]>
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 | |
9 | under the terms of the GNU General Public License as published by | |
10 | the Free Software Foundation; either version 2 of the License, or | |
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 | |
16 | General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
19 | along with systemd; If not, see <http://www.gnu.org/licenses/>. | |
20 | ***/ | |
21 | ||
22 | #include <sys/mman.h> | |
23 | #include <sys/types.h> | |
24 | #include <sys/reboot.h> | |
25 | #include <linux/reboot.h> | |
26 | #include <sys/wait.h> | |
7cb1094a HH |
27 | #include <sys/types.h> |
28 | #include <sys/stat.h> | |
29 | #include <sys/mount.h> | |
30 | #include <sys/syscall.h> | |
31 | #include <fcntl.h> | |
b1b2a107 FF |
32 | #include <dirent.h> |
33 | #include <errno.h> | |
34 | #include <unistd.h> | |
35 | #include <signal.h> | |
36 | #include <stdbool.h> | |
37 | #include <stdlib.h> | |
38 | #include <string.h> | |
39 | ||
7cb1094a | 40 | #include "missing.h" |
b1b2a107 FF |
41 | #include "log.h" |
42 | #include "umount.h" | |
43 | #include "util.h" | |
b52aae1d | 44 | #include "virt.h" |
b1b2a107 | 45 | |
567ea02a | 46 | #define TIMEOUT_USEC (5 * USEC_PER_SEC) |
b1b2a107 | 47 | #define FINALIZE_ATTEMPTS 50 |
b1b2a107 | 48 | |
b1b2a107 | 49 | static bool ignore_proc(pid_t pid) { |
7e4ab3c5 LP |
50 | char buf[PATH_MAX]; |
51 | FILE *f; | |
52 | char c; | |
53 | size_t count; | |
54 | uid_t uid; | |
55 | int r; | |
56 | ||
57 | /* We are PID 1, let's not commit suicide */ | |
b1b2a107 FF |
58 | if (pid == 1) |
59 | return true; | |
60 | ||
7e4ab3c5 LP |
61 | r = get_process_uid(pid, &uid); |
62 | if (r < 0) | |
63 | return true; /* not really, but better safe than sorry */ | |
b1b2a107 | 64 | |
7e4ab3c5 LP |
65 | /* Non-root processes otherwise are always subject to be killed */ |
66 | if (uid != 0) | |
67 | return false; | |
b1b2a107 | 68 | |
7e4ab3c5 LP |
69 | snprintf(buf, sizeof(buf), "/proc/%lu/cmdline", (unsigned long) pid); |
70 | char_array_0(buf); | |
b1b2a107 | 71 | |
b1b2a107 FF |
72 | f = fopen(buf, "re"); |
73 | if (!f) | |
74 | return true; /* not really, but has the desired effect */ | |
75 | ||
76 | count = fread(&c, 1, 1, f); | |
77 | fclose(f); | |
7e4ab3c5 LP |
78 | |
79 | /* Kernel threads have an empty cmdline */ | |
80 | if (count <= 0) | |
81 | return true; | |
82 | ||
83 | /* Processes with argv[0][0] = '@' we ignore from the killing | |
bd1a6981 LP |
84 | * spree. |
85 | * | |
86 | * http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons */ | |
7e4ab3c5 LP |
87 | if (count == 1 && c == '@') |
88 | return true; | |
89 | ||
90 | return false; | |
b1b2a107 FF |
91 | } |
92 | ||
93 | static int killall(int sign) { | |
94 | DIR *dir; | |
95 | struct dirent *d; | |
12aad1d0 | 96 | unsigned int n_processes = 0; |
b1b2a107 | 97 | |
7e4ab3c5 LP |
98 | dir = opendir("/proc"); |
99 | if (!dir) | |
b1b2a107 FF |
100 | return -errno; |
101 | ||
102 | while ((d = readdir(dir))) { | |
103 | pid_t pid; | |
104 | ||
105 | if (parse_pid(d->d_name, &pid) < 0) | |
106 | continue; | |
107 | ||
b1b2a107 FF |
108 | if (ignore_proc(pid)) |
109 | continue; | |
110 | ||
111 | if (kill(pid, sign) == 0) | |
12aad1d0 | 112 | n_processes++; |
b1b2a107 FF |
113 | else |
114 | log_warning("Could not kill %d: %m", pid); | |
115 | } | |
116 | ||
117 | closedir(dir); | |
118 | ||
12aad1d0 | 119 | return n_processes; |
b1b2a107 FF |
120 | } |
121 | ||
40e85d00 | 122 | static void wait_for_children(int n_processes, sigset_t *mask) { |
b1b2a107 | 123 | usec_t until; |
b1b2a107 | 124 | |
40e85d00 | 125 | assert(mask); |
b1b2a107 FF |
126 | |
127 | until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC; | |
128 | for (;;) { | |
40e85d00 | 129 | struct timespec ts; |
ab58e291 | 130 | int k; |
40e85d00 | 131 | usec_t n; |
ab58e291 | 132 | |
b1b2a107 FF |
133 | for (;;) { |
134 | pid_t pid = waitpid(-1, NULL, WNOHANG); | |
12aad1d0 | 135 | |
b1b2a107 FF |
136 | if (pid == 0) |
137 | break; | |
b1b2a107 | 138 | |
40e85d00 LP |
139 | if (pid < 0 && errno == ECHILD) |
140 | return; | |
141 | ||
142 | if (n_processes > 0) | |
143 | if (--n_processes == 0) | |
144 | return; | |
b1b2a107 FF |
145 | } |
146 | ||
40e85d00 | 147 | n = now(CLOCK_MONOTONIC); |
b1b2a107 | 148 | if (n >= until) |
40e85d00 | 149 | return; |
b1b2a107 FF |
150 | |
151 | timespec_store(&ts, until - n); | |
40e85d00 LP |
152 | |
153 | if ((k = sigtimedwait(mask, NULL, &ts)) != SIGCHLD) { | |
154 | ||
155 | if (k < 0 && errno != EAGAIN) { | |
156 | log_error("sigtimedwait() failed: %m"); | |
157 | return; | |
158 | } | |
159 | ||
ab58e291 LP |
160 | if (k >= 0) |
161 | log_warning("sigtimedwait() returned unexpected signal."); | |
ab58e291 | 162 | } |
b1b2a107 | 163 | } |
40e85d00 LP |
164 | } |
165 | ||
166 | static void send_signal(int sign) { | |
167 | sigset_t mask, oldmask; | |
168 | int n_processes; | |
169 | ||
170 | assert_se(sigemptyset(&mask) == 0); | |
171 | assert_se(sigaddset(&mask, SIGCHLD) == 0); | |
172 | assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0); | |
173 | ||
174 | if (kill(-1, SIGSTOP) < 0 && errno != ESRCH) | |
175 | log_warning("kill(-1, SIGSTOP) failed: %m"); | |
176 | ||
177 | n_processes = killall(sign); | |
178 | ||
179 | if (kill(-1, SIGCONT) < 0 && errno != ESRCH) | |
180 | log_warning("kill(-1, SIGCONT) failed: %m"); | |
181 | ||
182 | if (n_processes <= 0) | |
183 | goto finish; | |
184 | ||
185 | wait_for_children(n_processes, &mask); | |
b1b2a107 FF |
186 | |
187 | finish: | |
188 | sigprocmask(SIG_SETMASK, &oldmask, NULL); | |
b1b2a107 FF |
189 | } |
190 | ||
40e85d00 | 191 | static void ultimate_send_signal(int sign) { |
b1b2a107 | 192 | sigset_t mask, oldmask; |
b1b2a107 FF |
193 | int r; |
194 | ||
40e85d00 LP |
195 | assert_se(sigemptyset(&mask) == 0); |
196 | assert_se(sigaddset(&mask, SIGCHLD) == 0); | |
197 | assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0); | |
b1b2a107 | 198 | |
40e85d00 | 199 | if (kill(-1, SIGSTOP) < 0 && errno != ESRCH) |
ab58e291 | 200 | log_warning("kill(-1, SIGSTOP) failed: %m"); |
b1b2a107 FF |
201 | |
202 | r = kill(-1, sign); | |
40e85d00 LP |
203 | if (r < 0 && errno != ESRCH) |
204 | log_warning("kill(-1, %s) failed: %m", signal_to_string(sign)); | |
b1b2a107 | 205 | |
40e85d00 | 206 | if (kill(-1, SIGCONT) < 0 && errno != ESRCH) |
ab58e291 | 207 | log_warning("kill(-1, SIGCONT) failed: %m"); |
b1b2a107 FF |
208 | |
209 | if (r < 0) | |
210 | goto finish; | |
211 | ||
40e85d00 | 212 | wait_for_children(0, &mask); |
b1b2a107 FF |
213 | |
214 | finish: | |
215 | sigprocmask(SIG_SETMASK, &oldmask, NULL); | |
b1b2a107 FF |
216 | } |
217 | ||
89d471d5 LP |
218 | static int prepare_new_root(void) { |
219 | static const char dirs[] = | |
220 | "/run/initramfs/oldroot\0" | |
221 | "/run/initramfs/proc\0" | |
222 | "/run/initramfs/sys\0" | |
223 | "/run/initramfs/dev\0" | |
224 | "/run/initramfs/run\0"; | |
225 | ||
226 | const char *dir; | |
227 | ||
228 | if (mount("/run/initramfs", "/run/initramfs", NULL, MS_BIND, NULL) < 0) { | |
229 | log_error("Failed to mount bind /run/initramfs on /run/initramfs: %m"); | |
230 | return -errno; | |
231 | } | |
232 | ||
233 | if (mount(NULL, "/run/initramfs", NULL, MS_PRIVATE, NULL) < 0) { | |
234 | log_error("Failed to make /run/initramfs private mount: %m"); | |
235 | return -errno; | |
236 | } | |
237 | ||
238 | NULSTR_FOREACH(dir, dirs) | |
239 | if (mkdir_p(dir, 0755) < 0 && errno != EEXIST) { | |
240 | log_error("Failed to mkdir %s: %m", dir); | |
241 | return -errno; | |
7cb1094a | 242 | } |
89d471d5 LP |
243 | |
244 | if (mount("/sys", "/run/initramfs/sys", NULL, MS_BIND, NULL) < 0) { | |
245 | log_error("Failed to mount bind /sys on /run/initramfs/sys: %m"); | |
246 | return -errno; | |
247 | } | |
248 | ||
249 | if (mount("/proc", "/run/initramfs/proc", NULL, MS_BIND, NULL) < 0) { | |
250 | log_error("Failed to mount bind /proc on /run/initramfs/proc: %m"); | |
251 | return -errno; | |
252 | } | |
253 | ||
254 | if (mount("/dev", "/run/initramfs/dev", NULL, MS_BIND, NULL) < 0) { | |
255 | log_error("Failed to mount bind /dev on /run/initramfs/dev: %m"); | |
256 | return -errno; | |
7cb1094a HH |
257 | } |
258 | ||
89d471d5 LP |
259 | if (mount("/run", "/run/initramfs/run", NULL, MS_BIND, NULL) < 0) { |
260 | log_error("Failed to mount bind /run on /run/initramfs/run: %m"); | |
261 | return -errno; | |
262 | } | |
263 | ||
264 | return 0; | |
7cb1094a HH |
265 | } |
266 | ||
89d471d5 | 267 | static int pivot_to_new_root(void) { |
7cb1094a | 268 | int fd; |
89d471d5 | 269 | |
7cb1094a HH |
270 | chdir("/run/initramfs"); |
271 | ||
272 | /* | |
273 | In case some evil process made "/" MS_SHARED | |
274 | It works for pivot_root, but the ref count for the root device | |
275 | is not decreasing :-/ | |
276 | */ | |
89d471d5 LP |
277 | if (mount(NULL, "/", NULL, MS_PRIVATE, NULL) < 0) { |
278 | log_error("Failed to make \"/\" private mount %m"); | |
279 | return -errno; | |
7cb1094a HH |
280 | } |
281 | ||
89d471d5 | 282 | if (pivot_root(".", "oldroot") < 0) { |
7cb1094a | 283 | log_error("pivot failed: %m"); |
89d471d5 LP |
284 | /* only chroot if pivot root succeded */ |
285 | return -errno; | |
7cb1094a | 286 | } |
89d471d5 | 287 | |
7cb1094a | 288 | chroot("."); |
89d471d5 LP |
289 | log_info("Successfully changed into root pivot."); |
290 | ||
291 | fd = open("/dev/console", O_RDWR); | |
292 | if (fd < 0) | |
293 | log_error("Failed to open /dev/console: %m"); | |
294 | else { | |
295 | make_stdio(fd); | |
bccc1d88 LP |
296 | |
297 | /* Initialize the controlling terminal */ | |
298 | setsid(); | |
299 | ioctl(STDIN_FILENO, TIOCSCTTY, NULL); | |
89d471d5 LP |
300 | } |
301 | ||
302 | return 0; | |
7cb1094a HH |
303 | } |
304 | ||
b1b2a107 | 305 | int main(int argc, char *argv[]) { |
12aad1d0 LP |
306 | int cmd, r; |
307 | unsigned retries; | |
d48141ba | 308 | bool need_umount = true, need_swapoff = true, need_loop_detach = true, need_dm_detach = true; |
40e85d00 | 309 | bool killed_everbody = false, in_container; |
b1b2a107 FF |
310 | |
311 | log_parse_environment(); | |
2ca1b422 | 312 | log_set_target(LOG_TARGET_CONSOLE); /* syslog will die if not gone yet */ |
b1b2a107 FF |
313 | log_open(); |
314 | ||
4c12626c LP |
315 | umask(0022); |
316 | ||
b1b2a107 | 317 | if (getpid() != 1) { |
567ea02a | 318 | log_error("Not executed by init (pid 1)."); |
b1b2a107 FF |
319 | r = -EPERM; |
320 | goto error; | |
321 | } | |
322 | ||
323 | if (argc != 2) { | |
324 | log_error("Invalid number of arguments."); | |
325 | r = -EINVAL; | |
326 | goto error; | |
327 | } | |
328 | ||
40e85d00 LP |
329 | in_container = detect_container(NULL) > 0; |
330 | ||
b1b2a107 FF |
331 | if (streq(argv[1], "reboot")) |
332 | cmd = RB_AUTOBOOT; | |
333 | else if (streq(argv[1], "poweroff")) | |
334 | cmd = RB_POWER_OFF; | |
335 | else if (streq(argv[1], "halt")) | |
336 | cmd = RB_HALT_SYSTEM; | |
337 | else if (streq(argv[1], "kexec")) | |
338 | cmd = LINUX_REBOOT_CMD_KEXEC; | |
339 | else { | |
340 | log_error("Unknown action '%s'.", argv[1]); | |
341 | r = -EINVAL; | |
342 | goto error; | |
343 | } | |
344 | ||
345 | /* lock us into memory */ | |
346 | if (mlockall(MCL_CURRENT|MCL_FUTURE) != 0) | |
347 | log_warning("Cannot lock process memory: %m"); | |
348 | ||
ab58e291 | 349 | log_info("Sending SIGTERM to remaining processes..."); |
40e85d00 | 350 | send_signal(SIGTERM); |
b1b2a107 | 351 | |
ab58e291 | 352 | log_info("Sending SIGKILL to remaining processes..."); |
40e85d00 LP |
353 | send_signal(SIGKILL); |
354 | ||
355 | if (in_container) | |
356 | need_swapoff = false; | |
b1b2a107 | 357 | |
567ea02a | 358 | /* Unmount all mountpoints, swaps, and loopback devices */ |
12aad1d0 LP |
359 | for (retries = 0; retries < FINALIZE_ATTEMPTS; retries++) { |
360 | bool changed = false; | |
361 | ||
b1b2a107 | 362 | if (need_umount) { |
ab58e291 | 363 | log_info("Unmounting file systems."); |
12aad1d0 | 364 | r = umount_all(&changed); |
b1b2a107 FF |
365 | if (r == 0) |
366 | need_umount = false; | |
367 | else if (r > 0) | |
ab58e291 | 368 | log_info("Not all file systems unmounted, %d left.", r); |
b1b2a107 | 369 | else |
ab58e291 | 370 | log_error("Failed to unmount file systems: %s", strerror(-r)); |
b1b2a107 FF |
371 | } |
372 | ||
373 | if (need_swapoff) { | |
374 | log_info("Disabling swaps."); | |
12aad1d0 | 375 | r = swapoff_all(&changed); |
b1b2a107 FF |
376 | if (r == 0) |
377 | need_swapoff = false; | |
378 | else if (r > 0) | |
ab58e291 | 379 | log_info("Not all swaps are turned off, %d left.", r); |
b1b2a107 | 380 | else |
ab58e291 | 381 | log_error("Failed to turn off swaps: %s", strerror(-r)); |
b1b2a107 FF |
382 | } |
383 | ||
384 | if (need_loop_detach) { | |
385 | log_info("Detaching loop devices."); | |
12aad1d0 | 386 | r = loopback_detach_all(&changed); |
b1b2a107 FF |
387 | if (r == 0) |
388 | need_loop_detach = false; | |
389 | else if (r > 0) | |
ab58e291 | 390 | log_info("Not all loop devices detached, %d left.", r); |
b1b2a107 | 391 | else |
ab58e291 | 392 | log_error("Failed to detach loop devices: %s", strerror(-r)); |
d48141ba | 393 | } |
b1b2a107 | 394 | |
d48141ba LP |
395 | if (need_dm_detach) { |
396 | log_info("Detaching DM devices."); | |
12aad1d0 | 397 | r = dm_detach_all(&changed); |
d48141ba LP |
398 | if (r == 0) |
399 | need_dm_detach = false; | |
400 | else if (r > 0) | |
ab58e291 | 401 | log_warning("Not all DM devices detached, %d left.", r); |
d48141ba | 402 | else |
ab58e291 | 403 | log_error("Failed to detach DM devices: %s", strerror(-r)); |
b1b2a107 FF |
404 | } |
405 | ||
a27d2184 KS |
406 | if (!need_umount && !need_swapoff && !need_loop_detach && !need_dm_detach) { |
407 | if (retries > 0) | |
408 | log_info("All filesystems, swaps, loop devices, DM devices detached."); | |
12aad1d0 LP |
409 | /* Yay, done */ |
410 | break; | |
a27d2184 | 411 | } |
b1b2a107 | 412 | |
12aad1d0 LP |
413 | /* If in this iteration we didn't manage to |
414 | * unmount/deactivate anything, we either kill more | |
415 | * processes, or simply give up */ | |
416 | if (!changed) { | |
b1b2a107 | 417 | |
12aad1d0 LP |
418 | if (killed_everbody) { |
419 | /* Hmm, we already killed everybody, | |
420 | * let's just give up */ | |
ab58e291 | 421 | log_error("Cannot finalize remaining file systems and devices, giving up."); |
d37fb98b | 422 | break; |
b1b2a107 | 423 | } |
12aad1d0 | 424 | |
ab58e291 | 425 | log_warning("Cannot finalize remaining file systems and devices, trying to kill remaining processes."); |
40e85d00 LP |
426 | ultimate_send_signal(SIGTERM); |
427 | ultimate_send_signal(SIGKILL); | |
12aad1d0 LP |
428 | killed_everbody = true; |
429 | } | |
430 | ||
ab58e291 | 431 | log_debug("Couldn't finalize remaining file systems and devices after %u retries, trying again.", retries+1); |
b1b2a107 FF |
432 | } |
433 | ||
12aad1d0 | 434 | if (retries >= FINALIZE_ATTEMPTS) |
35b8ca3a | 435 | log_error("Too many iterations, giving up."); |
12aad1d0 | 436 | |
83cc030f LP |
437 | execute_directory(SYSTEM_SHUTDOWN_PATH, NULL, NULL); |
438 | ||
40e85d00 LP |
439 | /* If we are in a container, just exit, this will kill our |
440 | * container for good. */ | |
f41de959 LP |
441 | if (in_container) { |
442 | log_error("Exiting container."); | |
40e85d00 | 443 | exit(0); |
f41de959 | 444 | } |
40e85d00 | 445 | |
7cb1094a | 446 | if (access("/run/initramfs/shutdown", X_OK) == 0) { |
89d471d5 LP |
447 | |
448 | if (prepare_new_root() >= 0 && | |
449 | pivot_to_new_root() >= 0) { | |
450 | execv("/shutdown", argv); | |
7cb1094a HH |
451 | log_error("Failed to execute shutdown binary: %m"); |
452 | } | |
453 | } | |
454 | ||
89d471d5 LP |
455 | sync(); |
456 | ||
b1b2a107 | 457 | if (cmd == LINUX_REBOOT_CMD_KEXEC) { |
12aad1d0 | 458 | /* We cheat and exec kexec to avoid doing all its work */ |
b1b2a107 | 459 | pid_t pid = fork(); |
12aad1d0 | 460 | |
e61cd186 LP |
461 | if (pid < 0) |
462 | log_error("Could not fork: %m. Falling back to normal reboot."); | |
463 | else if (pid > 0) { | |
464 | wait_for_terminate_and_warn("kexec", pid); | |
465 | log_warning("kexec failed. Falling back to normal reboot."); | |
b1b2a107 | 466 | } else { |
e61cd186 | 467 | /* Child */ |
f8d0ceb4 | 468 | const char *args[3] = { "/sbin/kexec", "-e", NULL }; |
b1b2a107 FF |
469 | execv(args[0], (char * const *) args); |
470 | return EXIT_FAILURE; | |
471 | } | |
e61cd186 LP |
472 | |
473 | cmd = RB_AUTOBOOT; | |
b1b2a107 FF |
474 | } |
475 | ||
476 | reboot(cmd); | |
e61cd186 LP |
477 | log_error("Failed to invoke reboot(): %m"); |
478 | r = -errno; | |
b1b2a107 FF |
479 | |
480 | error: | |
e61cd186 LP |
481 | log_error("Critical error while doing system shutdown: %s", strerror(-r)); |
482 | ||
b1b2a107 | 483 | freeze(); |
3c14d26c | 484 | return EXIT_FAILURE; |
b1b2a107 | 485 | } |