, NULL) < 0) { log_error("Failed to bind mount journal from host into guest: %m"); - r = -errno; - goto finish; + return -errno; } - r = 0; - -finish: - free(p); - free(q); - free(d); - free(b); - return r; - + return 0; } static int drop_capabilities(void) { @@ -829,13 +899,18 @@ static int is_os_tree(const char *path) { return r < 0 ? 0 : 1; } -static int process_pty(int master, sigset_t *mask) { +static int process_pty(int master, pid_t pid, sigset_t *mask) { char in_buffer[LINE_MAX], out_buffer[LINE_MAX]; size_t in_buffer_full = 0, out_buffer_full = 0; struct epoll_event stdin_ev, stdout_ev, master_ev, signal_ev; bool stdin_readable = false, stdout_writable = false, master_readable = false, master_writable = false; int ep = -1, signal_fd = -1, r; + bool tried_orderly_shutdown = false; + + assert(master >= 0); + assert(pid > 0); + assert(mask); fd_nonblock(STDIN_FILENO, 1); fd_nonblock(STDOUT_FILENO, 1); @@ -855,9 +930,19 @@ static int process_pty(int master, sigset_t *mask) { goto finish; } - zero(stdin_ev); - stdin_ev.events = EPOLLIN|EPOLLET; - stdin_ev.data.fd = STDIN_FILENO; + /* We read from STDIN only if this is actually a TTY, + * otherwise we assume non-interactivity. */ + if (isatty(STDIN_FILENO)) { + zero(stdin_ev); + stdin_ev.events = EPOLLIN|EPOLLET; + stdin_ev.data.fd = STDIN_FILENO; + + if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0) { + log_error("Failed to register STDIN in epoll: %m"); + r = -errno; + goto finish; + } + } zero(stdout_ev); stdout_ev.events = EPOLLOUT|EPOLLET; @@ -871,11 +956,19 @@ static int process_pty(int master, sigset_t *mask) { signal_ev.events = EPOLLIN; signal_ev.data.fd = signal_fd; - if (epoll_ctl(ep, EPOLL_CTL_ADD, STDIN_FILENO, &stdin_ev) < 0 || - epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0 || - epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 || + if (epoll_ctl(ep, EPOLL_CTL_ADD, STDOUT_FILENO, &stdout_ev) < 0) { + if (errno != EPERM) { + log_error("Failed to register stdout in epoll: %m"); + r = -errno; + goto finish; + } + /* stdout without epoll support. Likely redirected to regular file. */ + stdout_writable = true; + } + + if (epoll_ctl(ep, EPOLL_CTL_ADD, master, &master_ev) < 0 || epoll_ctl(ep, EPOLL_CTL_ADD, signal_fd, &signal_ev) < 0) { - log_error("Failed to regiser fds in epoll: %m"); + log_error("Failed to register fds in epoll: %m"); r = -errno; goto finish; } @@ -943,6 +1036,14 @@ static int process_pty(int master, sigset_t *mask) { /* The window size changed, let's forward that. */ if (ioctl(STDIN_FILENO, TIOCGWINSZ, &ws) >= 0) ioctl(master, TIOCSWINSZ, &ws); + } else if (sfsi.ssi_signo == SIGTERM && arg_boot && !tried_orderly_shutdown) { + + log_info("Trying to halt container. Send SIGTERM again to trigger immediate termination."); + + /* This only works for systemd... */ + tried_orderly_shutdown = true; + kill(pid, SIGRTMIN+3); + } else { r = 0; goto finish; @@ -1045,13 +1146,14 @@ int main(int argc, char *argv[]) { int r = EXIT_FAILURE, k; char *oldcg = NULL, *newcg = NULL; char **controller = NULL; - int master = -1; + int master = -1, n_fd_passed; const char *console = NULL; struct termios saved_attr, raw_attr; sigset_t mask; bool saved_attr_valid = false; struct winsize ws; int kmsg_socket_pair[2] = { -1, -1 }; + FDSet *fds = NULL; log_parse_environment(); log_open(); @@ -1096,6 +1198,18 @@ int main(int argc, char *argv[]) { goto finish; } + log_close(); + n_fd_passed = sd_listen_fds(false); + if (n_fd_passed > 0) { + k = fdset_new_listen_fds(&fds, false); + if (k < 0) { + log_error("Failed to collect file descriptors: %s", strerror(-k)); + goto finish; + } + } + fdset_close_others(fds); + log_open(); + k = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 0, &oldcg); if (k < 0) { log_error("Failed to determine current cgroup: %s", strerror(-k)); @@ -1141,16 +1255,13 @@ int main(int argc, char *argv[]) { goto finish; } - if (tcgetattr(STDIN_FILENO, &saved_attr) < 0) { - log_error("Failed to get terminal attributes: %m"); - goto finish; - } - - saved_attr_valid = true; + if (tcgetattr(STDIN_FILENO, &saved_attr) >= 0) { + saved_attr_valid = true; - raw_attr = saved_attr; - cfmakeraw(&raw_attr); - raw_attr.c_lflag &= ~ECHO; + raw_attr = saved_attr; + cfmakeraw(&raw_attr); + raw_attr.c_lflag &= ~ECHO; + } if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) { log_error("Failed to create kmsg socket pair"); @@ -1163,9 +1274,10 @@ int main(int argc, char *argv[]) { for (;;) { siginfo_t status; + int pipefd[2]; - if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { - log_error("Failed to set terminal attributes: %m"); + if(pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) { + log_error("pipe2(): %m"); goto finish; } @@ -1181,10 +1293,10 @@ int main(int argc, char *argv[]) { if (pid == 0) { /* child */ - const char *home = NULL; uid_t uid = (uid_t) -1; gid_t gid = (gid_t) -1; + unsigned n_env = 2; const char *envp[] = { "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin", "container=systemd-nspawn", /* LXC sets container=lxc, so follow the scheme here */ @@ -1193,28 +1305,57 @@ int main(int argc, char *argv[]) { NULL, /* USER */ NULL, /* LOGNAME */ NULL, /* container_uuid */ + NULL, /* LISTEN_FDS */ + NULL, /* LISTEN_PID */ NULL }; - envp[2] = strv_find_prefix(environ, "TERM="); + envp[n_env] = strv_find_prefix(environ, "TERM="); + if (envp[n_env]) + n_env ++; + + close_nointr_nofail(pipefd[1]); + fd_wait_for_event(pipefd[0], POLLHUP, -1); + close_nointr_nofail(pipefd[0]); close_nointr_nofail(master); + master = -1; + + if (saved_attr_valid) { + if (tcsetattr(STDIN_FILENO, TCSANOW, &raw_attr) < 0) { + log_error("Failed to set terminal attributes: %m"); + goto child_fail; + } + } close_nointr(STDIN_FILENO); close_nointr(STDOUT_FILENO); close_nointr(STDERR_FILENO); - close_all_fds(&kmsg_socket_pair[1], 1); + close_nointr_nofail(kmsg_socket_pair[0]); + kmsg_socket_pair[0] = -1; reset_all_signal_handlers(); assert_se(sigemptyset(&mask) == 0); assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); - if (open_terminal(console, O_RDWR) != STDIN_FILENO || - dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || - dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) + k = open_terminal(console, O_RDWR); + if (k != STDIN_FILENO) { + if (k >= 0) { + close_nointr_nofail(k); + k = -EINVAL; + } + + log_error("Failed to open console: %s", strerror(-k)); goto child_fail; + } + + if (dup2(STDIN_FILENO, STDOUT_FILENO) != STDOUT_FILENO || + dup2(STDIN_FILENO, STDERR_FILENO) != STDERR_FILENO) { + log_error("Failed to duplicate console: %m"); + goto child_fail; + } if (setsid() < 0) { log_error("setsid() failed: %m"); @@ -1261,6 +1402,7 @@ int main(int argc, char *argv[]) { goto child_fail; close_nointr_nofail(kmsg_socket_pair[1]); + kmsg_socket_pair[1] = -1; if (setup_boot_id(arg_directory) < 0) goto child_fail; @@ -1274,6 +1416,12 @@ int main(int argc, char *argv[]) { if (setup_journal(arg_directory) < 0) goto child_fail; + if (mount_binds(arg_directory, arg_bind, 0) < 0) + goto child_fail; + + if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0) + goto child_fail; + if (chdir(arg_directory) < 0) { log_error("chdir(%s) failed: %m", arg_directory); goto child_fail; @@ -1305,6 +1453,12 @@ int main(int argc, char *argv[]) { if (arg_user) { + /* Note that this resolves user names + * inside the container, and hence + * accesses the NSS modules from the + * container and not the host. This is + * a bit weird... */ + if (get_user_creds((const char**)&arg_user, &uid, &gid, &home, NULL) < 0) { log_error("get_user_creds() failed: %m"); goto child_fail; @@ -1334,17 +1488,48 @@ int main(int argc, char *argv[]) { log_error("setreuid() failed: %m"); goto child_fail; } + } else { + /* Reset everything fully to 0, just in case */ + + if (setgroups(0, NULL) < 0) { + log_error("setgroups() failed: %m"); + goto child_fail; + } + + if (setresgid(0, 0, 0) < 0) { + log_error("setregid() failed: %m"); + goto child_fail; + } + + if (setresuid(0, 0, 0) < 0) { + log_error("setreuid() failed: %m"); + goto child_fail; + } } - if ((asprintf((char**)(envp + 3), "HOME=%s", home ? home: "/root") < 0) || - (asprintf((char**)(envp + 4), "USER=%s", arg_user ? arg_user : "root") < 0) || - (asprintf((char**)(envp + 5), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) { + if ((asprintf((char**)(envp + n_env++), "HOME=%s", home ? home: "/root") < 0) || + (asprintf((char**)(envp + n_env++), "USER=%s", arg_user ? arg_user : "root") < 0) || + (asprintf((char**)(envp + n_env++), "LOGNAME=%s", arg_user ? arg_user : "root") < 0)) { log_oom(); goto child_fail; } if (arg_uuid) { - if (asprintf((char**)(envp + 6), "container_uuid=%s", arg_uuid) < 0) { + if (asprintf((char**)(envp + n_env++), "container_uuid=%s", arg_uuid) < 0) { + log_oom(); + goto child_fail; + } + } + + if (fdset_size(fds) > 0) { + k = fdset_cloexec(fds, false); + if (k < 0) { + log_error("Failed to unset O_CLOEXEC for file descriptors."); + goto child_fail; + } + + 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)) { log_oom(); goto child_fail; } @@ -1383,9 +1568,15 @@ int main(int argc, char *argv[]) { _exit(EXIT_FAILURE); } - if (process_pty(master, &mask) < 0) - goto finish; + log_info("Init process in the container running as PID %d", pid); + close_nointr_nofail(pipefd[0]); + close_nointr_nofail(pipefd[1]); + + fdset_free(fds); + fds = NULL; + if (process_pty(master, pid, &mask) < 0) + goto finish; if (saved_attr_valid) tcsetattr(STDIN_FILENO, TCSANOW, &saved_attr); @@ -1447,5 +1638,7 @@ finish: free(oldcg); free(newcg); + fdset_free(fds); + return r; }