]> git.ipfire.org Git - thirdparty/systemd.git/blob - src/shutdown.c
shutdown: properly handle sigtimedwait() timing out
[thirdparty/systemd.git] / src / shutdown.c
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>
27 #include <dirent.h>
28 #include <errno.h>
29 #include <unistd.h>
30 #include <signal.h>
31 #include <stdbool.h>
32 #include <stdlib.h>
33 #include <string.h>
34
35 #include "log.h"
36 #include "umount.h"
37 #include "util.h"
38
39 #define TIMEOUT_USEC (5 * USEC_PER_SEC)
40 #define FINALIZE_ATTEMPTS 50
41
42 static bool ignore_proc(pid_t pid) {
43 if (pid == 1)
44 return true;
45
46 /* TODO: add more ignore rules here: device-mapper, etc */
47
48 return false;
49 }
50
51 static bool is_kernel_thread(pid_t pid)
52 {
53 char buf[PATH_MAX];
54 FILE *f;
55 char c;
56 size_t count;
57
58 snprintf(buf, sizeof(buf), "/proc/%lu/cmdline", (unsigned long)pid);
59 f = fopen(buf, "re");
60 if (!f)
61 return true; /* not really, but has the desired effect */
62
63 count = fread(&c, 1, 1, f);
64 fclose(f);
65 return count != 1;
66 }
67
68 static int killall(int sign) {
69 DIR *dir;
70 struct dirent *d;
71 unsigned int n_processes = 0;
72
73 if ((dir = opendir("/proc")) == NULL)
74 return -errno;
75
76 while ((d = readdir(dir))) {
77 pid_t pid;
78
79 if (parse_pid(d->d_name, &pid) < 0)
80 continue;
81
82 if (is_kernel_thread(pid))
83 continue;
84
85 if (ignore_proc(pid))
86 continue;
87
88 if (kill(pid, sign) == 0)
89 n_processes++;
90 else
91 log_warning("Could not kill %d: %m", pid);
92 }
93
94 closedir(dir);
95
96 return n_processes;
97 }
98
99 static int send_signal(int sign) {
100 sigset_t mask, oldmask;
101 usec_t until;
102 int n_processes;
103 struct timespec ts;
104
105 assert_se(sigemptyset(&mask) == 0);
106 assert_se(sigaddset(&mask, SIGCHLD) == 0);
107 if (sigprocmask(SIG_BLOCK, &mask, &oldmask) != 0)
108 return -errno;
109
110 if (kill(-1, SIGSTOP) < 0)
111 log_warning("Failed kill(-1, SIGSTOP): %m");
112
113 n_processes = killall(sign);
114
115 if (kill(-1, SIGCONT) < 0)
116 log_warning("Failed kill(-1, SIGCONT): %m");
117
118 if (n_processes <= 0)
119 goto finish;
120
121 until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
122 for (;;) {
123 usec_t n = now(CLOCK_MONOTONIC);
124 for (;;) {
125 pid_t pid = waitpid(-1, NULL, WNOHANG);
126
127 if (pid == 0)
128 break;
129 else if (pid < 0 && errno == ECHILD) {
130 n_processes = 0;
131 goto finish;
132 }
133
134 if (--n_processes == 0)
135 goto finish;
136 }
137
138 if (n >= until)
139 goto finish;
140
141 timespec_store(&ts, until - n);
142 if (sigtimedwait(&mask, NULL, &ts) != SIGCHLD)
143 if (errno != EAGAIN)
144 log_warning("Failed: sigtimedwait did not return SIGCHLD: %m");
145 }
146
147 finish:
148 sigprocmask(SIG_SETMASK, &oldmask, NULL);
149
150 return n_processes;
151 }
152
153 static int rescue_send_signal(int sign) {
154 sigset_t mask, oldmask;
155 usec_t until;
156 struct timespec ts;
157 int r;
158
159 sigemptyset(&mask);
160 sigaddset(&mask, SIGCHLD);
161 if (sigprocmask(SIG_BLOCK, &mask, &oldmask) != 0)
162 return -errno;
163
164 if (kill(-1, SIGSTOP) < 0)
165 log_warning("Failed kill(-1, SIGSTOP): %m");
166
167 r = kill(-1, sign);
168 if (r < 0)
169 log_warning("Failed kill(-1, %d): %m", sign);
170
171 if (kill(-1, SIGCONT) < 0)
172 log_warning("Failed kill(-1, SIGCONT): %m");
173
174 if (r < 0)
175 goto finish;
176
177 until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
178 for (;;) {
179 usec_t n = now(CLOCK_MONOTONIC);
180 for (;;) {
181 pid_t pid = waitpid(-1, NULL, WNOHANG);
182 if (pid == 0)
183 break;
184 else if (pid < 0 && errno == ECHILD)
185 goto finish;
186 }
187
188 if (n >= until)
189 goto finish;
190
191 timespec_store(&ts, until - n);
192 if (sigtimedwait(&mask, NULL, &ts) != SIGCHLD)
193 if (errno != EAGAIN)
194 log_warning("Failed: sigtimedwait did not return SIGCHLD: %m");
195 }
196
197 finish:
198 sigprocmask(SIG_SETMASK, &oldmask, NULL);
199
200 return r;
201 }
202
203 int main(int argc, char *argv[]) {
204 int cmd, r;
205 unsigned retries;
206 bool need_umount = true, need_swapoff = true, need_loop_detach = true, need_dm_detach = true;
207 bool killed_everbody = false;
208
209 log_parse_environment();
210 log_set_target(LOG_TARGET_CONSOLE); /* syslog will die if not gone yet */
211 log_open();
212
213 if (getpid() != 1) {
214 log_error("Not executed by init (pid 1).");
215 r = -EPERM;
216 goto error;
217 }
218
219 if (argc != 2) {
220 log_error("Invalid number of arguments.");
221 r = -EINVAL;
222 goto error;
223 }
224
225 if (streq(argv[1], "reboot"))
226 cmd = RB_AUTOBOOT;
227 else if (streq(argv[1], "poweroff"))
228 cmd = RB_POWER_OFF;
229 else if (streq(argv[1], "halt"))
230 cmd = RB_HALT_SYSTEM;
231 else if (streq(argv[1], "kexec"))
232 cmd = LINUX_REBOOT_CMD_KEXEC;
233 else {
234 log_error("Unknown action '%s'.", argv[1]);
235 r = -EINVAL;
236 goto error;
237 }
238
239 /* lock us into memory */
240 if (mlockall(MCL_CURRENT|MCL_FUTURE) != 0)
241 log_warning("Cannot lock process memory: %m");
242
243 log_info("Sending SIGTERM to processes");
244 r = send_signal(SIGTERM);
245 if (r < 0)
246 log_warning("Cannot send SIGTERM to all process: %s", strerror(r));
247
248 log_info("Sending SIGKILL to processes");
249 r = send_signal(SIGKILL);
250 if (r < 0)
251 log_warning("Cannot send SIGKILL to all process: %s", strerror(r));
252
253 /* Unmount all mountpoints, swaps, and loopback devices */
254 for (retries = 0; retries < FINALIZE_ATTEMPTS; retries++) {
255 bool changed = false;
256
257 if (need_umount) {
258 log_info("Unmounting filesystems.");
259 r = umount_all(&changed);
260 if (r == 0)
261 need_umount = false;
262 else if (r > 0)
263 log_warning("Not all filesystems unmounted, %d left.", r);
264 else
265 log_error("Error unmounting filesystems: %s", strerror(-r));
266 }
267
268 if (need_swapoff) {
269 log_info("Disabling swaps.");
270 r = swapoff_all(&changed);
271 if (r == 0)
272 need_swapoff = false;
273 else if (r > 0)
274 log_warning("Not all swaps are off, %d left.", r);
275 else
276 log_error("Error turning off swaps: %s", strerror(-r));
277 }
278
279 if (need_loop_detach) {
280 log_info("Detaching loop devices.");
281 r = loopback_detach_all(&changed);
282 if (r == 0)
283 need_loop_detach = false;
284 else if (r > 0)
285 log_warning("Not all loop devices detached, %d left.", r);
286 else
287 log_error("Error detaching loop devices: %s", strerror(-r));
288 }
289
290 if (need_dm_detach) {
291 log_info("Detaching DM devices.");
292 r = dm_detach_all(&changed);
293 if (r == 0)
294 need_dm_detach = false;
295 else if (r > 0)
296 log_warning("Not all dm devices detached, %d left.", r);
297 else
298 log_error("Error detaching dm devices: %s", strerror(-r));
299 }
300
301 if (!need_umount && !need_swapoff && !need_loop_detach && !need_dm_detach)
302 /* Yay, done */
303 break;
304
305 /* If in this iteration we didn't manage to
306 * unmount/deactivate anything, we either kill more
307 * processes, or simply give up */
308 if (!changed) {
309
310 if (killed_everbody) {
311 /* Hmm, we already killed everybody,
312 * let's just give up */
313 log_error("Cannot finalize all filesystems and devices, giving up.");
314 break;
315 }
316
317 log_warning("Cannot finalize filesystems and devices, trying to kill remaining processes.");
318 rescue_send_signal(SIGTERM);
319 rescue_send_signal(SIGKILL);
320 killed_everbody = true;
321 }
322
323 log_debug("Couldn't finalize filesystems and devices after %u retries, trying again.", retries+1);
324 }
325
326 if (retries >= FINALIZE_ATTEMPTS)
327 log_error("Too many interations, giving up.");
328
329 sync();
330
331 if (cmd == LINUX_REBOOT_CMD_KEXEC) {
332 /* We cheat and exec kexec to avoid doing all its work */
333 pid_t pid = fork();
334
335 if (pid < 0)
336 log_error("Could not fork: %m. Falling back to normal reboot.");
337 else if (pid > 0) {
338 wait_for_terminate_and_warn("kexec", pid);
339 log_warning("kexec failed. Falling back to normal reboot.");
340 } else {
341 /* Child */
342 const char *args[5] = { KEXEC_BINARY_PATH, "-e", "-f", "-x", NULL };
343 execv(args[0], (char * const *) args);
344 return EXIT_FAILURE;
345 }
346
347 cmd = RB_AUTOBOOT;
348 }
349
350 reboot(cmd);
351 log_error("Failed to invoke reboot(): %m");
352 r = -errno;
353
354 error:
355 sync();
356 log_error("Critical error while doing system shutdown: %s", strerror(-r));
357
358 freeze();
359 return EXIT_FAILURE;
360 }