#include <sys/signalfd.h>
#include <grp.h>
#include <linux/fs.h>
+#include <linux/netlink.h>
#include <sys/un.h>
#include <sys/socket.h>
+#ifdef HAVE_XATTR
+#include <attr/xattr.h>
+#endif
+
#include <systemd/sd-daemon.h>
#include "log.h"
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) {
break;
case ARG_UUID:
+ if (!id128_is_valid(optarg)) {
+ log_error("Invalid UUID: %s", optarg);
+ return -EINVAL;
+ }
+
arg_uuid = optarg;
break;
return 0;
}
+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;
+
+ assert(cgroup);
+ assert(pid >= 0);
+ assert(arg_directory);
+
+ 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
+}
+
static int drop_capabilities(void) {
return capability_bounding_set_drop(~arg_retain, false);
}
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;
- _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;
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;
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;
}
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;
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;
}
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);
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)
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]);
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. */
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;
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;
}
~ianmdlvl / elogind.git / blobdiff