#include <sys/reboot.h>
#include <linux/reboot.h>
#include <sys/wait.h>
-#include <sys/types.h>
#include <sys/stat.h>
#include <sys/mount.h>
#include <sys/syscall.h>
#include <stdbool.h>
#include <stdlib.h>
#include <string.h>
+#include <getopt.h>
#include "missing.h"
#include "log.h"
+#include "fileio.h"
#include "umount.h"
#include "util.h"
#include "mkdir.h"
#include "virt.h"
#include "watchdog.h"
+#include "killall.h"
+#include "cgroup-util.h"
+#include "def.h"
-#define TIMEOUT_USEC (5 * USEC_PER_SEC)
#define FINALIZE_ATTEMPTS 50
-static bool ignore_proc(pid_t pid) {
- char buf[PATH_MAX];
- FILE *f;
- char c;
- size_t count;
- uid_t uid;
- int r;
-
- /* We are PID 1, let's not commit suicide */
- if (pid == 1)
- return true;
-
- r = get_process_uid(pid, &uid);
- if (r < 0)
- return true; /* not really, but better safe than sorry */
-
- /* Non-root processes otherwise are always subject to be killed */
- if (uid != 0)
- return false;
-
- snprintf(buf, sizeof(buf), "/proc/%lu/cmdline", (unsigned long) pid);
- char_array_0(buf);
+static char* arg_verb;
- f = fopen(buf, "re");
- if (!f)
- return true; /* not really, but has the desired effect */
+static int parse_argv(int argc, char *argv[]) {
+ enum {
+ ARG_LOG_LEVEL = 0x100,
+ ARG_LOG_TARGET,
+ ARG_LOG_COLOR,
+ ARG_LOG_LOCATION,
+ };
- count = fread(&c, 1, 1, f);
- fclose(f);
+ static const struct option options[] = {
+ { "log-level", required_argument, NULL, ARG_LOG_LEVEL },
+ { "log-target", required_argument, NULL, ARG_LOG_TARGET },
+ { "log-color", optional_argument, NULL, ARG_LOG_COLOR },
+ { "log-location", optional_argument, NULL, ARG_LOG_LOCATION },
+ {}
+ };
- /* Kernel threads have an empty cmdline */
- if (count <= 0)
- return true;
+ int c, r;
- /* Processes with argv[0][0] = '@' we ignore from the killing
- * spree.
- *
- * http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons */
- if (count == 1 && c == '@')
- return true;
-
- return false;
-}
+ assert(argc >= 1);
+ assert(argv);
-static int killall(int sign) {
- DIR *dir;
- struct dirent *d;
- unsigned int n_processes = 0;
+ opterr = 0;
- dir = opendir("/proc");
- if (!dir)
- return -errno;
+ while ((c = getopt_long(argc, argv, ":", options, NULL)) >= 0)
+ switch (c) {
- while ((d = readdir(dir))) {
- pid_t pid;
+ case ARG_LOG_LEVEL:
+ r = log_set_max_level_from_string(optarg);
+ if (r < 0)
+ log_error("Failed to parse log level %s, ignoring.", optarg);
- if (parse_pid(d->d_name, &pid) < 0)
- continue;
+ break;
- if (ignore_proc(pid))
- continue;
+ case ARG_LOG_TARGET:
+ r = log_set_target_from_string(optarg);
+ if (r < 0)
+ log_error("Failed to parse log target %s, ignoring", optarg);
- if (kill(pid, sign) == 0)
- n_processes++;
- else
- log_warning("Could not kill %d: %m", pid);
- }
-
- closedir(dir);
+ break;
- return n_processes;
-}
+ case ARG_LOG_COLOR:
-static void wait_for_children(int n_processes, sigset_t *mask) {
- usec_t until;
+ if (optarg) {
+ r = log_show_color_from_string(optarg);
+ if (r < 0)
+ log_error("Failed to parse log color setting %s, ignoring", optarg);
+ } else
+ log_show_color(true);
- assert(mask);
+ break;
- until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
- for (;;) {
- struct timespec ts;
- int k;
- usec_t n;
+ case ARG_LOG_LOCATION:
+ if (optarg) {
+ r = log_show_location_from_string(optarg);
+ if (r < 0)
+ log_error("Failed to parse log location setting %s, ignoring", optarg);
+ } else
+ log_show_location(true);
- for (;;) {
- pid_t pid = waitpid(-1, NULL, WNOHANG);
+ break;
- if (pid == 0)
- break;
+ case '?':
+ log_error("Unknown option %s.", argv[optind-1]);
+ return -EINVAL;
- if (pid < 0 && errno == ECHILD)
- return;
+ case ':':
+ log_error("Missing argument to %s.", argv[optind-1]);
+ return -EINVAL;
- if (n_processes > 0)
- if (--n_processes == 0)
- return;
+ default:
+ assert_not_reached("Unhandled option code.");
}
- n = now(CLOCK_MONOTONIC);
- if (n >= until)
- return;
-
- timespec_store(&ts, until - n);
-
- if ((k = sigtimedwait(mask, NULL, &ts)) != SIGCHLD) {
-
- if (k < 0 && errno != EAGAIN) {
- log_error("sigtimedwait() failed: %m");
- return;
- }
-
- if (k >= 0)
- log_warning("sigtimedwait() returned unexpected signal.");
- }
+ if (optind >= argc) {
+ log_error("Verb argument missing.");
+ return -EINVAL;
}
-}
-
-static void send_signal(int sign) {
- sigset_t mask, oldmask;
- int n_processes;
-
- assert_se(sigemptyset(&mask) == 0);
- assert_se(sigaddset(&mask, SIGCHLD) == 0);
- assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0);
-
- if (kill(-1, SIGSTOP) < 0 && errno != ESRCH)
- log_warning("kill(-1, SIGSTOP) failed: %m");
-
- n_processes = killall(sign);
-
- if (kill(-1, SIGCONT) < 0 && errno != ESRCH)
- log_warning("kill(-1, SIGCONT) failed: %m");
-
- if (n_processes <= 0)
- goto finish;
-
- wait_for_children(n_processes, &mask);
-
-finish:
- sigprocmask(SIG_SETMASK, &oldmask, NULL);
-}
-
-static void ultimate_send_signal(int sign) {
- sigset_t mask, oldmask;
- int r;
- assert_se(sigemptyset(&mask) == 0);
- assert_se(sigaddset(&mask, SIGCHLD) == 0);
- assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0);
+ arg_verb = argv[optind];
- if (kill(-1, SIGSTOP) < 0 && errno != ESRCH)
- log_warning("kill(-1, SIGSTOP) failed: %m");
-
- r = kill(-1, sign);
- if (r < 0 && errno != ESRCH)
- log_warning("kill(-1, %s) failed: %m", signal_to_string(sign));
-
- if (kill(-1, SIGCONT) < 0 && errno != ESRCH)
- log_warning("kill(-1, SIGCONT) failed: %m");
-
- if (r < 0)
- goto finish;
-
- wait_for_children(0, &mask);
-
-finish:
- sigprocmask(SIG_SETMASK, &oldmask, NULL);
+ if (optind + 1 < argc)
+ log_error("Excess arguments, ignoring");
+ return 0;
}
static int prepare_new_root(void) {
}
static int pivot_to_new_root(void) {
+
if (chdir("/run/initramfs") < 0) {
log_error("Failed to change directory to /run/initramfs: %m");
return -errno;
}
- /*
- In case some evil process made "/" MS_SHARED
- It works for pivot_root, but the ref count for the root device
- is not decreasing :-/
- */
- if (mount(NULL, "/", NULL, MS_PRIVATE, NULL) < 0) {
- log_error("Failed to make \"/\" private mount %m");
- return -errno;
- }
+ /* Work-around for a kernel bug: for some reason the kernel
+ * refuses switching root if any file systems are mounted
+ * MS_SHARED. Hence remount them MS_PRIVATE here as a
+ * work-around.
+ *
+ * https://bugzilla.redhat.com/show_bug.cgi?id=847418 */
+ if (mount(NULL, "/", NULL, MS_REC|MS_PRIVATE, NULL) < 0)
+ log_warning("Failed to make \"/\" private mount: %m");
if (pivot_root(".", "oldroot") < 0) {
log_error("pivot failed: %m");
- /* only chroot if pivot root succeded */
+ /* only chroot if pivot root succeeded */
return -errno;
}
}
int main(int argc, char *argv[]) {
- int cmd, r;
- unsigned retries;
bool need_umount = true, need_swapoff = true, need_loop_detach = true, need_dm_detach = true;
- bool killed_everbody = false, in_container, use_watchdog = false;
+ bool in_container, use_watchdog = false;
+ _cleanup_free_ char *cgroup = NULL;
char *arguments[3];
+ unsigned retries;
+ int cmd, r;
log_parse_environment();
- log_set_target(LOG_TARGET_CONSOLE); /* syslog will die if not gone yet */
+ r = parse_argv(argc, argv);
+ if (r < 0)
+ goto error;
+
+ /* journald will die if not gone yet. The log target defaults
+ * to console, but may have been changed by commandline options. */
+
+ log_close_console(); /* force reopen of /dev/console */
log_open();
umask(0022);
if (getpid() != 1) {
- log_error("Not executed by init (pid 1).");
+ log_error("Not executed by init (PID 1).");
r = -EPERM;
goto error;
}
- if (argc != 2) {
- log_error("Invalid number of arguments.");
- r = -EINVAL;
- goto error;
- }
-
in_container = detect_container(NULL) > 0;
- if (streq(argv[1], "reboot"))
+ if (streq(arg_verb, "reboot"))
cmd = RB_AUTOBOOT;
- else if (streq(argv[1], "poweroff"))
+ else if (streq(arg_verb, "poweroff"))
cmd = RB_POWER_OFF;
- else if (streq(argv[1], "halt"))
+ else if (streq(arg_verb, "halt"))
cmd = RB_HALT_SYSTEM;
- else if (streq(argv[1], "kexec"))
+ else if (streq(arg_verb, "kexec"))
cmd = LINUX_REBOOT_CMD_KEXEC;
else {
- log_error("Unknown action '%s'.", argv[1]);
r = -EINVAL;
+ log_error("Unknown action '%s'.", arg_verb);
goto error;
}
+ cg_get_root_path(&cgroup);
+
use_watchdog = !!getenv("WATCHDOG_USEC");
/* lock us into memory */
mlockall(MCL_CURRENT|MCL_FUTURE);
log_info("Sending SIGTERM to remaining processes...");
- send_signal(SIGTERM);
+ broadcast_signal(SIGTERM, true, true);
log_info("Sending SIGKILL to remaining processes...");
- send_signal(SIGKILL);
+ broadcast_signal(SIGKILL, true, false);
if (in_container) {
need_swapoff = false;
need_dm_detach = false;
+ need_loop_detach = false;
}
/* Unmount all mountpoints, swaps, and loopback devices */
if (use_watchdog)
watchdog_ping();
+ /* Let's trim the cgroup tree on each iteration so
+ that we leave an empty cgroup tree around, so that
+ container managers get a nice notify event when we
+ are down */
+ if (cgroup)
+ cg_trim(SYSTEMD_CGROUP_CONTROLLER, cgroup, false);
+
if (need_umount) {
log_info("Unmounting file systems.");
r = umount_all(&changed);
- if (r == 0)
+ if (r == 0) {
need_umount = false;
- else if (r > 0)
+ log_info("All filesystems unmounted.");
+ } else if (r > 0)
log_info("Not all file systems unmounted, %d left.", r);
else
log_error("Failed to unmount file systems: %s", strerror(-r));
}
if (need_swapoff) {
- log_info("Disabling swaps.");
+ log_info("Deactivating swaps.");
r = swapoff_all(&changed);
- if (r == 0)
+ if (r == 0) {
need_swapoff = false;
- else if (r > 0)
- log_info("Not all swaps are turned off, %d left.", r);
+ log_info("All swaps deactivated.");
+ } else if (r > 0)
+ log_info("Not all swaps deactivated, %d left.", r);
else
- log_error("Failed to turn off swaps: %s", strerror(-r));
+ log_error("Failed to deactivate swaps: %s", strerror(-r));
}
if (need_loop_detach) {
log_info("Detaching loop devices.");
r = loopback_detach_all(&changed);
- if (r == 0)
+ if (r == 0) {
need_loop_detach = false;
- else if (r > 0)
+ log_info("All loop devices detached.");
+ } else if (r > 0)
log_info("Not all loop devices detached, %d left.", r);
else
log_error("Failed to detach loop devices: %s", strerror(-r));
if (need_dm_detach) {
log_info("Detaching DM devices.");
r = dm_detach_all(&changed);
- if (r == 0)
+ if (r == 0) {
need_dm_detach = false;
- else if (r > 0)
- log_warning("Not all DM devices detached, %d left.", r);
+ log_info("All DM devices detached.");
+ } else if (r > 0)
+ log_info("Not all DM devices detached, %d left.", r);
else
log_error("Failed to detach DM devices: %s", strerror(-r));
}
}
/* If in this iteration we didn't manage to
- * unmount/deactivate anything, we either kill more
- * processes, or simply give up */
+ * unmount/deactivate anything, we simply give up */
if (!changed) {
-
- if (killed_everbody) {
- /* Hmm, we already killed everybody,
- * let's just give up */
- log_error("Cannot finalize remaining file systems and devices, giving up.");
- break;
- }
-
- log_warning("Cannot finalize remaining file systems and devices, trying to kill remaining processes.");
- ultimate_send_signal(SIGTERM);
- ultimate_send_signal(SIGKILL);
- killed_everbody = true;
+ log_error("Cannot finalize remaining file systems and devices, giving up.");
+ break;
}
log_debug("Couldn't finalize remaining file systems and devices after %u retries, trying again.", retries+1);
if (retries >= FINALIZE_ATTEMPTS)
log_error("Too many iterations, giving up.");
+ else
+ log_info("Storage is finalized.");
arguments[0] = NULL;
- arguments[1] = argv[1];
+ arguments[1] = arg_verb;
arguments[2] = NULL;
- execute_directory(SYSTEM_SHUTDOWN_PATH, NULL, arguments);
-
- /* If we are in a container, just exit, this will kill our
- * container for good. */
- if (in_container) {
- log_error("Exiting container.");
- exit(0);
- }
+ execute_directory(SYSTEM_SHUTDOWN_PATH, NULL, DEFAULT_TIMEOUT_USEC, arguments);
- if (access("/run/initramfs/shutdown", X_OK) == 0) {
+ if (!in_container && !in_initrd() &&
+ access("/run/initramfs/shutdown", X_OK) == 0) {
if (prepare_new_root() >= 0 &&
pivot_to_new_root() >= 0) {
- execv("/shutdown", argv);
+ arguments[0] = (char*) "/shutdown";
+
+ log_info("Returning to initrd...");
+
+ execv("/shutdown", arguments);
log_error("Failed to execute shutdown binary: %m");
}
}
- sync();
-
- if (cmd == LINUX_REBOOT_CMD_KEXEC) {
- /* We cheat and exec kexec to avoid doing all its work */
- pid_t pid = fork();
-
- if (pid < 0)
- log_error("Could not fork: %m. Falling back to normal reboot.");
- else if (pid > 0) {
- wait_for_terminate_and_warn("kexec", pid);
- log_warning("kexec failed. Falling back to normal reboot.");
- } else {
- /* Child */
- const char *args[3] = { "/sbin/kexec", "-e", NULL };
- execv(args[0], (char * const *) args);
- return EXIT_FAILURE;
+ /* The kernel will automaticall flush ATA disks and suchlike
+ * on reboot(), but the file systems need to be synce'd
+ * explicitly in advance. So let's do this here, but not
+ * needlessly slow down containers. */
+ if (!in_container)
+ sync();
+
+ switch (cmd) {
+
+ case LINUX_REBOOT_CMD_KEXEC:
+
+ if (!in_container) {
+ /* We cheat and exec kexec to avoid doing all its work */
+ pid_t pid;
+
+ log_info("Rebooting with kexec.");
+
+ pid = fork();
+ if (pid < 0)
+ log_error("Failed to fork: %m");
+ else if (pid == 0) {
+
+ const char * const args[] = {
+ KEXEC, "-e", NULL
+ };
+
+ /* Child */
+
+ execv(args[0], (char * const *) args);
+ _exit(EXIT_FAILURE);
+ } else
+ wait_for_terminate_and_warn("kexec", pid);
}
cmd = RB_AUTOBOOT;
+ /* Fall through */
+
+ case RB_AUTOBOOT:
+
+ if (!in_container) {
+ _cleanup_free_ char *param = NULL;
+
+ if (read_one_line_file(REBOOT_PARAM_FILE, ¶m) >= 0) {
+ log_info("Rebooting with argument '%s'.", param);
+ syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2,
+ LINUX_REBOOT_CMD_RESTART2, param);
+ }
+ }
+
+ log_info("Rebooting.");
+ break;
+
+ case RB_POWER_OFF:
+ log_info("Powering off.");
+ break;
+
+ case RB_HALT_SYSTEM:
+ log_info("Halting system.");
+ break;
+
+ default:
+ assert_not_reached("Unknown magic");
}
reboot(cmd);
+ if (errno == EPERM && in_container) {
+ /* If we are in a container, and we lacked
+ * CAP_SYS_BOOT just exit, this will kill our
+ * container for good. */
+ log_info("Exiting container.");
+ exit(0);
+ }
+
log_error("Failed to invoke reboot(): %m");
r = -errno;
log_error("Critical error while doing system shutdown: %s", strerror(-r));
freeze();
- return EXIT_FAILURE;
}