]>
Commit | Line | Data |
---|---|---|
b1b2a107 FF |
1 | /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/ |
2 | ||
3 | /*** | |
4 | This file is part of systemd. | |
5 | ||
6 | Copyright 2010 ProFUSION embedded systems | |
7 | ||
8 | systemd is free software; you can redistribute it and/or modify it | |
5430f7f2 LP |
9 | under the terms of the GNU Lesser General Public License as published by |
10 | the Free Software Foundation; either version 2.1 of the License, or | |
b1b2a107 FF |
11 | (at your option) any later version. |
12 | ||
13 | systemd is distributed in the hope that it will be useful, but | |
14 | WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
5430f7f2 | 16 | Lesser General Public License for more details. |
b1b2a107 | 17 | |
5430f7f2 | 18 | You should have received a copy of the GNU Lesser General Public License |
b1b2a107 FF |
19 | along with systemd; If not, see <http://www.gnu.org/licenses/>. |
20 | ***/ | |
21 | ||
22 | #include <sys/mman.h> | |
b1b2a107 FF |
23 | #include <sys/reboot.h> |
24 | #include <linux/reboot.h> | |
7cb1094a HH |
25 | #include <sys/stat.h> |
26 | #include <sys/mount.h> | |
b1b2a107 FF |
27 | #include <errno.h> |
28 | #include <unistd.h> | |
29 | #include <signal.h> | |
30 | #include <stdbool.h> | |
31 | #include <stdlib.h> | |
b1e90ec5 | 32 | #include <getopt.h> |
b1b2a107 | 33 | |
7cb1094a | 34 | #include "missing.h" |
b1b2a107 | 35 | #include "log.h" |
ec26be51 | 36 | #include "fileio.h" |
b1b2a107 FF |
37 | #include "umount.h" |
38 | #include "util.h" | |
b52aae1d | 39 | #include "virt.h" |
e96d6be7 | 40 | #include "watchdog.h" |
39d6464c | 41 | #include "killall.h" |
41f85451 | 42 | #include "cgroup-util.h" |
37185ec8 | 43 | #include "def.h" |
5a4bf02f | 44 | #include "switch-root.h" |
0b452006 | 45 | #include "process-util.h" |
b1b2a107 | 46 | |
b1b2a107 | 47 | #define FINALIZE_ATTEMPTS 50 |
b1b2a107 | 48 | |
b1e90ec5 ZJS |
49 | static char* arg_verb; |
50 | ||
51 | static int parse_argv(int argc, char *argv[]) { | |
52 | enum { | |
53 | ARG_LOG_LEVEL = 0x100, | |
54 | ARG_LOG_TARGET, | |
55 | ARG_LOG_COLOR, | |
56 | ARG_LOG_LOCATION, | |
57 | }; | |
58 | ||
59 | static const struct option options[] = { | |
60 | { "log-level", required_argument, NULL, ARG_LOG_LEVEL }, | |
61 | { "log-target", required_argument, NULL, ARG_LOG_TARGET }, | |
62 | { "log-color", optional_argument, NULL, ARG_LOG_COLOR }, | |
63 | { "log-location", optional_argument, NULL, ARG_LOG_LOCATION }, | |
64 | {} | |
65 | }; | |
66 | ||
67 | int c, r; | |
68 | ||
69 | assert(argc >= 1); | |
70 | assert(argv); | |
71 | ||
4b5d8d0f MS |
72 | /* "-" prevents getopt from permuting argv[] and moving the verb away |
73 | * from argv[1]. Our interface to initrd promises it'll be there. */ | |
74 | while ((c = getopt_long(argc, argv, "-", options, NULL)) >= 0) | |
b1e90ec5 ZJS |
75 | switch (c) { |
76 | ||
77 | case ARG_LOG_LEVEL: | |
78 | r = log_set_max_level_from_string(optarg); | |
79 | if (r < 0) | |
80 | log_error("Failed to parse log level %s, ignoring.", optarg); | |
81 | ||
82 | break; | |
83 | ||
84 | case ARG_LOG_TARGET: | |
85 | r = log_set_target_from_string(optarg); | |
86 | if (r < 0) | |
87 | log_error("Failed to parse log target %s, ignoring", optarg); | |
88 | ||
89 | break; | |
90 | ||
91 | case ARG_LOG_COLOR: | |
92 | ||
93 | if (optarg) { | |
94 | r = log_show_color_from_string(optarg); | |
95 | if (r < 0) | |
96 | log_error("Failed to parse log color setting %s, ignoring", optarg); | |
97 | } else | |
98 | log_show_color(true); | |
99 | ||
100 | break; | |
101 | ||
102 | case ARG_LOG_LOCATION: | |
103 | if (optarg) { | |
104 | r = log_show_location_from_string(optarg); | |
105 | if (r < 0) | |
106 | log_error("Failed to parse log location setting %s, ignoring", optarg); | |
107 | } else | |
108 | log_show_location(true); | |
109 | ||
110 | break; | |
111 | ||
4b5d8d0f MS |
112 | case '\001': |
113 | if (!arg_verb) | |
114 | arg_verb = optarg; | |
115 | else | |
116 | log_error("Excess arguments, ignoring"); | |
117 | break; | |
118 | ||
b1e90ec5 | 119 | case '?': |
b1e90ec5 ZJS |
120 | return -EINVAL; |
121 | ||
122 | default: | |
123 | assert_not_reached("Unhandled option code."); | |
124 | } | |
125 | ||
4b5d8d0f | 126 | if (!arg_verb) { |
b1e90ec5 ZJS |
127 | log_error("Verb argument missing."); |
128 | return -EINVAL; | |
129 | } | |
130 | ||
b1e90ec5 ZJS |
131 | return 0; |
132 | } | |
133 | ||
5a4bf02f | 134 | static int switch_root_initramfs(void) { |
4a62c710 MS |
135 | if (mount("/run/initramfs", "/run/initramfs", NULL, MS_BIND, NULL) < 0) |
136 | return log_error_errno(errno, "Failed to mount bind /run/initramfs on /run/initramfs: %m"); | |
89d471d5 | 137 | |
4a62c710 MS |
138 | if (mount(NULL, "/run/initramfs", NULL, MS_PRIVATE, NULL) < 0) |
139 | return log_error_errno(errno, "Failed to make /run/initramfs private mount: %m"); | |
89d471d5 | 140 | |
f131770b | 141 | /* switch_root with MS_BIND, because there might still be processes lurking around, which have open file descriptors. |
5a4bf02f HH |
142 | * /run/initramfs/shutdown will take care of these. |
143 | * Also do not detach the old root, because /run/initramfs/shutdown needs to access it. | |
144 | */ | |
145 | return switch_root("/run/initramfs", "/oldroot", false, MS_BIND); | |
7cb1094a HH |
146 | } |
147 | ||
7cb1094a | 148 | |
b1b2a107 | 149 | int main(int argc, char *argv[]) { |
8c977838 | 150 | bool need_umount, need_swapoff, need_loop_detach, need_dm_detach; |
bd3fa1d2 | 151 | bool in_container, use_watchdog = false; |
06beed6d | 152 | _cleanup_free_ char *cgroup = NULL; |
6edd7d0a | 153 | char *arguments[3]; |
41f85451 LP |
154 | unsigned retries; |
155 | int cmd, r; | |
e801700e | 156 | static const char* const dirs[] = {SYSTEM_SHUTDOWN_PATH, NULL}; |
b1b2a107 | 157 | |
b1e90ec5 ZJS |
158 | log_parse_environment(); |
159 | r = parse_argv(argc, argv); | |
160 | if (r < 0) | |
161 | goto error; | |
ec26be51 | 162 | |
b1e90ec5 | 163 | /* journald will die if not gone yet. The log target defaults |
3f85ef0f | 164 | * to console, but may have been changed by command line options. */ |
ec26be51 | 165 | |
99f09825 | 166 | log_close_console(); /* force reopen of /dev/console */ |
b1b2a107 FF |
167 | log_open(); |
168 | ||
4c12626c LP |
169 | umask(0022); |
170 | ||
b1b2a107 | 171 | if (getpid() != 1) { |
b1e90ec5 | 172 | log_error("Not executed by init (PID 1)."); |
b1b2a107 FF |
173 | r = -EPERM; |
174 | goto error; | |
175 | } | |
176 | ||
b1e90ec5 | 177 | if (streq(arg_verb, "reboot")) |
b1b2a107 | 178 | cmd = RB_AUTOBOOT; |
b1e90ec5 | 179 | else if (streq(arg_verb, "poweroff")) |
b1b2a107 | 180 | cmd = RB_POWER_OFF; |
b1e90ec5 | 181 | else if (streq(arg_verb, "halt")) |
b1b2a107 | 182 | cmd = RB_HALT_SYSTEM; |
b1e90ec5 | 183 | else if (streq(arg_verb, "kexec")) |
b1b2a107 FF |
184 | cmd = LINUX_REBOOT_CMD_KEXEC; |
185 | else { | |
b1b2a107 | 186 | r = -EINVAL; |
b1e90ec5 | 187 | log_error("Unknown action '%s'.", arg_verb); |
b1b2a107 FF |
188 | goto error; |
189 | } | |
190 | ||
41f85451 LP |
191 | cg_get_root_path(&cgroup); |
192 | ||
e96d6be7 LP |
193 | use_watchdog = !!getenv("WATCHDOG_USEC"); |
194 | ||
b1b2a107 | 195 | /* lock us into memory */ |
b55d0651 | 196 | mlockall(MCL_CURRENT|MCL_FUTURE); |
b1b2a107 | 197 | |
ab58e291 | 198 | log_info("Sending SIGTERM to remaining processes..."); |
0bee65f0 | 199 | broadcast_signal(SIGTERM, true, true); |
b1b2a107 | 200 | |
ab58e291 | 201 | log_info("Sending SIGKILL to remaining processes..."); |
0bee65f0 | 202 | broadcast_signal(SIGKILL, true, false); |
40e85d00 | 203 | |
8c977838 ZJS |
204 | in_container = detect_container(NULL) > 0; |
205 | ||
d89b5fed | 206 | need_umount = !in_container; |
8c977838 ZJS |
207 | need_swapoff = !in_container; |
208 | need_loop_detach = !in_container; | |
209 | need_dm_detach = !in_container; | |
b1b2a107 | 210 | |
567ea02a | 211 | /* Unmount all mountpoints, swaps, and loopback devices */ |
12aad1d0 LP |
212 | for (retries = 0; retries < FINALIZE_ATTEMPTS; retries++) { |
213 | bool changed = false; | |
214 | ||
e96d6be7 LP |
215 | if (use_watchdog) |
216 | watchdog_ping(); | |
217 | ||
41f85451 LP |
218 | /* Let's trim the cgroup tree on each iteration so |
219 | that we leave an empty cgroup tree around, so that | |
220 | container managers get a nice notify event when we | |
221 | are down */ | |
222 | if (cgroup) | |
223 | cg_trim(SYSTEMD_CGROUP_CONTROLLER, cgroup, false); | |
224 | ||
b1b2a107 | 225 | if (need_umount) { |
ab58e291 | 226 | log_info("Unmounting file systems."); |
12aad1d0 | 227 | r = umount_all(&changed); |
bce93b7a | 228 | if (r == 0) { |
b1b2a107 | 229 | need_umount = false; |
bce93b7a MS |
230 | log_info("All filesystems unmounted."); |
231 | } else if (r > 0) | |
ab58e291 | 232 | log_info("Not all file systems unmounted, %d left.", r); |
b1b2a107 | 233 | else |
da927ba9 | 234 | log_error_errno(r, "Failed to unmount file systems: %m"); |
b1b2a107 FF |
235 | } |
236 | ||
237 | if (need_swapoff) { | |
735e0712 | 238 | log_info("Deactivating swaps."); |
12aad1d0 | 239 | r = swapoff_all(&changed); |
bce93b7a | 240 | if (r == 0) { |
b1b2a107 | 241 | need_swapoff = false; |
735e0712 | 242 | log_info("All swaps deactivated."); |
bce93b7a | 243 | } else if (r > 0) |
735e0712 | 244 | log_info("Not all swaps deactivated, %d left.", r); |
b1b2a107 | 245 | else |
da927ba9 | 246 | log_error_errno(r, "Failed to deactivate swaps: %m"); |
b1b2a107 FF |
247 | } |
248 | ||
249 | if (need_loop_detach) { | |
250 | log_info("Detaching loop devices."); | |
12aad1d0 | 251 | r = loopback_detach_all(&changed); |
bce93b7a | 252 | if (r == 0) { |
b1b2a107 | 253 | need_loop_detach = false; |
bce93b7a MS |
254 | log_info("All loop devices detached."); |
255 | } else if (r > 0) | |
ab58e291 | 256 | log_info("Not all loop devices detached, %d left.", r); |
b1b2a107 | 257 | else |
da927ba9 | 258 | log_error_errno(r, "Failed to detach loop devices: %m"); |
d48141ba | 259 | } |
b1b2a107 | 260 | |
d48141ba LP |
261 | if (need_dm_detach) { |
262 | log_info("Detaching DM devices."); | |
12aad1d0 | 263 | r = dm_detach_all(&changed); |
bce93b7a | 264 | if (r == 0) { |
d48141ba | 265 | need_dm_detach = false; |
bce93b7a MS |
266 | log_info("All DM devices detached."); |
267 | } else if (r > 0) | |
2569a5ce | 268 | log_info("Not all DM devices detached, %d left.", r); |
d48141ba | 269 | else |
da927ba9 | 270 | log_error_errno(r, "Failed to detach DM devices: %m"); |
b1b2a107 FF |
271 | } |
272 | ||
a27d2184 KS |
273 | if (!need_umount && !need_swapoff && !need_loop_detach && !need_dm_detach) { |
274 | if (retries > 0) | |
275 | log_info("All filesystems, swaps, loop devices, DM devices detached."); | |
12aad1d0 | 276 | /* Yay, done */ |
8c977838 | 277 | goto initrd_jump; |
a27d2184 | 278 | } |
b1b2a107 | 279 | |
12aad1d0 | 280 | /* If in this iteration we didn't manage to |
bd3fa1d2 | 281 | * unmount/deactivate anything, we simply give up */ |
12aad1d0 | 282 | if (!changed) { |
8c977838 ZJS |
283 | log_info("Cannot finalize remaining%s%s%s%s continuing.", |
284 | need_umount ? " file systems," : "", | |
285 | need_swapoff ? " swap devices," : "", | |
286 | need_loop_detach ? " loop devices," : "", | |
287 | need_dm_detach ? " DM devices," : ""); | |
288 | goto initrd_jump; | |
12aad1d0 LP |
289 | } |
290 | ||
8c977838 ZJS |
291 | log_debug("After %u retries, couldn't finalize remaining %s%s%s%s trying again.", |
292 | retries + 1, | |
293 | need_umount ? " file systems," : "", | |
294 | need_swapoff ? " swap devices," : "", | |
295 | need_loop_detach ? " loop devices," : "", | |
296 | need_dm_detach ? " DM devices," : ""); | |
b1b2a107 FF |
297 | } |
298 | ||
8c977838 ZJS |
299 | log_error("Too many iterations, giving up."); |
300 | ||
301 | initrd_jump: | |
12aad1d0 | 302 | |
6edd7d0a | 303 | arguments[0] = NULL; |
b1e90ec5 | 304 | arguments[1] = arg_verb; |
6edd7d0a | 305 | arguments[2] = NULL; |
e801700e | 306 | execute_directories(dirs, DEFAULT_TIMEOUT_USEC, arguments); |
83cc030f | 307 | |
02eaa788 | 308 | if (!in_container && !in_initrd() && |
cb7ec564 | 309 | access("/run/initramfs/shutdown", X_OK) == 0) { |
5a4bf02f HH |
310 | r = switch_root_initramfs(); |
311 | if (r >= 0) { | |
a2726e5c | 312 | argv[0] = (char*) "/shutdown"; |
30d743f4 | 313 | |
5a4bf02f HH |
314 | setsid(); |
315 | make_console_stdio(); | |
316 | ||
317 | log_info("Successfully changed into root pivot.\n" | |
318 | "Returning to initrd..."); | |
30d743f4 | 319 | |
a2726e5c | 320 | execv("/shutdown", argv); |
56f64d95 | 321 | log_error_errno(errno, "Failed to execute shutdown binary: %m"); |
5a4bf02f | 322 | } else |
da927ba9 | 323 | log_error_errno(r, "Failed to switch root to \"/run/initramfs\": %m"); |
5a4bf02f | 324 | |
7cb1094a HH |
325 | } |
326 | ||
8c977838 ZJS |
327 | if (need_umount || need_swapoff || need_loop_detach || need_dm_detach) |
328 | log_error("Failed to finalize %s%s%s%s ignoring", | |
329 | need_umount ? " file systems," : "", | |
330 | need_swapoff ? " swap devices," : "", | |
331 | need_loop_detach ? " loop devices," : "", | |
332 | need_dm_detach ? " DM devices," : ""); | |
333 | ||
0049f05a LP |
334 | /* The kernel will automaticall flush ATA disks and suchlike |
335 | * on reboot(), but the file systems need to be synce'd | |
336 | * explicitly in advance. So let's do this here, but not | |
337 | * needlessly slow down containers. */ | |
338 | if (!in_container) | |
339 | sync(); | |
340 | ||
477def80 LP |
341 | switch (cmd) { |
342 | ||
343 | case LINUX_REBOOT_CMD_KEXEC: | |
cb7ec564 LP |
344 | |
345 | if (!in_container) { | |
346 | /* We cheat and exec kexec to avoid doing all its work */ | |
477def80 LP |
347 | pid_t pid; |
348 | ||
349 | log_info("Rebooting with kexec."); | |
cb7ec564 | 350 | |
477def80 | 351 | pid = fork(); |
cb7ec564 | 352 | if (pid < 0) |
56f64d95 | 353 | log_error_errno(errno, "Failed to fork: %m"); |
477def80 LP |
354 | else if (pid == 0) { |
355 | ||
356 | const char * const args[] = { | |
357 | KEXEC, "-e", NULL | |
358 | }; | |
359 | ||
cb7ec564 | 360 | /* Child */ |
477def80 | 361 | |
cb7ec564 | 362 | execv(args[0], (char * const *) args); |
477def80 LP |
363 | _exit(EXIT_FAILURE); |
364 | } else | |
820d3acf | 365 | wait_for_terminate_and_warn("kexec", pid, true); |
b1b2a107 | 366 | } |
e61cd186 LP |
367 | |
368 | cmd = RB_AUTOBOOT; | |
477def80 | 369 | /* Fall through */ |
b1b2a107 | 370 | |
477def80 LP |
371 | case RB_AUTOBOOT: |
372 | ||
373 | if (!in_container) { | |
374 | _cleanup_free_ char *param = NULL; | |
375 | ||
376 | if (read_one_line_file(REBOOT_PARAM_FILE, ¶m) >= 0) { | |
377 | log_info("Rebooting with argument '%s'.", param); | |
f07756bf | 378 | syscall(SYS_reboot, LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, LINUX_REBOOT_CMD_RESTART2, param); |
477def80 LP |
379 | } |
380 | } | |
381 | ||
382 | log_info("Rebooting."); | |
383 | break; | |
384 | ||
385 | case RB_POWER_OFF: | |
386 | log_info("Powering off."); | |
387 | break; | |
388 | ||
389 | case RB_HALT_SYSTEM: | |
390 | log_info("Halting system."); | |
391 | break; | |
392 | ||
393 | default: | |
394 | assert_not_reached("Unknown magic"); | |
395 | } | |
cb7ec564 | 396 | |
477def80 | 397 | reboot(cmd); |
cb7ec564 LP |
398 | if (errno == EPERM && in_container) { |
399 | /* If we are in a container, and we lacked | |
400 | * CAP_SYS_BOOT just exit, this will kill our | |
401 | * container for good. */ | |
477def80 | 402 | log_info("Exiting container."); |
cb7ec564 LP |
403 | exit(0); |
404 | } | |
405 | ||
56f64d95 | 406 | log_error_errno(errno, "Failed to invoke reboot(): %m"); |
e61cd186 | 407 | r = -errno; |
b1b2a107 FF |
408 | |
409 | error: | |
da927ba9 | 410 | log_emergency_errno(r, "Critical error while doing system shutdown: %m"); |
e61cd186 | 411 | |
b1b2a107 | 412 | freeze(); |
b1b2a107 | 413 | } |