#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 <fcntl.h>
#include <dirent.h>
#include <errno.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
+#include "missing.h"
#include "log.h"
#include "umount.h"
#include "util.h"
+#include "virt.h"
#define TIMEOUT_USEC (5 * USEC_PER_SEC)
#define FINALIZE_ATTEMPTS 50
-#define FINALIZE_CRITICAL_ATTEMPTS 10
static bool ignore_proc(pid_t pid) {
if (pid == 1)
static int killall(int sign) {
DIR *dir;
struct dirent *d;
- unsigned int processes = 0;
+ unsigned int n_processes = 0;
if ((dir = opendir("/proc")) == NULL)
return -errno;
continue;
if (kill(pid, sign) == 0)
- processes++;
+ n_processes++;
else
log_warning("Could not kill %d: %m", pid);
}
closedir(dir);
- return processes;
+ return n_processes;
}
-static int send_signal(int sign) {
- sigset_t mask, oldmask;
+static void wait_for_children(int n_processes, sigset_t *mask) {
usec_t until;
- int processes;
- struct timespec ts;
-
- assert_se(sigemptyset(&mask) == 0);
- assert_se(sigaddset(&mask, SIGCHLD) == 0);
- if (sigprocmask(SIG_BLOCK, &mask, &oldmask) != 0)
- return -errno;
-
- if (kill(-1, SIGSTOP) < 0)
- log_warning("Failed kill(-1, SIGSTOP): %m");
- processes = killall(sign);
-
- if (kill(-1, SIGCONT) < 0)
- log_warning("Failed kill(-1, SIGCONT): %m");
-
- if (processes <= 0)
- goto finish;
+ assert(mask);
until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
for (;;) {
- usec_t n = now(CLOCK_MONOTONIC);
+ struct timespec ts;
+ int k;
+ usec_t n;
+
for (;;) {
pid_t pid = waitpid(-1, NULL, WNOHANG);
+
if (pid == 0)
break;
- else if (pid < 0 && errno == ECHILD) {
- processes = 0;
- goto finish;
- }
- if (--processes == 0)
- goto finish;
+ if (pid < 0 && errno == ECHILD)
+ return;
+
+ if (n_processes > 0)
+ if (--n_processes == 0)
+ return;
}
+ n = now(CLOCK_MONOTONIC);
if (n >= until)
- goto finish;
+ return;
timespec_store(&ts, until - n);
- if (sigtimedwait(&mask, NULL, &ts) != SIGCHLD)
- log_warning("Failed: sigtimedwait did not return SIGCHLD: %m");
+
+ 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.");
+ }
}
+}
+
+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);
-
- return processes;
}
-static int rescue_send_signal(int sign) {
+static void ultimate_send_signal(int sign) {
sigset_t mask, oldmask;
- usec_t until;
- struct timespec ts;
int r;
- sigemptyset(&mask);
- sigaddset(&mask, SIGCHLD);
- if (sigprocmask(SIG_BLOCK, &mask, &oldmask) != 0)
- return -errno;
+ assert_se(sigemptyset(&mask) == 0);
+ assert_se(sigaddset(&mask, SIGCHLD) == 0);
+ assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0);
- if (kill(-1, SIGSTOP) < 0)
- log_warning("Failed kill(-1, SIGSTOP): %m");
+ if (kill(-1, SIGSTOP) < 0 && errno != ESRCH)
+ log_warning("kill(-1, SIGSTOP) failed: %m");
r = kill(-1, sign);
- if (r < 0)
- log_warning("Failed kill(-1, %d): %m", sign);
+ if (r < 0 && errno != ESRCH)
+ log_warning("kill(-1, %s) failed: %m", signal_to_string(sign));
- if (kill(-1, SIGCONT) < 0)
- log_warning("Failed kill(-1, SIGCONT): %m");
+ if (kill(-1, SIGCONT) < 0 && errno != ESRCH)
+ log_warning("kill(-1, SIGCONT) failed: %m");
if (r < 0)
goto finish;
- until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
- for (;;) {
- usec_t n = now(CLOCK_MONOTONIC);
- for (;;) {
- pid_t pid = waitpid(-1, NULL, WNOHANG);
- if (pid == 0)
- break;
- else if (pid < 0 && errno == ECHILD)
- goto finish;
+ wait_for_children(0, &mask);
+
+finish:
+ sigprocmask(SIG_SETMASK, &oldmask, NULL);
+}
+
+static int prepare_new_root(void) {
+ static const char dirs[] =
+ "/run/initramfs/oldroot\0"
+ "/run/initramfs/proc\0"
+ "/run/initramfs/sys\0"
+ "/run/initramfs/dev\0"
+ "/run/initramfs/run\0";
+
+ const char *dir;
+
+ if (mount("/run/initramfs", "/run/initramfs", NULL, MS_BIND, NULL) < 0) {
+ log_error("Failed to mount bind /run/initramfs on /run/initramfs: %m");
+ return -errno;
+ }
+
+ if (mount(NULL, "/run/initramfs", NULL, MS_PRIVATE, NULL) < 0) {
+ log_error("Failed to make /run/initramfs private mount: %m");
+ return -errno;
+ }
+
+ NULSTR_FOREACH(dir, dirs)
+ if (mkdir_p(dir, 0755) < 0 && errno != EEXIST) {
+ log_error("Failed to mkdir %s: %m", dir);
+ return -errno;
}
- if (n >= until)
- goto finish;
+ if (mount("/sys", "/run/initramfs/sys", NULL, MS_BIND, NULL) < 0) {
+ log_error("Failed to mount bind /sys on /run/initramfs/sys: %m");
+ return -errno;
+ }
- timespec_store(&ts, until - n);
- if (sigtimedwait(&mask, NULL, &ts) != SIGCHLD)
- log_warning("Failed: sigtimedwait did not return SIGCHLD: %m");
+ if (mount("/proc", "/run/initramfs/proc", NULL, MS_BIND, NULL) < 0) {
+ log_error("Failed to mount bind /proc on /run/initramfs/proc: %m");
+ return -errno;
}
-finish:
- sigprocmask(SIG_SETMASK, &oldmask, NULL);
+ if (mount("/dev", "/run/initramfs/dev", NULL, MS_BIND, NULL) < 0) {
+ log_error("Failed to mount bind /dev on /run/initramfs/dev: %m");
+ return -errno;
+ }
+
+ if (mount("/run", "/run/initramfs/run", NULL, MS_BIND, NULL) < 0) {
+ log_error("Failed to mount bind /run on /run/initramfs/run: %m");
+ return -errno;
+ }
+
+ return 0;
+}
+
+static int pivot_to_new_root(void) {
+ int fd;
+
+ chdir("/run/initramfs");
+
+ /*
+ 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;
+ }
+
+ if (pivot_root(".", "oldroot") < 0) {
+ log_error("pivot failed: %m");
+ /* only chroot if pivot root succeded */
+ return -errno;
+ }
+
+ chroot(".");
+ log_info("Successfully changed into root pivot.");
+
+ fd = open("/dev/console", O_RDWR);
+ if (fd < 0)
+ log_error("Failed to open /dev/console: %m");
+ else {
+ make_stdio(fd);
+
+ /* Initialize the controlling terminal */
+ setsid();
+ ioctl(STDIN_FILENO, TIOCSCTTY, NULL);
+ }
- return r;
+ return 0;
}
int main(int argc, char *argv[]) {
- int cmd, r, retries;
- bool need_umount = true, need_swapoff = true, need_loop_detach = true;
+ 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;
log_parse_environment();
log_set_target(LOG_TARGET_CONSOLE); /* syslog will die if not gone yet */
log_open();
+ umask(0022);
+
if (getpid() != 1) {
log_error("Not executed by init (pid 1).");
r = -EPERM;
goto error;
}
+ in_container = detect_container(NULL) > 0;
+
if (streq(argv[1], "reboot"))
cmd = RB_AUTOBOOT;
else if (streq(argv[1], "poweroff"))
if (mlockall(MCL_CURRENT|MCL_FUTURE) != 0)
log_warning("Cannot lock process memory: %m");
- log_info("Sending SIGTERM to processes");
- r = send_signal(SIGTERM);
- if (r < 0)
- log_warning("Cannot send SIGTERM to all process: %s", strerror(r));
+ log_info("Sending SIGTERM to remaining processes...");
+ send_signal(SIGTERM);
- log_info("Sending SIGKILL to processes");
- r = send_signal(SIGKILL);
- if (r < 0)
- log_warning("Cannot send SIGKILL to all process: %s", strerror(r));
+ log_info("Sending SIGKILL to remaining processes...");
+ send_signal(SIGKILL);
+
+ if (in_container)
+ need_swapoff = false;
/* Unmount all mountpoints, swaps, and loopback devices */
- retries = FINALIZE_ATTEMPTS;
- while (need_umount || need_swapoff || need_loop_detach) {
+ for (retries = 0; retries < FINALIZE_ATTEMPTS; retries++) {
+ bool changed = false;
+
if (need_umount) {
- log_info("Unmounting filesystems.");
- r = umount_all();
+ log_info("Unmounting file systems.");
+ r = umount_all(&changed);
if (r == 0)
need_umount = false;
else if (r > 0)
- log_warning("Not all filesystems unmounted, %d left.", r);
+ log_info("Not all file systems unmounted, %d left.", r);
else
- log_error("Error unmounting filesystems: %s", strerror(-r));
+ log_error("Failed to unmount file systems: %s", strerror(-r));
}
if (need_swapoff) {
log_info("Disabling swaps.");
- r = swapoff_all();
+ r = swapoff_all(&changed);
if (r == 0)
need_swapoff = false;
else if (r > 0)
- log_warning("Not all swaps are off, %d left.", r);
+ log_info("Not all swaps are turned off, %d left.", r);
else
- log_error("Error turning off swaps: %s", strerror(-r));
+ log_error("Failed to turn off swaps: %s", strerror(-r));
}
if (need_loop_detach) {
log_info("Detaching loop devices.");
- r = loopback_detach_all();
+ r = loopback_detach_all(&changed);
if (r == 0)
need_loop_detach = false;
else if (r > 0)
- log_warning("Not all loop devices detached, %d left.", r);
+ log_info("Not all loop devices detached, %d left.", r);
else
- log_error("Error detaching loop devices: %s", strerror(-r));
+ 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)
+ need_dm_detach = false;
+ else if (r > 0)
+ log_warning("Not all DM devices detached, %d left.", r);
+ else
+ log_error("Failed to detach DM devices: %s", strerror(-r));
}
- if (need_umount || need_swapoff || need_loop_detach) {
- retries--;
+ if (!need_umount && !need_swapoff && !need_loop_detach && !need_dm_detach) {
+ if (retries > 0)
+ log_info("All filesystems, swaps, loop devices, DM devices detached.");
+ /* Yay, done */
+ break;
+ }
- if (retries == FINALIZE_CRITICAL_ATTEMPTS) {
- log_warning("Approaching critical level to finalize filesystem and devices, try to kill all processes.");
- rescue_send_signal(SIGTERM);
- rescue_send_signal(SIGKILL);
- }
+ /* If in this iteration we didn't manage to
+ * unmount/deactivate anything, we either kill more
+ * processes, or simply give up */
+ if (!changed) {
- if (retries > 0)
- log_info("Action still required, %d tries left.", retries);
- else {
- log_error("Giving up. Actions left: Umount=%s, Swap off=%s, Loop detach=%s",
- yes_no(need_umount), yes_no(need_swapoff), yes_no(need_loop_detach));
+ 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_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.");
+
+ execute_directory(SYSTEM_SHUTDOWN_PATH, NULL, NULL);
+
+ /* If we are in a container, just exit, this will kill our
+ * container for good. */
+ if (in_container) {
+ log_error("Exiting container.");
+ exit(0);
+ }
+
+ if (access("/run/initramfs/shutdown", X_OK) == 0) {
+
+ if (prepare_new_root() >= 0 &&
+ pivot_to_new_root() >= 0) {
+ execv("/shutdown", argv);
+ 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 */
+ /* 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) {
log_warning("kexec failed. Falling back to normal reboot.");
} else {
/* Child */
- const char *args[5] = { KEXEC_BINARY_PATH, "-e", "-f", "-x", NULL };
+ const char *args[3] = { "/sbin/kexec", "-e", NULL };
execv(args[0], (char * const *) args);
return EXIT_FAILURE;
}
r = -errno;
error:
- sync();
log_error("Critical error while doing system shutdown: %s", strerror(-r));
freeze();