1 /* SPDX-License-Identifier: LGPL-2.1+ */
3 This file is part of systemd.
5 Copyright 2010 ProFUSION embedded systems
7 systemd is free software; you can redistribute it and/or modify it
8 under the terms of the GNU Lesser General Public License as published by
9 the Free Software Foundation; either version 2.1 of the License, or
10 (at your option) any later version.
12 systemd is distributed in the hope that it will be useful, but
13 WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public License
18 along with systemd; If not, see <http://www.gnu.org/licenses/>.
23 #include <linux/reboot.h>
28 #include <sys/mount.h>
29 #include <sys/reboot.h>
33 #include "alloc-util.h"
34 #include "cgroup-util.h"
37 #include "exec-util.h"
42 #include "parse-util.h"
43 #include "process-util.h"
44 #include "signal-util.h"
45 #include "string-util.h"
46 #include "switch-root.h"
47 #include "terminal-util.h"
53 #define FINALIZE_ATTEMPTS 50
55 #define SYNC_PROGRESS_ATTEMPTS 3
56 #define SYNC_TIMEOUT_USEC (10*USEC_PER_SEC)
58 static char* arg_verb
;
59 static uint8_t arg_exit_code
;
61 static int parse_argv(int argc
, char *argv
[]) {
63 ARG_LOG_LEVEL
= 0x100,
70 static const struct option options
[] = {
71 { "log-level", required_argument
, NULL
, ARG_LOG_LEVEL
},
72 { "log-target", required_argument
, NULL
, ARG_LOG_TARGET
},
73 { "log-color", optional_argument
, NULL
, ARG_LOG_COLOR
},
74 { "log-location", optional_argument
, NULL
, ARG_LOG_LOCATION
},
75 { "exit-code", required_argument
, NULL
, ARG_EXIT_CODE
},
84 /* "-" prevents getopt from permuting argv[] and moving the verb away
85 * from argv[1]. Our interface to initrd promises it'll be there. */
86 while ((c
= getopt_long(argc
, argv
, "-", options
, NULL
)) >= 0)
90 r
= log_set_max_level_from_string(optarg
);
92 log_error("Failed to parse log level %s, ignoring.", optarg
);
97 r
= log_set_target_from_string(optarg
);
99 log_error("Failed to parse log target %s, ignoring", optarg
);
106 r
= log_show_color_from_string(optarg
);
108 log_error("Failed to parse log color setting %s, ignoring", optarg
);
110 log_show_color(true);
114 case ARG_LOG_LOCATION
:
116 r
= log_show_location_from_string(optarg
);
118 log_error("Failed to parse log location setting %s, ignoring", optarg
);
120 log_show_location(true);
125 r
= safe_atou8(optarg
, &arg_exit_code
);
127 log_error("Failed to parse exit code %s, ignoring", optarg
);
135 log_error("Excess arguments, ignoring");
142 assert_not_reached("Unhandled option code.");
146 log_error("Verb argument missing.");
153 static int switch_root_initramfs(void) {
154 if (mount("/run/initramfs", "/run/initramfs", NULL
, MS_BIND
, NULL
) < 0)
155 return log_error_errno(errno
, "Failed to mount bind /run/initramfs on /run/initramfs: %m");
157 if (mount(NULL
, "/run/initramfs", NULL
, MS_PRIVATE
, NULL
) < 0)
158 return log_error_errno(errno
, "Failed to make /run/initramfs private mount: %m");
160 /* switch_root with MS_BIND, because there might still be processes lurking around, which have open file descriptors.
161 * /run/initramfs/shutdown will take care of these.
162 * Also do not detach the old root, because /run/initramfs/shutdown needs to access it.
164 return switch_root("/run/initramfs", "/oldroot", false, MS_BIND
);
167 /* Read the following fields from /proc/meminfo:
173 * Return true if the sum of these fields is greater than the previous
174 * value input. For all other issues, report the failure and indicate that
175 * the sync is not making progress.
177 static bool sync_making_progress(unsigned long long *prev_dirty
) {
178 _cleanup_fclose_
FILE *f
= NULL
;
181 unsigned long long val
= 0;
183 f
= fopen("/proc/meminfo", "re");
185 return log_warning_errno(errno
, "Failed to open /proc/meminfo: %m");
187 FOREACH_LINE(line
, f
, log_warning_errno(errno
, "Failed to parse /proc/meminfo: %m")) {
188 unsigned long long ull
= 0;
190 if (!first_word(line
, "NFS_Unstable:") && !first_word(line
, "Writeback:") && !first_word(line
, "Dirty:"))
194 if (sscanf(line
, "%*s %llu %*s", &ull
) != 1) {
196 log_warning_errno(errno
, "Failed to parse /proc/meminfo: %m");
198 log_warning("Failed to parse /proc/meminfo");
206 r
= *prev_dirty
> val
;
213 static void sync_with_progress(void) {
217 unsigned long long dirty
= ULONG_LONG_MAX
;
219 BLOCK_SIGNALS(SIGCHLD
);
221 /* Due to the possiblity of the sync operation hanging, we fork
222 * a child process and monitor the progress. If the timeout
223 * lapses, the assumption is that that particular sync stalled. */
224 r
= safe_fork("(sd-sync)", FORK_RESET_SIGNALS
|FORK_CLOSE_ALL_FDS
, &pid
);
226 log_error_errno(r
, "Failed to fork: %m");
230 /* Start the sync operation here in the child */
235 log_info("Syncing filesystems and block devices.");
237 /* Start monitoring the sync operation. If more than
238 * SYNC_PROGRESS_ATTEMPTS lapse without progress being made,
239 * we assume that the sync is stalled */
240 for (checks
= 0; checks
< SYNC_PROGRESS_ATTEMPTS
; checks
++) {
241 r
= wait_for_terminate_with_timeout(pid
, SYNC_TIMEOUT_USEC
);
243 /* Sync finished without error.
244 * (The sync itself does not return an error code) */
246 else if (r
== -ETIMEDOUT
) {
247 /* Reset the check counter if the "Dirty" value is
249 if (sync_making_progress(&dirty
))
252 log_error_errno(r
, "Failed to sync filesystems and block devices: %m");
257 /* Only reached in the event of a timeout. We should issue a kill
258 * to the stray process. */
259 log_error("Syncing filesystems and block devices - timed out, issuing SIGKILL to PID "PID_FMT
".", pid
);
260 (void) kill(pid
, SIGKILL
);
263 int main(int argc
, char *argv
[]) {
264 bool need_umount
, need_swapoff
, need_loop_detach
, need_dm_detach
;
265 bool in_container
, use_watchdog
= false;
266 _cleanup_free_
char *cgroup
= NULL
;
270 static const char* const dirs
[] = {SYSTEM_SHUTDOWN_PATH
, NULL
};
271 char *watchdog_device
;
273 log_parse_environment();
274 r
= parse_argv(argc
, argv
);
278 /* journald will die if not gone yet. The log target defaults
279 * to console, but may have been changed by command line options. */
281 log_close_console(); /* force reopen of /dev/console */
286 if (getpid_cached() != 1) {
287 log_error("Not executed by init (PID 1).");
292 if (streq(arg_verb
, "reboot"))
294 else if (streq(arg_verb
, "poweroff"))
296 else if (streq(arg_verb
, "halt"))
297 cmd
= RB_HALT_SYSTEM
;
298 else if (streq(arg_verb
, "kexec"))
299 cmd
= LINUX_REBOOT_CMD_KEXEC
;
300 else if (streq(arg_verb
, "exit"))
301 cmd
= 0; /* ignored, just checking that arg_verb is valid */
304 log_error("Unknown action '%s'.", arg_verb
);
308 (void) cg_get_root_path(&cgroup
);
309 in_container
= detect_container() > 0;
311 use_watchdog
= !!getenv("WATCHDOG_USEC");
312 watchdog_device
= getenv("WATCHDOG_DEVICE");
313 if (watchdog_device
) {
314 r
= watchdog_set_device(watchdog_device
);
316 log_warning_errno(r
, "Failed to set watchdog device to %s, ignoring: %m",
320 /* Lock us into memory */
321 mlockall(MCL_CURRENT
|MCL_FUTURE
);
323 /* Synchronize everything that is not written to disk yet at this point already. This is a good idea so that
324 * slow IO is processed here already and the final process killing spree is not impacted by processes
325 * desperately trying to sync IO to disk within their timeout. Do not remove this sync, data corruption will
328 sync_with_progress();
330 log_info("Sending SIGTERM to remaining processes...");
331 broadcast_signal(SIGTERM
, true, true);
333 log_info("Sending SIGKILL to remaining processes...");
334 broadcast_signal(SIGKILL
, true, false);
336 need_umount
= !in_container
;
337 need_swapoff
= !in_container
;
338 need_loop_detach
= !in_container
;
339 need_dm_detach
= !in_container
;
341 /* Unmount all mountpoints, swaps, and loopback devices */
342 for (retries
= 0; retries
< FINALIZE_ATTEMPTS
; retries
++) {
343 bool changed
= false;
348 /* Let's trim the cgroup tree on each iteration so
349 that we leave an empty cgroup tree around, so that
350 container managers get a nice notify event when we
353 cg_trim(SYSTEMD_CGROUP_CONTROLLER
, cgroup
, false);
356 log_info("Unmounting file systems.");
357 r
= umount_all(&changed
);
360 log_info("All filesystems unmounted.");
362 log_info("Not all file systems unmounted, %d left.", r
);
364 log_error_errno(r
, "Failed to unmount file systems: %m");
368 log_info("Deactivating swaps.");
369 r
= swapoff_all(&changed
);
371 need_swapoff
= false;
372 log_info("All swaps deactivated.");
374 log_info("Not all swaps deactivated, %d left.", r
);
376 log_error_errno(r
, "Failed to deactivate swaps: %m");
379 if (need_loop_detach
) {
380 log_info("Detaching loop devices.");
381 r
= loopback_detach_all(&changed
);
383 need_loop_detach
= false;
384 log_info("All loop devices detached.");
386 log_info("Not all loop devices detached, %d left.", r
);
388 log_error_errno(r
, "Failed to detach loop devices: %m");
391 if (need_dm_detach
) {
392 log_info("Detaching DM devices.");
393 r
= dm_detach_all(&changed
);
395 need_dm_detach
= false;
396 log_info("All DM devices detached.");
398 log_info("Not all DM devices detached, %d left.", r
);
400 log_error_errno(r
, "Failed to detach DM devices: %m");
403 if (!need_umount
&& !need_swapoff
&& !need_loop_detach
&& !need_dm_detach
) {
405 log_info("All filesystems, swaps, loop devices, DM devices detached.");
410 /* If in this iteration we didn't manage to
411 * unmount/deactivate anything, we simply give up */
413 log_info("Cannot finalize remaining%s%s%s%s continuing.",
414 need_umount
? " file systems," : "",
415 need_swapoff
? " swap devices," : "",
416 need_loop_detach
? " loop devices," : "",
417 need_dm_detach
? " DM devices," : "");
421 log_debug("After %u retries, couldn't finalize remaining %s%s%s%s trying again.",
423 need_umount
? " file systems," : "",
424 need_swapoff
? " swap devices," : "",
425 need_loop_detach
? " loop devices," : "",
426 need_dm_detach
? " DM devices," : "");
429 log_error("Too many iterations, giving up.");
433 /* We're done with the watchdog. */
434 watchdog_free_device();
437 arguments
[1] = arg_verb
;
439 execute_directories(dirs
, DEFAULT_TIMEOUT_USEC
, NULL
, NULL
, arguments
);
441 if (!in_container
&& !in_initrd() &&
442 access("/run/initramfs/shutdown", X_OK
) == 0) {
443 r
= switch_root_initramfs();
445 argv
[0] = (char*) "/shutdown";
448 make_console_stdio();
450 log_info("Successfully changed into root pivot.\n"
451 "Returning to initrd...");
453 execv("/shutdown", argv
);
454 log_error_errno(errno
, "Failed to execute shutdown binary: %m");
456 log_error_errno(r
, "Failed to switch root to \"/run/initramfs\": %m");
460 if (need_umount
|| need_swapoff
|| need_loop_detach
|| need_dm_detach
)
461 log_error("Failed to finalize %s%s%s%s ignoring",
462 need_umount
? " file systems," : "",
463 need_swapoff
? " swap devices," : "",
464 need_loop_detach
? " loop devices," : "",
465 need_dm_detach
? " DM devices," : "");
467 /* The kernel will automatically flush ATA disks and suchlike on reboot(), but the file systems need to be
468 * sync'ed explicitly in advance. So let's do this here, but not needlessly slow down containers. Note that we
469 * sync'ed things already once above, but we did some more work since then which might have caused IO, hence
470 * let's do it once more. Do not remove this sync, data corruption will result. */
472 sync_with_progress();
474 if (streq(arg_verb
, "exit")) {
478 /* We cannot exit() on the host, fallback on another
486 case LINUX_REBOOT_CMD_KEXEC
:
489 /* We cheat and exec kexec to avoid doing all its work */
492 log_info("Rebooting with kexec.");
494 r
= safe_fork("(sd-kexec)", FORK_RESET_SIGNALS
|FORK_CLOSE_ALL_FDS
, &pid
);
496 log_error_errno(r
, "Failed to fork: %m");
499 const char * const args
[] = {
505 execv(args
[0], (char * const *) args
);
509 (void) wait_for_terminate_and_warn("kexec", pid
, true);
517 _cleanup_free_
char *param
= NULL
;
519 r
= read_one_line_file("/run/systemd/reboot-param", ¶m
);
520 if (r
< 0 && r
!= -ENOENT
)
521 log_warning_errno(r
, "Failed to read reboot parameter file: %m");
523 if (!isempty(param
)) {
524 log_info("Rebooting with argument '%s'.", param
);
525 syscall(SYS_reboot
, LINUX_REBOOT_MAGIC1
, LINUX_REBOOT_MAGIC2
, LINUX_REBOOT_CMD_RESTART2
, param
);
526 log_warning_errno(errno
, "Failed to reboot with parameter, retrying without: %m");
530 log_info("Rebooting.");
534 log_info("Powering off.");
538 log_info("Halting system.");
542 assert_not_reached("Unknown magic");
546 if (errno
== EPERM
&& in_container
) {
547 /* If we are in a container, and we lacked
548 * CAP_SYS_BOOT just exit, this will kill our
549 * container for good. */
550 log_info("Exiting container.");
554 r
= log_error_errno(errno
, "Failed to invoke reboot(): %m");
557 log_emergency_errno(r
, "Critical error while doing system shutdown: %m");