]>
Commit | Line | Data |
---|---|---|
53e1b683 | 1 | /* SPDX-License-Identifier: LGPL-2.1+ */ |
a7334b09 | 2 | |
60918275 | 3 | #include <errno.h> |
3dfc9763 | 4 | #include <fcntl.h> |
f170852a | 5 | #include <getopt.h> |
97c4f35c | 6 | #include <signal.h> |
3dfc9763 LP |
7 | #include <stdio.h> |
8 | #include <string.h> | |
664f88a7 | 9 | #include <sys/mount.h> |
3dfc9763 | 10 | #include <sys/prctl.h> |
b9e74c39 | 11 | #include <sys/reboot.h> |
3dfc9763 LP |
12 | #include <sys/stat.h> |
13 | #include <unistd.h> | |
349cc4a5 | 14 | #if HAVE_SECCOMP |
b64a3d86 LP |
15 | #include <seccomp.h> |
16 | #endif | |
349cc4a5 | 17 | #if HAVE_VALGRIND_VALGRIND_H |
3dfc9763 LP |
18 | #include <valgrind/valgrind.h> |
19 | #endif | |
54b434b1 | 20 | |
718db961 | 21 | #include "sd-bus.h" |
cf0fbc49 | 22 | #include "sd-daemon.h" |
b2e7486c | 23 | #include "sd-messages.h" |
3dfc9763 | 24 | |
b5efdb8a | 25 | #include "alloc-util.h" |
d9d93745 | 26 | #include "architecture.h" |
3dfc9763 LP |
27 | #include "build.h" |
28 | #include "bus-error.h" | |
29 | #include "bus-util.h" | |
430f0182 | 30 | #include "capability-util.h" |
a88c5b8a | 31 | #include "cgroup-util.h" |
24efb112 | 32 | #include "clock-util.h" |
3dfc9763 | 33 | #include "conf-parser.h" |
618234a5 | 34 | #include "cpu-set-util.h" |
bbc1acab | 35 | #include "dbus.h" |
3dfc9763 LP |
36 | #include "dbus-manager.h" |
37 | #include "def.h" | |
eee8b7ab | 38 | #include "emergency-action.h" |
3dfc9763 | 39 | #include "env-util.h" |
57b7a260 | 40 | #include "exit-status.h" |
3ffd4af2 | 41 | #include "fd-util.h" |
3dfc9763 | 42 | #include "fdset.h" |
718db961 | 43 | #include "fileio.h" |
f97b34a6 | 44 | #include "format-util.h" |
f4f15635 | 45 | #include "fs-util.h" |
3dfc9763 LP |
46 | #include "hostname-setup.h" |
47 | #include "ima-setup.h" | |
48 | #include "killall.h" | |
49 | #include "kmod-setup.h" | |
d7b8eec7 | 50 | #include "load-fragment.h" |
3dfc9763 | 51 | #include "log.h" |
b6e66135 | 52 | #include "loopback-setup.h" |
b6e66135 | 53 | #include "machine-id-setup.h" |
3dfc9763 LP |
54 | #include "manager.h" |
55 | #include "missing.h" | |
56 | #include "mount-setup.h" | |
d58ad743 | 57 | #include "os-util.h" |
3dfc9763 | 58 | #include "pager.h" |
6bedfcbb | 59 | #include "parse-util.h" |
7d5ceb64 | 60 | #include "path-util.h" |
4e731273 | 61 | #include "proc-cmdline.h" |
3dfc9763 | 62 | #include "process-util.h" |
8869a0b4 | 63 | #include "raw-clone.h" |
78f22b97 | 64 | #include "rlimit-util.h" |
349cc4a5 | 65 | #if HAVE_SECCOMP |
83f12b27 FS |
66 | #include "seccomp-util.h" |
67 | #endif | |
b6e66135 | 68 | #include "selinux-setup.h" |
3dfc9763 LP |
69 | #include "selinux-util.h" |
70 | #include "signal-util.h" | |
ffbd2c4d | 71 | #include "smack-setup.h" |
3dfc9763 | 72 | #include "special.h" |
8fcde012 | 73 | #include "stat-util.h" |
15a5e950 | 74 | #include "stdio-util.h" |
3dfc9763 LP |
75 | #include "strv.h" |
76 | #include "switch-root.h" | |
a8b627aa | 77 | #include "sysctl-util.h" |
3dfc9763 | 78 | #include "terminal-util.h" |
8612da97 | 79 | #include "umask-util.h" |
b1d4f8e1 | 80 | #include "user-util.h" |
9ce17593 | 81 | #include "util.h" |
3dfc9763 LP |
82 | #include "virt.h" |
83 | #include "watchdog.h" | |
b6e66135 | 84 | |
f170852a LP |
85 | static enum { |
86 | ACTION_RUN, | |
e965d56d | 87 | ACTION_HELP, |
9ba0bc4e | 88 | ACTION_VERSION, |
e537352b | 89 | ACTION_TEST, |
bbc1acab YW |
90 | ACTION_DUMP_CONFIGURATION_ITEMS, |
91 | ACTION_DUMP_BUS_PROPERTIES, | |
fa0f4d8a | 92 | } arg_action = ACTION_RUN; |
fa0f4d8a | 93 | static char *arg_default_unit = NULL; |
463d0d15 | 94 | static bool arg_system = false; |
fa0f4d8a | 95 | static bool arg_dump_core = true; |
fa0f4d8a | 96 | static int arg_crash_chvt = -1; |
b9e74c39 LP |
97 | static bool arg_crash_shell = false; |
98 | static bool arg_crash_reboot = false; | |
7d5ceb64 | 99 | static char *arg_confirm_spawn = NULL; |
7a293242 | 100 | static ShowStatus arg_show_status = _SHOW_STATUS_INVALID; |
bf4df7c3 | 101 | static bool arg_switched_root = false; |
0221d68a | 102 | static PagerFlags arg_pager_flags = 0; |
2a12e32e | 103 | static bool arg_service_watchdogs = true; |
706343f4 | 104 | static ExecOutput arg_default_std_output = EXEC_OUTPUT_JOURNAL; |
0a494f1f | 105 | static ExecOutput arg_default_std_error = EXEC_OUTPUT_INHERIT; |
1f19a534 OS |
106 | static usec_t arg_default_restart_usec = DEFAULT_RESTART_USEC; |
107 | static usec_t arg_default_timeout_start_usec = DEFAULT_TIMEOUT_USEC; | |
108 | static usec_t arg_default_timeout_stop_usec = DEFAULT_TIMEOUT_USEC; | |
3f41e1e5 LN |
109 | static usec_t arg_default_start_limit_interval = DEFAULT_START_LIMIT_INTERVAL; |
110 | static unsigned arg_default_start_limit_burst = DEFAULT_START_LIMIT_BURST; | |
e96d6be7 LP |
111 | static usec_t arg_runtime_watchdog = 0; |
112 | static usec_t arg_shutdown_watchdog = 10 * USEC_PER_MINUTE; | |
c6885f5f | 113 | static char *arg_early_core_pattern = NULL; |
8a2c1fbf | 114 | static char *arg_watchdog_device = NULL; |
97d0e5f8 | 115 | static char **arg_default_environment = NULL; |
517d56b1 | 116 | static struct rlimit *arg_default_rlimit[_RLIMIT_MAX] = {}; |
a103496c | 117 | static uint64_t arg_capability_bounding_set = CAP_ALL; |
39362f6f | 118 | static bool arg_no_new_privs = false; |
3a43da28 | 119 | static nsec_t arg_timer_slack_nsec = NSEC_INFINITY; |
bd8f585b | 120 | static usec_t arg_default_timer_accuracy_usec = 1 * USEC_PER_MINUTE; |
d3b1c508 LP |
121 | static Set* arg_syscall_archs = NULL; |
122 | static FILE* arg_serialization = NULL; | |
a88c5b8a | 123 | static int arg_default_cpu_accounting = -1; |
13c31542 | 124 | static bool arg_default_io_accounting = false; |
377bfd2d | 125 | static bool arg_default_ip_accounting = false; |
085afe36 | 126 | static bool arg_default_blockio_accounting = false; |
444d5863 | 127 | static bool arg_default_memory_accounting = MEMORY_ACCOUNTING_DEFAULT; |
9ded9cd1 | 128 | static bool arg_default_tasks_accounting = true; |
79baeeb9 | 129 | static uint64_t arg_default_tasks_max = UINT64_MAX; |
ee48dbd5 | 130 | static sd_id128_t arg_machine_id = {}; |
ae8c7939 | 131 | static EmergencyAction arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE; |
4fc935ca | 132 | |
848e863a | 133 | _noreturn_ static void freeze_or_reboot(void) { |
b9e74c39 LP |
134 | |
135 | if (arg_crash_reboot) { | |
136 | log_notice("Rebooting in 10s..."); | |
137 | (void) sleep(10); | |
138 | ||
139 | log_notice("Rebooting now..."); | |
140 | (void) reboot(RB_AUTOBOOT); | |
141 | log_emergency_errno(errno, "Failed to reboot: %m"); | |
142 | } | |
143 | ||
144 | log_emergency("Freezing execution."); | |
145 | freeze(); | |
146 | } | |
147 | ||
848e863a | 148 | _noreturn_ static void crash(int sig) { |
7d06dad9 MS |
149 | struct sigaction sa; |
150 | pid_t pid; | |
97c4f35c | 151 | |
df0ff127 | 152 | if (getpid_cached() != 1) |
abb26902 | 153 | /* Pass this on immediately, if this is not PID 1 */ |
92ca4cac | 154 | (void) raise(sig); |
abb26902 | 155 | else if (!arg_dump_core) |
4104970e | 156 | log_emergency("Caught <%s>, not dumping core.", signal_to_string(sig)); |
97c4f35c | 157 | else { |
7d06dad9 | 158 | sa = (struct sigaction) { |
189d5bac | 159 | .sa_handler = nop_signal_handler, |
b92bea5d ZJS |
160 | .sa_flags = SA_NOCLDSTOP|SA_RESTART, |
161 | }; | |
97c4f35c | 162 | |
6f5e3f35 | 163 | /* We want to wait for the core process, hence let's enable SIGCHLD */ |
92ca4cac | 164 | (void) sigaction(SIGCHLD, &sa, NULL); |
6f5e3f35 | 165 | |
8869a0b4 | 166 | pid = raw_clone(SIGCHLD); |
e62d8c39 | 167 | if (pid < 0) |
56f64d95 | 168 | log_emergency_errno(errno, "Caught <%s>, cannot fork for core dump: %m", signal_to_string(sig)); |
97c4f35c | 169 | else if (pid == 0) { |
97c4f35c | 170 | /* Enable default signal handler for core dump */ |
15a90032 | 171 | |
92ca4cac LP |
172 | sa = (struct sigaction) { |
173 | .sa_handler = SIG_DFL, | |
174 | }; | |
175 | (void) sigaction(sig, &sa, NULL); | |
97c4f35c | 176 | |
15a90032 LP |
177 | /* Don't limit the coredump size */ |
178 | (void) setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)); | |
97c4f35c LP |
179 | |
180 | /* Just to be sure... */ | |
e62d9b81 | 181 | (void) chdir("/"); |
97c4f35c LP |
182 | |
183 | /* Raise the signal again */ | |
ee05e779 | 184 | pid = raw_getpid(); |
92ca4cac | 185 | (void) kill(pid, sig); /* raise() would kill the parent */ |
97c4f35c LP |
186 | |
187 | assert_not_reached("We shouldn't be here..."); | |
92ca4cac | 188 | _exit(EXIT_FAILURE); |
4fc935ca | 189 | } else { |
8e12a6ae LP |
190 | siginfo_t status; |
191 | int r; | |
4fc935ca LP |
192 | |
193 | /* Order things nicely. */ | |
e62d8c39 ZJS |
194 | r = wait_for_terminate(pid, &status); |
195 | if (r < 0) | |
da927ba9 | 196 | log_emergency_errno(r, "Caught <%s>, waitpid() failed: %m", signal_to_string(sig)); |
8e12a6ae | 197 | else if (status.si_code != CLD_DUMPED) |
ee05e779 ZJS |
198 | log_emergency("Caught <%s>, core dump failed (child "PID_FMT", code=%s, status=%i/%s).", |
199 | signal_to_string(sig), | |
200 | pid, sigchld_code_to_string(status.si_code), | |
201 | status.si_status, | |
202 | strna(status.si_code == CLD_EXITED | |
38107f5a | 203 | ? exit_status_to_string(status.si_status, EXIT_STATUS_MINIMAL) |
ee05e779 | 204 | : signal_to_string(status.si_status))); |
4fc935ca | 205 | else |
4104970e | 206 | log_emergency("Caught <%s>, dumped core as pid "PID_FMT".", signal_to_string(sig), pid); |
97c4f35c LP |
207 | } |
208 | } | |
209 | ||
b9e74c39 | 210 | if (arg_crash_chvt >= 0) |
92ca4cac | 211 | (void) chvt(arg_crash_chvt); |
601f6a1e | 212 | |
7d06dad9 MS |
213 | sa = (struct sigaction) { |
214 | .sa_handler = SIG_IGN, | |
215 | .sa_flags = SA_NOCLDSTOP|SA_NOCLDWAIT|SA_RESTART, | |
216 | }; | |
217 | ||
218 | /* Let the kernel reap children for us */ | |
219 | (void) sigaction(SIGCHLD, &sa, NULL); | |
8c43883a | 220 | |
7d06dad9 | 221 | if (arg_crash_shell) { |
b9e74c39 | 222 | log_notice("Executing crash shell in 10s..."); |
92ca4cac | 223 | (void) sleep(10); |
4fc935ca | 224 | |
8869a0b4 | 225 | pid = raw_clone(SIGCHLD); |
cd3bd60a | 226 | if (pid < 0) |
56f64d95 | 227 | log_emergency_errno(errno, "Failed to fork off crash shell: %m"); |
6f5e3f35 | 228 | else if (pid == 0) { |
b9e74c39 | 229 | (void) setsid(); |
92ca4cac LP |
230 | (void) make_console_stdio(); |
231 | (void) execle("/bin/sh", "/bin/sh", NULL, environ); | |
6f5e3f35 | 232 | |
ee05e779 | 233 | log_emergency_errno(errno, "execle() failed: %m"); |
92ca4cac | 234 | _exit(EXIT_FAILURE); |
b9e74c39 LP |
235 | } else { |
236 | log_info("Spawned crash shell as PID "PID_FMT".", pid); | |
4cf0b03b | 237 | (void) wait_for_terminate(pid, NULL); |
b9e74c39 | 238 | } |
4fc935ca LP |
239 | } |
240 | ||
b9e74c39 | 241 | freeze_or_reboot(); |
97c4f35c LP |
242 | } |
243 | ||
244 | static void install_crash_handler(void) { | |
297d563d | 245 | static const struct sigaction sa = { |
b92bea5d | 246 | .sa_handler = crash, |
297d563d | 247 | .sa_flags = SA_NODEFER, /* So that we can raise the signal again from the signal handler */ |
b92bea5d | 248 | }; |
297d563d | 249 | int r; |
97c4f35c | 250 | |
297d563d LP |
251 | /* We ignore the return value here, since, we don't mind if we |
252 | * cannot set up a crash handler */ | |
253 | r = sigaction_many(&sa, SIGNALS_CRASH_HANDLER, -1); | |
254 | if (r < 0) | |
255 | log_debug_errno(r, "I had trouble setting up the crash handler, ignoring: %m"); | |
97c4f35c | 256 | } |
f170852a | 257 | |
56d96fc0 LP |
258 | static int console_setup(void) { |
259 | _cleanup_close_ int tty_fd = -1; | |
260 | int r; | |
80876c20 | 261 | |
512947d4 | 262 | tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); |
23bbb0de MS |
263 | if (tty_fd < 0) |
264 | return log_error_errno(tty_fd, "Failed to open /dev/console: %m"); | |
80876c20 | 265 | |
56d96fc0 LP |
266 | /* We don't want to force text mode. plymouth may be showing |
267 | * pictures already from initrd. */ | |
512947d4 | 268 | r = reset_terminal_fd(tty_fd, false); |
23bbb0de MS |
269 | if (r < 0) |
270 | return log_error_errno(r, "Failed to reset /dev/console: %m"); | |
843d2643 | 271 | |
56d96fc0 | 272 | return 0; |
80876c20 LP |
273 | } |
274 | ||
b9e74c39 LP |
275 | static int parse_crash_chvt(const char *value) { |
276 | int b; | |
277 | ||
278 | if (safe_atoi(value, &arg_crash_chvt) >= 0) | |
279 | return 0; | |
280 | ||
281 | b = parse_boolean(value); | |
282 | if (b < 0) | |
283 | return b; | |
284 | ||
285 | if (b > 0) | |
286 | arg_crash_chvt = 0; /* switch to where kmsg goes */ | |
287 | else | |
288 | arg_crash_chvt = -1; /* turn off switching */ | |
289 | ||
290 | return 0; | |
291 | } | |
292 | ||
7d5ceb64 FB |
293 | static int parse_confirm_spawn(const char *value, char **console) { |
294 | char *s; | |
295 | int r; | |
296 | ||
297 | r = value ? parse_boolean(value) : 1; | |
298 | if (r == 0) { | |
299 | *console = NULL; | |
300 | return 0; | |
301 | } | |
302 | ||
303 | if (r > 0) /* on with default tty */ | |
304 | s = strdup("/dev/console"); | |
305 | else if (is_path(value)) /* on with fully qualified path */ | |
306 | s = strdup(value); | |
307 | else /* on with only a tty file name, not a fully qualified path */ | |
308 | s = strjoin("/dev/", value); | |
309 | if (!s) | |
310 | return -ENOMEM; | |
5b65ae15 | 311 | |
7d5ceb64 FB |
312 | *console = s; |
313 | return 0; | |
314 | } | |
315 | ||
ee48dbd5 | 316 | static int set_machine_id(const char *m) { |
e042eab7 | 317 | sd_id128_t t; |
8b26cdbd | 318 | assert(m); |
ee48dbd5 | 319 | |
e042eab7 | 320 | if (sd_id128_from_string(m, &t) < 0) |
ee48dbd5 NC |
321 | return -EINVAL; |
322 | ||
e042eab7 | 323 | if (sd_id128_is_null(t)) |
ee48dbd5 NC |
324 | return -EINVAL; |
325 | ||
e042eab7 | 326 | arg_machine_id = t; |
ee48dbd5 NC |
327 | return 0; |
328 | } | |
329 | ||
96287a49 | 330 | static int parse_proc_cmdline_item(const char *key, const char *value, void *data) { |
f170852a | 331 | |
059cb385 | 332 | int r; |
f170852a | 333 | |
059cb385 | 334 | assert(key); |
5192bd19 | 335 | |
1d84ad94 | 336 | if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) { |
bf4df7c3 | 337 | |
1d84ad94 LP |
338 | if (proc_cmdline_value_missing(key, value)) |
339 | return 0; | |
bf4df7c3 | 340 | |
1d84ad94 LP |
341 | if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE)) |
342 | log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value); | |
343 | else if (in_initrd() == !!startswith(key, "rd.")) { | |
344 | if (free_and_strdup(&arg_default_unit, value) < 0) | |
345 | return log_oom(); | |
346 | } | |
f170852a | 347 | |
1d84ad94 | 348 | } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) { |
4fc935ca | 349 | |
1d84ad94 | 350 | r = value ? parse_boolean(value) : true; |
fb472900 | 351 | if (r < 0) |
5e1ee764 | 352 | log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value); |
4fc935ca | 353 | else |
fa0f4d8a | 354 | arg_dump_core = r; |
4fc935ca | 355 | |
c6885f5f FB |
356 | } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) { |
357 | ||
358 | if (proc_cmdline_value_missing(key, value)) | |
359 | return 0; | |
360 | ||
361 | if (path_is_absolute(value)) | |
362 | (void) parse_path_argument_and_warn(value, false, &arg_early_core_pattern); | |
363 | else | |
364 | log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value); | |
365 | ||
1d84ad94 | 366 | } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) { |
b9e74c39 | 367 | |
1d84ad94 LP |
368 | if (!value) |
369 | arg_crash_chvt = 0; /* turn on */ | |
5e1ee764 YW |
370 | else { |
371 | r = parse_crash_chvt(value); | |
372 | if (r < 0) | |
373 | log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value); | |
374 | } | |
b9e74c39 | 375 | |
1d84ad94 | 376 | } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) { |
4fc935ca | 377 | |
1d84ad94 | 378 | r = value ? parse_boolean(value) : true; |
fb472900 | 379 | if (r < 0) |
5e1ee764 | 380 | log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value); |
4fc935ca | 381 | else |
fa0f4d8a | 382 | arg_crash_shell = r; |
5e7ee61c | 383 | |
1d84ad94 | 384 | } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) { |
5e7ee61c | 385 | |
1d84ad94 | 386 | r = value ? parse_boolean(value) : true; |
b9e74c39 | 387 | if (r < 0) |
5e1ee764 | 388 | log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value); |
5e7ee61c | 389 | else |
b9e74c39 | 390 | arg_crash_reboot = r; |
5e7ee61c | 391 | |
1d84ad94 LP |
392 | } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) { |
393 | char *s; | |
7d5ceb64 | 394 | |
1d84ad94 | 395 | r = parse_confirm_spawn(value, &s); |
059cb385 | 396 | if (r < 0) |
5e1ee764 YW |
397 | log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value); |
398 | else | |
399 | free_and_replace(arg_confirm_spawn, s); | |
601f6a1e | 400 | |
2a12e32e JK |
401 | } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) { |
402 | ||
403 | r = value ? parse_boolean(value) : true; | |
404 | if (r < 0) | |
5e1ee764 | 405 | log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value); |
2a12e32e JK |
406 | else |
407 | arg_service_watchdogs = r; | |
408 | ||
1d84ad94 | 409 | } else if (proc_cmdline_key_streq(key, "systemd.show_status")) { |
9e58ff9c | 410 | |
1d84ad94 LP |
411 | if (value) { |
412 | r = parse_show_status(value, &arg_show_status); | |
413 | if (r < 0) | |
5e1ee764 | 414 | log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value); |
1d84ad94 LP |
415 | } else |
416 | arg_show_status = SHOW_STATUS_YES; | |
059cb385 | 417 | |
1d84ad94 LP |
418 | } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) { |
419 | ||
420 | if (proc_cmdline_value_missing(key, value)) | |
421 | return 0; | |
0a494f1f | 422 | |
059cb385 | 423 | r = exec_output_from_string(value); |
fb472900 | 424 | if (r < 0) |
5e1ee764 | 425 | log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value); |
0a494f1f LP |
426 | else |
427 | arg_default_std_output = r; | |
0a494f1f | 428 | |
1d84ad94 LP |
429 | } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) { |
430 | ||
431 | if (proc_cmdline_value_missing(key, value)) | |
432 | return 0; | |
059cb385 LP |
433 | |
434 | r = exec_output_from_string(value); | |
fb472900 | 435 | if (r < 0) |
5e1ee764 | 436 | log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value); |
0a494f1f LP |
437 | else |
438 | arg_default_std_error = r; | |
9e7c5357 | 439 | |
1d84ad94 LP |
440 | } else if (streq(key, "systemd.setenv")) { |
441 | ||
442 | if (proc_cmdline_value_missing(key, value)) | |
443 | return 0; | |
059cb385 LP |
444 | |
445 | if (env_assignment_is_valid(value)) { | |
e21fea24 KS |
446 | char **env; |
447 | ||
059cb385 | 448 | env = strv_env_set(arg_default_environment, value); |
1d84ad94 LP |
449 | if (!env) |
450 | return log_oom(); | |
451 | ||
452 | arg_default_environment = env; | |
e21fea24 | 453 | } else |
059cb385 | 454 | log_warning("Environment variable name '%s' is not valid. Ignoring.", value); |
9e58ff9c | 455 | |
1d84ad94 LP |
456 | } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) { |
457 | ||
458 | if (proc_cmdline_value_missing(key, value)) | |
459 | return 0; | |
460 | ||
461 | r = set_machine_id(value); | |
462 | if (r < 0) | |
5e1ee764 | 463 | log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value); |
ee48dbd5 | 464 | |
1d84ad94 LP |
465 | } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) { |
466 | ||
467 | if (proc_cmdline_value_missing(key, value)) | |
468 | return 0; | |
469 | ||
470 | r = parse_sec(value, &arg_default_timeout_start_usec); | |
471 | if (r < 0) | |
5e1ee764 | 472 | log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value); |
1d84ad94 LP |
473 | |
474 | if (arg_default_timeout_start_usec <= 0) | |
475 | arg_default_timeout_start_usec = USEC_INFINITY; | |
ee48dbd5 | 476 | |
8a2c1fbf EJ |
477 | } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) { |
478 | ||
479 | if (proc_cmdline_value_missing(key, value)) | |
480 | return 0; | |
481 | ||
c6885f5f | 482 | (void) parse_path_argument_and_warn(value, false, &arg_watchdog_device); |
8a2c1fbf | 483 | |
059cb385 | 484 | } else if (streq(key, "quiet") && !value) { |
d7b15e0a | 485 | |
7a293242 | 486 | if (arg_show_status == _SHOW_STATUS_INVALID) |
d450b6f2 | 487 | arg_show_status = SHOW_STATUS_AUTO; |
059cb385 LP |
488 | |
489 | } else if (streq(key, "debug") && !value) { | |
d7b15e0a | 490 | |
1de1c9c3 LP |
491 | /* Note that log_parse_environment() handles 'debug' |
492 | * too, and sets the log level to LOG_DEBUG. */ | |
d7b15e0a | 493 | |
75f86906 | 494 | if (detect_container() > 0) |
b2103dcc | 495 | log_set_target(LOG_TARGET_CONSOLE); |
059cb385 | 496 | |
dcd61450 | 497 | } else if (!value) { |
e2c9a131 | 498 | const char *target; |
f170852a LP |
499 | |
500 | /* SysV compatibility */ | |
e2c9a131 EV |
501 | target = runlevel_to_target(key); |
502 | if (target) | |
503 | return free_and_strdup(&arg_default_unit, target); | |
f170852a LP |
504 | } |
505 | ||
506 | return 0; | |
507 | } | |
508 | ||
e8e581bf ZJS |
509 | #define DEFINE_SETTER(name, func, descr) \ |
510 | static int name(const char *unit, \ | |
511 | const char *filename, \ | |
512 | unsigned line, \ | |
513 | const char *section, \ | |
71a61510 | 514 | unsigned section_line, \ |
e8e581bf ZJS |
515 | const char *lvalue, \ |
516 | int ltype, \ | |
517 | const char *rvalue, \ | |
518 | void *data, \ | |
519 | void *userdata) { \ | |
520 | \ | |
521 | int r; \ | |
522 | \ | |
523 | assert(filename); \ | |
524 | assert(lvalue); \ | |
525 | assert(rvalue); \ | |
526 | \ | |
527 | r = func(rvalue); \ | |
528 | if (r < 0) \ | |
d1cefe0a LP |
529 | log_syntax(unit, LOG_ERR, filename, line, r, \ |
530 | "Invalid " descr "'%s': %m", \ | |
531 | rvalue); \ | |
e8e581bf ZJS |
532 | \ |
533 | return 0; \ | |
534 | } | |
487393e9 | 535 | |
a6ecbf83 FB |
536 | DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level"); |
537 | DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target"); | |
538 | DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color" ); | |
539 | DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location"); | |
487393e9 | 540 | |
4d7213b2 LP |
541 | static int config_parse_cpu_affinity2( |
542 | const char *unit, | |
543 | const char *filename, | |
544 | unsigned line, | |
545 | const char *section, | |
546 | unsigned section_line, | |
547 | const char *lvalue, | |
548 | int ltype, | |
549 | const char *rvalue, | |
550 | void *data, | |
551 | void *userdata) { | |
487393e9 | 552 | |
4b40bc38 | 553 | _cleanup_cpu_free_ cpu_set_t *c = NULL; |
f5c72b73 | 554 | int ncpus; |
487393e9 | 555 | |
765d143b | 556 | ncpus = parse_cpu_set_and_warn(rvalue, &c, unit, filename, line, lvalue); |
f5c72b73 FB |
557 | if (ncpus < 0) |
558 | return ncpus; | |
487393e9 | 559 | |
f5c72b73 | 560 | if (sched_setaffinity(0, CPU_ALLOC_SIZE(ncpus), c) < 0) |
25f027c5 | 561 | log_warning_errno(errno, "Failed to set CPU affinity: %m"); |
487393e9 | 562 | |
487393e9 LP |
563 | return 0; |
564 | } | |
565 | ||
4d7213b2 LP |
566 | static int config_parse_show_status( |
567 | const char* unit, | |
568 | const char *filename, | |
569 | unsigned line, | |
570 | const char *section, | |
571 | unsigned section_line, | |
572 | const char *lvalue, | |
573 | int ltype, | |
574 | const char *rvalue, | |
575 | void *data, | |
576 | void *userdata) { | |
577 | ||
578 | int k; | |
579 | ShowStatus *b = data; | |
580 | ||
581 | assert(filename); | |
582 | assert(lvalue); | |
583 | assert(rvalue); | |
584 | assert(data); | |
585 | ||
586 | k = parse_show_status(rvalue, b); | |
587 | if (k < 0) { | |
12ca818f | 588 | log_syntax(unit, LOG_ERR, filename, line, k, "Failed to parse show status setting, ignoring: %s", rvalue); |
4d7213b2 LP |
589 | return 0; |
590 | } | |
591 | ||
592 | return 0; | |
593 | } | |
594 | ||
5db98187 LP |
595 | static int config_parse_output_restricted( |
596 | const char* unit, | |
597 | const char *filename, | |
598 | unsigned line, | |
599 | const char *section, | |
600 | unsigned section_line, | |
601 | const char *lvalue, | |
602 | int ltype, | |
603 | const char *rvalue, | |
604 | void *data, | |
605 | void *userdata) { | |
606 | ||
607 | ExecOutput t, *eo = data; | |
608 | ||
609 | assert(filename); | |
610 | assert(lvalue); | |
611 | assert(rvalue); | |
612 | assert(data); | |
613 | ||
614 | t = exec_output_from_string(rvalue); | |
615 | if (t < 0) { | |
616 | log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse output type, ignoring: %s", rvalue); | |
617 | return 0; | |
618 | } | |
619 | ||
566b7d23 ZD |
620 | if (IN_SET(t, EXEC_OUTPUT_SOCKET, EXEC_OUTPUT_NAMED_FD, EXEC_OUTPUT_FILE, EXEC_OUTPUT_FILE_APPEND)) { |
621 | log_syntax(unit, LOG_ERR, filename, line, 0, "Standard output types socket, fd:, file:, append: are not supported as defaults, ignoring: %s", rvalue); | |
5db98187 LP |
622 | return 0; |
623 | } | |
624 | ||
625 | *eo = t; | |
626 | return 0; | |
627 | } | |
628 | ||
b9e74c39 LP |
629 | static int config_parse_crash_chvt( |
630 | const char* unit, | |
631 | const char *filename, | |
632 | unsigned line, | |
633 | const char *section, | |
634 | unsigned section_line, | |
635 | const char *lvalue, | |
636 | int ltype, | |
637 | const char *rvalue, | |
638 | void *data, | |
639 | void *userdata) { | |
640 | ||
641 | int r; | |
642 | ||
643 | assert(filename); | |
644 | assert(lvalue); | |
645 | assert(rvalue); | |
b9e74c39 LP |
646 | |
647 | r = parse_crash_chvt(rvalue); | |
648 | if (r < 0) { | |
649 | log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse CrashChangeVT= setting, ignoring: %s", rvalue); | |
650 | return 0; | |
651 | } | |
652 | ||
653 | return 0; | |
654 | } | |
655 | ||
487393e9 LP |
656 | static int parse_config_file(void) { |
657 | ||
f975e971 | 658 | const ConfigTableItem items[] = { |
d3b1c508 LP |
659 | { "Manager", "LogLevel", config_parse_level2, 0, NULL }, |
660 | { "Manager", "LogTarget", config_parse_target, 0, NULL }, | |
661 | { "Manager", "LogColor", config_parse_color, 0, NULL }, | |
662 | { "Manager", "LogLocation", config_parse_location, 0, NULL }, | |
663 | { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core }, | |
b9e74c39 LP |
664 | { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, NULL }, |
665 | { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, NULL }, | |
d3b1c508 | 666 | { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell }, |
b9e74c39 | 667 | { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot }, |
d3b1c508 | 668 | { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status }, |
d3b1c508 | 669 | { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, NULL }, |
143fadf3 | 670 | { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, |
d3b1c508 LP |
671 | { "Manager", "RuntimeWatchdogSec", config_parse_sec, 0, &arg_runtime_watchdog }, |
672 | { "Manager", "ShutdownWatchdogSec", config_parse_sec, 0, &arg_shutdown_watchdog }, | |
8a2c1fbf | 673 | { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device }, |
a103496c | 674 | { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set }, |
39362f6f | 675 | { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs }, |
349cc4a5 | 676 | #if HAVE_SECCOMP |
d3b1c508 | 677 | { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs }, |
89fffa27 | 678 | #endif |
d3b1c508 | 679 | { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec }, |
bd8f585b | 680 | { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_default_timer_accuracy_usec }, |
5db98187 LP |
681 | { "Manager", "DefaultStandardOutput", config_parse_output_restricted,0, &arg_default_std_output }, |
682 | { "Manager", "DefaultStandardError", config_parse_output_restricted,0, &arg_default_std_error }, | |
d3b1c508 LP |
683 | { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_default_timeout_start_usec }, |
684 | { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_default_timeout_stop_usec }, | |
685 | { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_default_restart_usec }, | |
f0367da7 LP |
686 | { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_default_start_limit_interval }, /* obsolete alias */ |
687 | { "Manager", "DefaultStartLimitIntervalSec",config_parse_sec, 0, &arg_default_start_limit_interval }, | |
d3b1c508 LP |
688 | { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_default_start_limit_burst }, |
689 | { "Manager", "DefaultEnvironment", config_parse_environ, 0, &arg_default_environment }, | |
4f424df7 LP |
690 | { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_default_rlimit }, |
691 | { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_default_rlimit }, | |
692 | { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_default_rlimit }, | |
693 | { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_default_rlimit }, | |
694 | { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_default_rlimit }, | |
695 | { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_default_rlimit }, | |
696 | { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_default_rlimit }, | |
697 | { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_default_rlimit }, | |
698 | { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_default_rlimit }, | |
699 | { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_default_rlimit }, | |
700 | { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_default_rlimit }, | |
701 | { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_default_rlimit }, | |
702 | { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_default_rlimit }, | |
703 | { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_default_rlimit }, | |
704 | { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_default_rlimit }, | |
705 | { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_default_rlimit }, | |
a88c5b8a | 706 | { "Manager", "DefaultCPUAccounting", config_parse_tristate, 0, &arg_default_cpu_accounting }, |
13c31542 | 707 | { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_default_io_accounting }, |
377bfd2d | 708 | { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_default_ip_accounting }, |
085afe36 LP |
709 | { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_default_blockio_accounting }, |
710 | { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_default_memory_accounting }, | |
03a7b521 | 711 | { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_default_tasks_accounting }, |
0af20ea2 | 712 | { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_default_tasks_max }, |
ae8c7939 | 713 | { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, 0, &arg_cad_burst_action }, |
d3b1c508 | 714 | {} |
487393e9 LP |
715 | }; |
716 | ||
1b907b5c | 717 | const char *fn, *conf_dirs_nulstr; |
487393e9 | 718 | |
463d0d15 | 719 | fn = arg_system ? |
75eb6154 LP |
720 | PKGSYSCONFDIR "/system.conf" : |
721 | PKGSYSCONFDIR "/user.conf"; | |
722 | ||
463d0d15 | 723 | conf_dirs_nulstr = arg_system ? |
75eb6154 LP |
724 | CONF_PATHS_NULSTR("systemd/system.conf.d") : |
725 | CONF_PATHS_NULSTR("systemd/user.conf.d"); | |
726 | ||
bcde742e | 727 | (void) config_parse_many_nulstr(fn, conf_dirs_nulstr, "Manager\0", config_item_table_lookup, items, CONFIG_PARSE_WARN, NULL); |
36c16a7c LP |
728 | |
729 | /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we used USEC_INFINITY | |
730 | * like everywhere else. */ | |
731 | if (arg_default_timeout_start_usec <= 0) | |
732 | arg_default_timeout_start_usec = USEC_INFINITY; | |
733 | if (arg_default_timeout_stop_usec <= 0) | |
734 | arg_default_timeout_stop_usec = USEC_INFINITY; | |
487393e9 | 735 | |
487393e9 LP |
736 | return 0; |
737 | } | |
738 | ||
85cb4151 | 739 | static void set_manager_defaults(Manager *m) { |
06af2a04 TB |
740 | |
741 | assert(m); | |
742 | ||
5b65ae15 LP |
743 | /* Propagates the various default unit property settings into the manager object, i.e. properties that do not |
744 | * affect the manager itself, but are just what newly allocated units will have set if they haven't set | |
745 | * anything else. (Also see set_manager_settings() for the settings that affect the manager's own behaviour) */ | |
746 | ||
06af2a04 TB |
747 | m->default_timer_accuracy_usec = arg_default_timer_accuracy_usec; |
748 | m->default_std_output = arg_default_std_output; | |
749 | m->default_std_error = arg_default_std_error; | |
750 | m->default_timeout_start_usec = arg_default_timeout_start_usec; | |
751 | m->default_timeout_stop_usec = arg_default_timeout_stop_usec; | |
752 | m->default_restart_usec = arg_default_restart_usec; | |
753 | m->default_start_limit_interval = arg_default_start_limit_interval; | |
754 | m->default_start_limit_burst = arg_default_start_limit_burst; | |
a88c5b8a CD |
755 | |
756 | /* On 4.15+ with unified hierarchy, CPU accounting is essentially free as it doesn't require the CPU | |
757 | * controller to be enabled, so the default is to enable it unless we got told otherwise. */ | |
758 | if (arg_default_cpu_accounting >= 0) | |
759 | m->default_cpu_accounting = arg_default_cpu_accounting; | |
760 | else | |
761 | m->default_cpu_accounting = cpu_accounting_is_cheap(); | |
762 | ||
13c31542 | 763 | m->default_io_accounting = arg_default_io_accounting; |
377bfd2d | 764 | m->default_ip_accounting = arg_default_ip_accounting; |
06af2a04 TB |
765 | m->default_blockio_accounting = arg_default_blockio_accounting; |
766 | m->default_memory_accounting = arg_default_memory_accounting; | |
03a7b521 | 767 | m->default_tasks_accounting = arg_default_tasks_accounting; |
0af20ea2 | 768 | m->default_tasks_max = arg_default_tasks_max; |
06af2a04 | 769 | |
79a224c4 LP |
770 | (void) manager_set_default_rlimits(m, arg_default_rlimit); |
771 | ||
772 | (void) manager_default_environment(m); | |
773 | (void) manager_transient_environment_add(m, arg_default_environment); | |
06af2a04 TB |
774 | } |
775 | ||
7b46fc6a LP |
776 | static void set_manager_settings(Manager *m) { |
777 | ||
778 | assert(m); | |
779 | ||
5b65ae15 LP |
780 | /* Propagates the various manager settings into the manager object, i.e. properties that effect the manager |
781 | * itself (as opposed to just being inherited into newly allocated units, see set_manager_defaults() above). */ | |
782 | ||
7b46fc6a | 783 | m->confirm_spawn = arg_confirm_spawn; |
2a12e32e | 784 | m->service_watchdogs = arg_service_watchdogs; |
7b46fc6a LP |
785 | m->runtime_watchdog = arg_runtime_watchdog; |
786 | m->shutdown_watchdog = arg_shutdown_watchdog; | |
787 | m->cad_burst_action = arg_cad_burst_action; | |
788 | ||
789 | manager_set_show_status(m, arg_show_status); | |
790 | } | |
791 | ||
f170852a | 792 | static int parse_argv(int argc, char *argv[]) { |
f170852a LP |
793 | enum { |
794 | ARG_LOG_LEVEL = 0x100, | |
795 | ARG_LOG_TARGET, | |
bbe63281 LP |
796 | ARG_LOG_COLOR, |
797 | ARG_LOG_LOCATION, | |
2f198e2f | 798 | ARG_UNIT, |
edb9aaa8 | 799 | ARG_SYSTEM, |
af2d49f7 | 800 | ARG_USER, |
e537352b | 801 | ARG_TEST, |
b87c2aa6 | 802 | ARG_NO_PAGER, |
9ba0bc4e | 803 | ARG_VERSION, |
80876c20 | 804 | ARG_DUMP_CONFIGURATION_ITEMS, |
bbc1acab | 805 | ARG_DUMP_BUS_PROPERTIES, |
9e58ff9c | 806 | ARG_DUMP_CORE, |
b9e74c39 | 807 | ARG_CRASH_CHVT, |
9e58ff9c | 808 | ARG_CRASH_SHELL, |
b9e74c39 | 809 | ARG_CRASH_REBOOT, |
a16e1123 | 810 | ARG_CONFIRM_SPAWN, |
9e58ff9c | 811 | ARG_SHOW_STATUS, |
4288f619 | 812 | ARG_DESERIALIZE, |
2660882b | 813 | ARG_SWITCHED_ROOT, |
0a494f1f | 814 | ARG_DEFAULT_STD_OUTPUT, |
ee48dbd5 | 815 | ARG_DEFAULT_STD_ERROR, |
2a12e32e JK |
816 | ARG_MACHINE_ID, |
817 | ARG_SERVICE_WATCHDOGS, | |
f170852a LP |
818 | }; |
819 | ||
820 | static const struct option options[] = { | |
a16e1123 LP |
821 | { "log-level", required_argument, NULL, ARG_LOG_LEVEL }, |
822 | { "log-target", required_argument, NULL, ARG_LOG_TARGET }, | |
bbe63281 LP |
823 | { "log-color", optional_argument, NULL, ARG_LOG_COLOR }, |
824 | { "log-location", optional_argument, NULL, ARG_LOG_LOCATION }, | |
2f198e2f | 825 | { "unit", required_argument, NULL, ARG_UNIT }, |
edb9aaa8 | 826 | { "system", no_argument, NULL, ARG_SYSTEM }, |
af2d49f7 | 827 | { "user", no_argument, NULL, ARG_USER }, |
a16e1123 | 828 | { "test", no_argument, NULL, ARG_TEST }, |
b87c2aa6 | 829 | { "no-pager", no_argument, NULL, ARG_NO_PAGER }, |
a16e1123 | 830 | { "help", no_argument, NULL, 'h' }, |
9ba0bc4e | 831 | { "version", no_argument, NULL, ARG_VERSION }, |
a16e1123 | 832 | { "dump-configuration-items", no_argument, NULL, ARG_DUMP_CONFIGURATION_ITEMS }, |
bbc1acab | 833 | { "dump-bus-properties", no_argument, NULL, ARG_DUMP_BUS_PROPERTIES }, |
a5d87bf0 | 834 | { "dump-core", optional_argument, NULL, ARG_DUMP_CORE }, |
b9e74c39 | 835 | { "crash-chvt", required_argument, NULL, ARG_CRASH_CHVT }, |
a5d87bf0 | 836 | { "crash-shell", optional_argument, NULL, ARG_CRASH_SHELL }, |
b9e74c39 | 837 | { "crash-reboot", optional_argument, NULL, ARG_CRASH_REBOOT }, |
a5d87bf0 | 838 | { "confirm-spawn", optional_argument, NULL, ARG_CONFIRM_SPAWN }, |
6e98720f | 839 | { "show-status", optional_argument, NULL, ARG_SHOW_STATUS }, |
a16e1123 | 840 | { "deserialize", required_argument, NULL, ARG_DESERIALIZE }, |
2660882b | 841 | { "switched-root", no_argument, NULL, ARG_SWITCHED_ROOT }, |
0a494f1f LP |
842 | { "default-standard-output", required_argument, NULL, ARG_DEFAULT_STD_OUTPUT, }, |
843 | { "default-standard-error", required_argument, NULL, ARG_DEFAULT_STD_ERROR, }, | |
ee48dbd5 | 844 | { "machine-id", required_argument, NULL, ARG_MACHINE_ID }, |
2a12e32e | 845 | { "service-watchdogs", required_argument, NULL, ARG_SERVICE_WATCHDOGS }, |
fb472900 | 846 | {} |
f170852a LP |
847 | }; |
848 | ||
849 | int c, r; | |
850 | ||
851 | assert(argc >= 1); | |
852 | assert(argv); | |
853 | ||
df0ff127 | 854 | if (getpid_cached() == 1) |
b770165a LP |
855 | opterr = 0; |
856 | ||
099663ff | 857 | while ((c = getopt_long(argc, argv, "hDbsz:", options, NULL)) >= 0) |
f170852a LP |
858 | |
859 | switch (c) { | |
860 | ||
861 | case ARG_LOG_LEVEL: | |
fb472900 | 862 | r = log_set_max_level_from_string(optarg); |
2b5107e1 ZJS |
863 | if (r < 0) |
864 | return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg); | |
f170852a LP |
865 | |
866 | break; | |
867 | ||
868 | case ARG_LOG_TARGET: | |
fb472900 | 869 | r = log_set_target_from_string(optarg); |
2b5107e1 ZJS |
870 | if (r < 0) |
871 | return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg); | |
f170852a LP |
872 | |
873 | break; | |
874 | ||
bbe63281 LP |
875 | case ARG_LOG_COLOR: |
876 | ||
d0b170c8 | 877 | if (optarg) { |
fb472900 | 878 | r = log_show_color_from_string(optarg); |
2b5107e1 ZJS |
879 | if (r < 0) |
880 | return log_error_errno(r, "Failed to parse log color setting \"%s\": %m", | |
881 | optarg); | |
d0b170c8 LP |
882 | } else |
883 | log_show_color(true); | |
bbe63281 LP |
884 | |
885 | break; | |
886 | ||
887 | case ARG_LOG_LOCATION: | |
d0b170c8 | 888 | if (optarg) { |
fb472900 | 889 | r = log_show_location_from_string(optarg); |
2b5107e1 ZJS |
890 | if (r < 0) |
891 | return log_error_errno(r, "Failed to parse log location setting \"%s\": %m", | |
892 | optarg); | |
d0b170c8 LP |
893 | } else |
894 | log_show_location(true); | |
bbe63281 LP |
895 | |
896 | break; | |
897 | ||
0a494f1f | 898 | case ARG_DEFAULT_STD_OUTPUT: |
fb472900 | 899 | r = exec_output_from_string(optarg); |
2b5107e1 ZJS |
900 | if (r < 0) |
901 | return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m", | |
902 | optarg); | |
903 | arg_default_std_output = r; | |
0a494f1f LP |
904 | break; |
905 | ||
906 | case ARG_DEFAULT_STD_ERROR: | |
fb472900 | 907 | r = exec_output_from_string(optarg); |
2b5107e1 ZJS |
908 | if (r < 0) |
909 | return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m", | |
910 | optarg); | |
911 | arg_default_std_error = r; | |
0a494f1f LP |
912 | break; |
913 | ||
2f198e2f | 914 | case ARG_UNIT: |
e6e242ad | 915 | r = free_and_strdup(&arg_default_unit, optarg); |
23bbb0de | 916 | if (r < 0) |
2b5107e1 | 917 | return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg); |
f170852a LP |
918 | |
919 | break; | |
920 | ||
edb9aaa8 | 921 | case ARG_SYSTEM: |
463d0d15 | 922 | arg_system = true; |
edb9aaa8 | 923 | break; |
a5dab5ce | 924 | |
af2d49f7 | 925 | case ARG_USER: |
463d0d15 | 926 | arg_system = false; |
a5dab5ce | 927 | break; |
a5dab5ce | 928 | |
e965d56d | 929 | case ARG_TEST: |
fa0f4d8a | 930 | arg_action = ACTION_TEST; |
b87c2aa6 ZJS |
931 | break; |
932 | ||
933 | case ARG_NO_PAGER: | |
0221d68a | 934 | arg_pager_flags |= PAGER_DISABLE; |
e965d56d LP |
935 | break; |
936 | ||
9ba0bc4e ZJS |
937 | case ARG_VERSION: |
938 | arg_action = ACTION_VERSION; | |
939 | break; | |
940 | ||
e537352b | 941 | case ARG_DUMP_CONFIGURATION_ITEMS: |
fa0f4d8a | 942 | arg_action = ACTION_DUMP_CONFIGURATION_ITEMS; |
e537352b LP |
943 | break; |
944 | ||
bbc1acab YW |
945 | case ARG_DUMP_BUS_PROPERTIES: |
946 | arg_action = ACTION_DUMP_BUS_PROPERTIES; | |
947 | break; | |
948 | ||
9e58ff9c | 949 | case ARG_DUMP_CORE: |
b9e74c39 LP |
950 | if (!optarg) |
951 | arg_dump_core = true; | |
952 | else { | |
953 | r = parse_boolean(optarg); | |
954 | if (r < 0) | |
2b5107e1 ZJS |
955 | return log_error_errno(r, "Failed to parse dump core boolean: \"%s\": %m", |
956 | optarg); | |
b9e74c39 | 957 | arg_dump_core = r; |
a5d87bf0 | 958 | } |
b9e74c39 LP |
959 | break; |
960 | ||
961 | case ARG_CRASH_CHVT: | |
962 | r = parse_crash_chvt(optarg); | |
963 | if (r < 0) | |
2b5107e1 ZJS |
964 | return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m", |
965 | optarg); | |
9e58ff9c LP |
966 | break; |
967 | ||
968 | case ARG_CRASH_SHELL: | |
b9e74c39 LP |
969 | if (!optarg) |
970 | arg_crash_shell = true; | |
971 | else { | |
972 | r = parse_boolean(optarg); | |
973 | if (r < 0) | |
2b5107e1 ZJS |
974 | return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m", |
975 | optarg); | |
b9e74c39 LP |
976 | arg_crash_shell = r; |
977 | } | |
978 | break; | |
979 | ||
980 | case ARG_CRASH_REBOOT: | |
981 | if (!optarg) | |
982 | arg_crash_reboot = true; | |
983 | else { | |
984 | r = parse_boolean(optarg); | |
985 | if (r < 0) | |
2b5107e1 ZJS |
986 | return log_error_errno(r, "Failed to parse crash shell boolean: \"%s\": %m", |
987 | optarg); | |
b9e74c39 | 988 | arg_crash_reboot = r; |
a5d87bf0 | 989 | } |
9e58ff9c LP |
990 | break; |
991 | ||
80876c20 | 992 | case ARG_CONFIRM_SPAWN: |
7d5ceb64 FB |
993 | arg_confirm_spawn = mfree(arg_confirm_spawn); |
994 | ||
995 | r = parse_confirm_spawn(optarg, &arg_confirm_spawn); | |
996 | if (r < 0) | |
2b5107e1 ZJS |
997 | return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m", |
998 | optarg); | |
80876c20 LP |
999 | break; |
1000 | ||
2a12e32e JK |
1001 | case ARG_SERVICE_WATCHDOGS: |
1002 | r = parse_boolean(optarg); | |
1003 | if (r < 0) | |
2b5107e1 ZJS |
1004 | return log_error_errno(r, "Failed to parse service watchdogs boolean: \"%s\": %m", |
1005 | optarg); | |
2a12e32e JK |
1006 | arg_service_watchdogs = r; |
1007 | break; | |
1008 | ||
9e58ff9c | 1009 | case ARG_SHOW_STATUS: |
d450b6f2 ZJS |
1010 | if (optarg) { |
1011 | r = parse_show_status(optarg, &arg_show_status); | |
ac7ec288 | 1012 | if (r < 0) |
2b5107e1 ZJS |
1013 | return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m", |
1014 | optarg); | |
d450b6f2 ZJS |
1015 | } else |
1016 | arg_show_status = SHOW_STATUS_YES; | |
6e98720f | 1017 | break; |
a5d87bf0 | 1018 | |
a16e1123 LP |
1019 | case ARG_DESERIALIZE: { |
1020 | int fd; | |
1021 | FILE *f; | |
1022 | ||
01e10de3 | 1023 | r = safe_atoi(optarg, &fd); |
2b5107e1 ZJS |
1024 | if (r < 0) |
1025 | log_error_errno(r, "Failed to parse deserialize option \"%s\": %m", optarg); | |
1026 | if (fd < 0) { | |
1027 | log_error("Invalid deserialize fd: %d", fd); | |
b9e74c39 | 1028 | return -EINVAL; |
a16e1123 LP |
1029 | } |
1030 | ||
b9e74c39 | 1031 | (void) fd_cloexec(fd, true); |
01e10de3 LP |
1032 | |
1033 | f = fdopen(fd, "r"); | |
4a62c710 | 1034 | if (!f) |
2b5107e1 | 1035 | return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd); |
a16e1123 | 1036 | |
74ca738f | 1037 | safe_fclose(arg_serialization); |
d3b1c508 | 1038 | arg_serialization = f; |
a16e1123 LP |
1039 | |
1040 | break; | |
1041 | } | |
1042 | ||
2660882b | 1043 | case ARG_SWITCHED_ROOT: |
bf4df7c3 | 1044 | arg_switched_root = true; |
d03bc1b8 HH |
1045 | break; |
1046 | ||
ee48dbd5 NC |
1047 | case ARG_MACHINE_ID: |
1048 | r = set_machine_id(optarg); | |
54500613 | 1049 | if (r < 0) |
2b5107e1 | 1050 | return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg); |
ee48dbd5 NC |
1051 | break; |
1052 | ||
f170852a | 1053 | case 'h': |
fa0f4d8a | 1054 | arg_action = ACTION_HELP; |
f170852a LP |
1055 | break; |
1056 | ||
1d2e23ab LP |
1057 | case 'D': |
1058 | log_set_max_level(LOG_DEBUG); | |
1059 | break; | |
1060 | ||
099663ff LP |
1061 | case 'b': |
1062 | case 's': | |
1063 | case 'z': | |
1064 | /* Just to eat away the sysvinit kernel | |
1065 | * cmdline args without getopt() error | |
1066 | * messages that we'll parse in | |
1067 | * parse_proc_cmdline_word() or ignore. */ | |
f170852a | 1068 | |
099663ff | 1069 | case '?': |
df0ff127 | 1070 | if (getpid_cached() != 1) |
099663ff | 1071 | return -EINVAL; |
601185b4 ZJS |
1072 | else |
1073 | return 0; | |
099663ff | 1074 | |
601185b4 ZJS |
1075 | default: |
1076 | assert_not_reached("Unhandled option code."); | |
f170852a LP |
1077 | } |
1078 | ||
df0ff127 | 1079 | if (optind < argc && getpid_cached() != 1) { |
d821e6d6 LP |
1080 | /* Hmm, when we aren't run as init system |
1081 | * let's complain about excess arguments */ | |
1082 | ||
1083 | log_error("Excess arguments."); | |
1084 | return -EINVAL; | |
1085 | } | |
1086 | ||
f170852a LP |
1087 | return 0; |
1088 | } | |
1089 | ||
1090 | static int help(void) { | |
37ec0fdd LP |
1091 | _cleanup_free_ char *link = NULL; |
1092 | int r; | |
1093 | ||
1094 | r = terminal_urlify_man("systemd", "1", &link); | |
1095 | if (r < 0) | |
1096 | return log_oom(); | |
f170852a | 1097 | |
2e33c433 | 1098 | printf("%s [OPTIONS...]\n\n" |
af2d49f7 | 1099 | "Starts up and maintains the system or user services.\n\n" |
e537352b | 1100 | " -h --help Show this help\n" |
cb4069d9 | 1101 | " --version Show version\n" |
e537352b | 1102 | " --test Determine startup sequence, dump it and exit\n" |
b87c2aa6 | 1103 | " --no-pager Do not pipe output into a pager\n" |
80876c20 | 1104 | " --dump-configuration-items Dump understood unit configuration items\n" |
bbc1acab | 1105 | " --dump-bus-properties Dump exposed bus properties\n" |
9e58ff9c | 1106 | " --unit=UNIT Set default unit\n" |
edb9aaa8 | 1107 | " --system Run a system instance, even if PID != 1\n" |
af2d49f7 | 1108 | " --user Run a user instance\n" |
b9e74c39 LP |
1109 | " --dump-core[=BOOL] Dump core on crash\n" |
1110 | " --crash-vt=NR Change to specified VT on crash\n" | |
1111 | " --crash-reboot[=BOOL] Reboot on crash\n" | |
1112 | " --crash-shell[=BOOL] Run shell on crash\n" | |
1113 | " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n" | |
1114 | " --show-status[=BOOL] Show status updates on the console during bootup\n" | |
c1dc6153 | 1115 | " --log-target=TARGET Set log target (console, journal, kmsg, journal-or-kmsg, null)\n" |
9e58ff9c | 1116 | " --log-level=LEVEL Set log level (debug, info, notice, warning, err, crit, alert, emerg)\n" |
b9e74c39 LP |
1117 | " --log-color[=BOOL] Highlight important log messages\n" |
1118 | " --log-location[=BOOL] Include code location in log messages\n" | |
0a494f1f | 1119 | " --default-standard-output= Set default standard output for services\n" |
37ec0fdd LP |
1120 | " --default-standard-error= Set default standard error output for services\n" |
1121 | "\nSee the %s for details.\n" | |
1122 | , program_invocation_short_name | |
1123 | , link | |
1124 | ); | |
f170852a LP |
1125 | |
1126 | return 0; | |
1127 | } | |
1128 | ||
2cc856ac LP |
1129 | static int prepare_reexecute( |
1130 | Manager *m, | |
1131 | FILE **ret_f, | |
1132 | FDSet **ret_fds, | |
1133 | bool switching_root) { | |
1134 | ||
48b90859 LP |
1135 | _cleanup_fdset_free_ FDSet *fds = NULL; |
1136 | _cleanup_fclose_ FILE *f = NULL; | |
a16e1123 LP |
1137 | int r; |
1138 | ||
1139 | assert(m); | |
2cc856ac LP |
1140 | assert(ret_f); |
1141 | assert(ret_fds); | |
a16e1123 | 1142 | |
6b78f9b4 | 1143 | r = manager_open_serialization(m, &f); |
48b90859 LP |
1144 | if (r < 0) |
1145 | return log_error_errno(r, "Failed to create serialization file: %m"); | |
a16e1123 | 1146 | |
71445ae7 | 1147 | /* Make sure nothing is really destructed when we shut down */ |
313cefa1 | 1148 | m->n_reloading++; |
718db961 | 1149 | bus_manager_send_reloading(m, true); |
71445ae7 | 1150 | |
6b78f9b4 | 1151 | fds = fdset_new(); |
48b90859 LP |
1152 | if (!fds) |
1153 | return log_oom(); | |
a16e1123 | 1154 | |
b3680f49 | 1155 | r = manager_serialize(m, f, fds, switching_root); |
48b90859 | 1156 | if (r < 0) |
d68c645b | 1157 | return r; |
a16e1123 | 1158 | |
48b90859 LP |
1159 | if (fseeko(f, 0, SEEK_SET) == (off_t) -1) |
1160 | return log_error_errno(errno, "Failed to rewind serialization fd: %m"); | |
a16e1123 | 1161 | |
6b78f9b4 | 1162 | r = fd_cloexec(fileno(f), false); |
48b90859 LP |
1163 | if (r < 0) |
1164 | return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m"); | |
a16e1123 | 1165 | |
6b78f9b4 | 1166 | r = fdset_cloexec(fds, false); |
48b90859 LP |
1167 | if (r < 0) |
1168 | return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m"); | |
a16e1123 | 1169 | |
2cc856ac LP |
1170 | *ret_f = TAKE_PTR(f); |
1171 | *ret_fds = TAKE_PTR(fds); | |
a16e1123 | 1172 | |
48b90859 | 1173 | return 0; |
a16e1123 LP |
1174 | } |
1175 | ||
a8b627aa LP |
1176 | static void bump_file_max_and_nr_open(void) { |
1177 | ||
1178 | /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large numbers of file | |
1179 | * descriptors are no longer a performance problem and their memory is properly tracked by memcg, thus counting | |
1180 | * them and limiting them in another two layers of limits is unnecessary and just complicates things. This | |
1181 | * function hence turns off 2 of the 4 levels of limits on file descriptors, and makes RLIMIT_NOLIMIT (soft + | |
1182 | * hard) the only ones that really matter. */ | |
1183 | ||
1184 | #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN | |
1185 | _cleanup_free_ char *t = NULL; | |
1186 | int r; | |
1187 | #endif | |
1188 | ||
1189 | #if BUMP_PROC_SYS_FS_FILE_MAX | |
1190 | /* I so wanted to use STRINGIFY(ULONG_MAX) here, but alas we can't as glibc/gcc define that as | |
1191 | * "(0x7fffffffffffffffL * 2UL + 1UL)". Seriously. 😢 */ | |
1192 | if (asprintf(&t, "%lu\n", ULONG_MAX) < 0) { | |
1193 | log_oom(); | |
1194 | return; | |
1195 | } | |
1196 | ||
1197 | r = sysctl_write("fs/file-max", t); | |
1198 | if (r < 0) | |
1199 | log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.file-max, ignoring: %m"); | |
1200 | #endif | |
1201 | ||
1202 | #if BUMP_PROC_SYS_FS_FILE_MAX && BUMP_PROC_SYS_FS_NR_OPEN | |
1203 | t = mfree(t); | |
1204 | #endif | |
1205 | ||
1206 | #if BUMP_PROC_SYS_FS_NR_OPEN | |
1207 | int v = INT_MAX; | |
1208 | ||
1209 | /* Arg! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know what they | |
1210 | * are. The expression by which the maximum is determined is dependent on the architecture, and is something we | |
1211 | * don't really want to copy to userspace, as it is dependent on implementation details of the kernel. Since | |
1212 | * the kernel doesn't expose the maximum value to us, we can only try and hope. Hence, let's start with | |
1213 | * INT_MAX, and then keep halving the value until we find one that works. Ugly? Yes, absolutely, but kernel | |
1214 | * APIs are kernel APIs, so what do can we do... 🤯 */ | |
1215 | ||
1216 | for (;;) { | |
1217 | int k; | |
1218 | ||
1219 | v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */ | |
1220 | if (v < 1024) { | |
1221 | log_warning("Can't bump fs.nr_open, value too small."); | |
1222 | break; | |
1223 | } | |
1224 | ||
1225 | k = read_nr_open(); | |
1226 | if (k < 0) { | |
1227 | log_error_errno(k, "Failed to read fs.nr_open: %m"); | |
1228 | break; | |
1229 | } | |
1230 | if (k >= v) { /* Already larger */ | |
1231 | log_debug("Skipping bump, value is already larger."); | |
1232 | break; | |
1233 | } | |
1234 | ||
1235 | if (asprintf(&t, "%i\n", v) < 0) { | |
1236 | log_oom(); | |
1237 | return; | |
1238 | } | |
1239 | ||
1240 | r = sysctl_write("fs/nr_open", t); | |
1241 | t = mfree(t); | |
1242 | if (r == -EINVAL) { | |
1243 | log_debug("Couldn't write fs.nr_open as %i, halving it.", v); | |
1244 | v /= 2; | |
1245 | continue; | |
1246 | } | |
1247 | if (r < 0) { | |
1248 | log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m"); | |
1249 | break; | |
1250 | } | |
1251 | ||
1252 | log_debug("Successfully bumped fs.nr_open to %i", v); | |
1253 | break; | |
1254 | } | |
1255 | #endif | |
1256 | } | |
1257 | ||
4096d6f5 | 1258 | static int bump_rlimit_nofile(struct rlimit *saved_rlimit) { |
9264cc39 | 1259 | int r, nr; |
4096d6f5 LP |
1260 | |
1261 | assert(saved_rlimit); | |
1262 | ||
52d62075 | 1263 | /* Save the original RLIMIT_NOFILE so that we can reset it later when transitioning from the initrd to the main |
4096d6f5 | 1264 | * systemd or suchlike. */ |
4a62c710 | 1265 | if (getrlimit(RLIMIT_NOFILE, saved_rlimit) < 0) |
3ce40911 | 1266 | return log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m"); |
4096d6f5 | 1267 | |
52d62075 LP |
1268 | /* Get the underlying absolute limit the kernel enforces */ |
1269 | nr = read_nr_open(); | |
1270 | ||
1271 | /* Make sure forked processes get limits based on the original kernel setting */ | |
4096d6f5 LP |
1272 | if (!arg_default_rlimit[RLIMIT_NOFILE]) { |
1273 | struct rlimit *rl; | |
1274 | ||
1275 | rl = newdup(struct rlimit, saved_rlimit, 1); | |
1276 | if (!rl) | |
1277 | return log_oom(); | |
1278 | ||
52d62075 LP |
1279 | /* Bump the hard limit for system services to a substantially higher value. The default hard limit |
1280 | * current kernels set is pretty low (4K), mostly for historical reasons. According to kernel | |
1281 | * developers, the fd handling in recent kernels has been optimized substantially enough, so that we | |
1282 | * can bump the limit now, without paying too high a price in memory or performance. Note however that | |
1283 | * we only bump the hard limit, not the soft limit. That's because select() works the way it works, and | |
1284 | * chokes on fds >= 1024. If we'd bump the soft limit globally, it might accidentally happen to | |
1285 | * unexpecting programs that they get fds higher than what they can process using select(). By only | |
1286 | * bumping the hard limit but leaving the low limit as it is we avoid this pitfall: programs that are | |
1287 | * written by folks aware of the select() problem in mind (and thus use poll()/epoll instead of | |
1288 | * select(), the way everybody should) can explicitly opt into high fds by bumping their soft limit | |
1289 | * beyond 1024, to the hard limit we pass. */ | |
1290 | if (arg_system) | |
1291 | rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE)); | |
1292 | ||
4096d6f5 LP |
1293 | arg_default_rlimit[RLIMIT_NOFILE] = rl; |
1294 | } | |
1295 | ||
52d62075 LP |
1296 | /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for |
1297 | * both hard and soft. */ | |
9264cc39 | 1298 | r = setrlimit_closest(RLIMIT_NOFILE, &RLIMIT_MAKE_CONST(nr)); |
23bbb0de | 1299 | if (r < 0) |
3ce40911 | 1300 | return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m"); |
4096d6f5 LP |
1301 | |
1302 | return 0; | |
1303 | } | |
1304 | ||
fb3ae275 LP |
1305 | static int bump_rlimit_memlock(struct rlimit *saved_rlimit) { |
1306 | int r; | |
1307 | ||
1308 | assert(saved_rlimit); | |
fb3ae275 | 1309 | |
a17c1712 LP |
1310 | /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK which should |
1311 | * normally disable such checks. We need them to implement IPAccessAllow= and IPAccessDeny=, hence let's bump | |
1312 | * the value high enough for our user. */ | |
fb3ae275 LP |
1313 | |
1314 | if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit) < 0) | |
1315 | return log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m"); | |
1316 | ||
c8884ace | 1317 | r = setrlimit_closest(RLIMIT_MEMLOCK, &RLIMIT_MAKE_CONST(HIGH_RLIMIT_MEMLOCK)); |
fb3ae275 LP |
1318 | if (r < 0) |
1319 | return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m"); | |
1320 | ||
1321 | return 0; | |
1322 | } | |
1323 | ||
80758717 | 1324 | static void test_usr(void) { |
80758717 | 1325 | |
ed1c99fc | 1326 | /* Check that /usr is not a separate fs */ |
80758717 | 1327 | |
871c44a7 LP |
1328 | if (dir_is_empty("/usr") <= 0) |
1329 | return; | |
1330 | ||
8b173b5e | 1331 | log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. " |
871c44a7 LP |
1332 | "Some things will probably break (sometimes even silently) in mysterious ways. " |
1333 | "Consult http://freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information."); | |
1334 | } | |
1335 | ||
d3b1c508 | 1336 | static int enforce_syscall_archs(Set *archs) { |
349cc4a5 | 1337 | #if HAVE_SECCOMP |
d3b1c508 LP |
1338 | int r; |
1339 | ||
83f12b27 FS |
1340 | if (!is_seccomp_available()) |
1341 | return 0; | |
1342 | ||
469830d1 | 1343 | r = seccomp_restrict_archs(arg_syscall_archs); |
d3b1c508 | 1344 | if (r < 0) |
469830d1 | 1345 | return log_error_errno(r, "Failed to enforce system call architecture restrication: %m"); |
d3b1c508 | 1346 | #endif |
469830d1 | 1347 | return 0; |
d3b1c508 LP |
1348 | } |
1349 | ||
b6e2f329 LP |
1350 | static int status_welcome(void) { |
1351 | _cleanup_free_ char *pretty_name = NULL, *ansi_color = NULL; | |
1352 | int r; | |
1353 | ||
7a293242 | 1354 | if (IN_SET(arg_show_status, SHOW_STATUS_NO, SHOW_STATUS_AUTO)) |
fd8c85c6 LP |
1355 | return 0; |
1356 | ||
d58ad743 LP |
1357 | r = parse_os_release(NULL, |
1358 | "PRETTY_NAME", &pretty_name, | |
1359 | "ANSI_COLOR", &ansi_color, | |
1360 | NULL); | |
1361 | if (r < 0) | |
1362 | log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, | |
1363 | "Failed to read os-release file, ignoring: %m"); | |
b6e2f329 | 1364 | |
dc9b5816 ZJS |
1365 | if (log_get_show_color()) |
1366 | return status_printf(NULL, false, false, | |
1367 | "\nWelcome to \x1B[%sm%s\x1B[0m!\n", | |
1368 | isempty(ansi_color) ? "1" : ansi_color, | |
1369 | isempty(pretty_name) ? "Linux" : pretty_name); | |
1370 | else | |
1371 | return status_printf(NULL, false, false, | |
1372 | "\nWelcome to %s!\n", | |
1373 | isempty(pretty_name) ? "Linux" : pretty_name); | |
b6e2f329 LP |
1374 | } |
1375 | ||
fdd25311 LP |
1376 | static int write_container_id(void) { |
1377 | const char *c; | |
19854865 | 1378 | int r; |
fdd25311 LP |
1379 | |
1380 | c = getenv("container"); | |
1381 | if (isempty(c)) | |
1382 | return 0; | |
1383 | ||
8612da97 LP |
1384 | RUN_WITH_UMASK(0022) |
1385 | r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE); | |
19854865 | 1386 | if (r < 0) |
f1f849b0 | 1387 | return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m"); |
19854865 LP |
1388 | |
1389 | return 1; | |
1390 | } | |
1391 | ||
1392 | static int bump_unix_max_dgram_qlen(void) { | |
1393 | _cleanup_free_ char *qlen = NULL; | |
1394 | unsigned long v; | |
1395 | int r; | |
1396 | ||
3130fca5 LP |
1397 | /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set the value |
1398 | * really really early during boot, so that it is actually applied to all our sockets, including the | |
1399 | * $NOTIFY_SOCKET one. */ | |
19854865 LP |
1400 | |
1401 | r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen); | |
1402 | if (r < 0) | |
875622c3 | 1403 | return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, "Failed to read AF_UNIX datagram queue length, ignoring: %m"); |
19854865 LP |
1404 | |
1405 | r = safe_atolu(qlen, &v); | |
1406 | if (r < 0) | |
3130fca5 | 1407 | return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen); |
19854865 LP |
1408 | |
1409 | if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN) | |
1410 | return 0; | |
1411 | ||
57512c89 | 1412 | r = write_string_filef("/proc/sys/net/unix/max_dgram_qlen", WRITE_STRING_FILE_DISABLE_BUFFER, "%lu", DEFAULT_UNIX_MAX_DGRAM_QLEN); |
19854865 LP |
1413 | if (r < 0) |
1414 | return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, | |
1415 | "Failed to bump AF_UNIX datagram queue length, ignoring: %m"); | |
1416 | ||
1417 | return 1; | |
fdd25311 LP |
1418 | } |
1419 | ||
32391275 FB |
1420 | static int fixup_environment(void) { |
1421 | _cleanup_free_ char *term = NULL; | |
4dc63c4b | 1422 | const char *t; |
32391275 FB |
1423 | int r; |
1424 | ||
43db615b LP |
1425 | /* Only fix up the environment when we are started as PID 1 */ |
1426 | if (getpid_cached() != 1) | |
1427 | return 0; | |
1428 | ||
1429 | /* We expect the environment to be set correctly if run inside a container. */ | |
84af7821 LP |
1430 | if (detect_container() > 0) |
1431 | return 0; | |
1432 | ||
43db615b LP |
1433 | /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the backend |
1434 | * device used by the console. We try to make a better guess here since some consoles might not have support | |
1435 | * for color mode for example. | |
32391275 | 1436 | * |
43db615b | 1437 | * However if TERM was configured through the kernel command line then leave it alone. */ |
1d84ad94 | 1438 | r = proc_cmdline_get_key("TERM", 0, &term); |
32391275 FB |
1439 | if (r < 0) |
1440 | return r; | |
32391275 | 1441 | |
4dc63c4b LP |
1442 | t = term ?: default_term_for_tty("/dev/console"); |
1443 | ||
1444 | if (setenv("TERM", t, 1) < 0) | |
32391275 FB |
1445 | return -errno; |
1446 | ||
1447 | return 0; | |
1448 | } | |
1449 | ||
6808a0bc LP |
1450 | static void redirect_telinit(int argc, char *argv[]) { |
1451 | ||
1452 | /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */ | |
1453 | ||
1454 | #if HAVE_SYSV_COMPAT | |
1455 | if (getpid_cached() == 1) | |
1456 | return; | |
1457 | ||
1458 | if (!strstr(program_invocation_short_name, "init")) | |
1459 | return; | |
1460 | ||
1461 | execv(SYSTEMCTL_BINARY_PATH, argv); | |
1462 | log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m"); | |
a45d7127 | 1463 | exit(EXIT_FAILURE); |
6808a0bc LP |
1464 | #endif |
1465 | } | |
1466 | ||
4a36297c LP |
1467 | static int become_shutdown( |
1468 | const char *shutdown_verb, | |
7eb35049 | 1469 | int retval) { |
4a36297c LP |
1470 | |
1471 | char log_level[DECIMAL_STR_MAX(int) + 1], | |
e73c54b8 JK |
1472 | exit_code[DECIMAL_STR_MAX(uint8_t) + 1], |
1473 | timeout[DECIMAL_STR_MAX(usec_t) + 1]; | |
4a36297c | 1474 | |
e73c54b8 | 1475 | const char* command_line[13] = { |
4a36297c LP |
1476 | SYSTEMD_SHUTDOWN_BINARY_PATH, |
1477 | shutdown_verb, | |
e73c54b8 | 1478 | "--timeout", timeout, |
4a36297c LP |
1479 | "--log-level", log_level, |
1480 | "--log-target", | |
1481 | }; | |
1482 | ||
1483 | _cleanup_strv_free_ char **env_block = NULL; | |
e73c54b8 | 1484 | size_t pos = 7; |
4a36297c LP |
1485 | int r; |
1486 | ||
7eb35049 | 1487 | assert(shutdown_verb); |
234519ae | 1488 | assert(!command_line[pos]); |
4a36297c LP |
1489 | env_block = strv_copy(environ); |
1490 | ||
1491 | xsprintf(log_level, "%d", log_get_max_level()); | |
e73c54b8 | 1492 | xsprintf(timeout, "%" PRI_USEC "us", arg_default_timeout_stop_usec); |
4a36297c LP |
1493 | |
1494 | switch (log_get_target()) { | |
1495 | ||
1496 | case LOG_TARGET_KMSG: | |
1497 | case LOG_TARGET_JOURNAL_OR_KMSG: | |
1498 | case LOG_TARGET_SYSLOG_OR_KMSG: | |
1499 | command_line[pos++] = "kmsg"; | |
1500 | break; | |
1501 | ||
1502 | case LOG_TARGET_NULL: | |
1503 | command_line[pos++] = "null"; | |
1504 | break; | |
1505 | ||
1506 | case LOG_TARGET_CONSOLE: | |
1507 | default: | |
1508 | command_line[pos++] = "console"; | |
1509 | break; | |
1510 | }; | |
1511 | ||
1512 | if (log_get_show_color()) | |
1513 | command_line[pos++] = "--log-color"; | |
1514 | ||
1515 | if (log_get_show_location()) | |
1516 | command_line[pos++] = "--log-location"; | |
1517 | ||
1518 | if (streq(shutdown_verb, "exit")) { | |
1519 | command_line[pos++] = "--exit-code"; | |
1520 | command_line[pos++] = exit_code; | |
1521 | xsprintf(exit_code, "%d", retval); | |
1522 | } | |
1523 | ||
1524 | assert(pos < ELEMENTSOF(command_line)); | |
1525 | ||
7eb35049 LP |
1526 | if (streq(shutdown_verb, "reboot") && |
1527 | arg_shutdown_watchdog > 0 && | |
1528 | arg_shutdown_watchdog != USEC_INFINITY) { | |
1529 | ||
4a36297c LP |
1530 | char *e; |
1531 | ||
1532 | /* If we reboot let's set the shutdown | |
1533 | * watchdog and tell the shutdown binary to | |
1534 | * repeatedly ping it */ | |
1535 | r = watchdog_set_timeout(&arg_shutdown_watchdog); | |
1536 | watchdog_close(r < 0); | |
1537 | ||
1538 | /* Tell the binary how often to ping, ignore failure */ | |
1539 | if (asprintf(&e, "WATCHDOG_USEC="USEC_FMT, arg_shutdown_watchdog) > 0) | |
8a2c1fbf EJ |
1540 | (void) strv_consume(&env_block, e); |
1541 | ||
1542 | if (arg_watchdog_device && | |
1543 | asprintf(&e, "WATCHDOG_DEVICE=%s", arg_watchdog_device) > 0) | |
1544 | (void) strv_consume(&env_block, e); | |
4a36297c LP |
1545 | } else |
1546 | watchdog_close(true); | |
1547 | ||
1548 | /* Avoid the creation of new processes forked by the | |
1549 | * kernel; at this point, we will not listen to the | |
1550 | * signals anyway */ | |
1551 | if (detect_container() <= 0) | |
1552 | (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER); | |
1553 | ||
1554 | execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block); | |
1555 | return -errno; | |
1556 | } | |
1557 | ||
e839bafd LP |
1558 | static void initialize_clock(void) { |
1559 | int r; | |
1560 | ||
1561 | if (clock_is_localtime(NULL) > 0) { | |
1562 | int min; | |
1563 | ||
1564 | /* | |
1565 | * The very first call of settimeofday() also does a time warp in the kernel. | |
1566 | * | |
1567 | * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to take care | |
1568 | * of maintaining the RTC and do all adjustments. This matches the behavior of Windows, which leaves | |
1569 | * the RTC alone if the registry tells that the RTC runs in UTC. | |
1570 | */ | |
1571 | r = clock_set_timezone(&min); | |
1572 | if (r < 0) | |
1573 | log_error_errno(r, "Failed to apply local time delta, ignoring: %m"); | |
1574 | else | |
1575 | log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min); | |
1576 | ||
1577 | } else if (!in_initrd()) { | |
1578 | /* | |
1579 | * Do a dummy very first call to seal the kernel's time warp magic. | |
1580 | * | |
1581 | * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with LOCAL, but the | |
1582 | * real system could be set up that way. In such case, we need to delay the time-warp or the sealing | |
1583 | * until we reach the real system. | |
1584 | * | |
1585 | * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, the time | |
1586 | * will jump or be incorrect at every daylight saving time change. All kernel local time concepts will | |
1587 | * be treated as UTC that way. | |
1588 | */ | |
1589 | (void) clock_reset_timewarp(); | |
1590 | } | |
1591 | ||
1592 | r = clock_apply_epoch(); | |
1593 | if (r < 0) | |
1594 | log_error_errno(r, "Current system time is before build time, but cannot correct: %m"); | |
1595 | else if (r > 0) | |
1596 | log_info("System time before build time, advancing clock."); | |
1597 | } | |
1598 | ||
1e41242e | 1599 | static void initialize_coredump(bool skip_setup) { |
752bcb77 | 1600 | #if ENABLE_COREDUMP |
1e41242e LP |
1601 | if (getpid_cached() != 1) |
1602 | return; | |
1603 | ||
1604 | /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour the limit) | |
1605 | * will process core dumps for system services by default. */ | |
1606 | if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0) | |
1607 | log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m"); | |
1608 | ||
c6885f5f FB |
1609 | /* But at the same time, turn off the core_pattern logic by default, so that no |
1610 | * coredumps are stored until the systemd-coredump tool is enabled via | |
1611 | * sysctl. However it can be changed via the kernel command line later so core | |
1612 | * dumps can still be generated during early startup and in initramfs. */ | |
1e41242e | 1613 | if (!skip_setup) |
e557b1a6 | 1614 | disable_coredumps(); |
752bcb77 | 1615 | #endif |
1e41242e LP |
1616 | } |
1617 | ||
c6885f5f FB |
1618 | static void initialize_core_pattern(bool skip_setup) { |
1619 | int r; | |
1620 | ||
1621 | if (skip_setup || !arg_early_core_pattern) | |
1622 | return; | |
1623 | ||
1624 | if (getpid_cached() != 1) | |
1625 | return; | |
1626 | ||
57512c89 | 1627 | r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER); |
c6885f5f FB |
1628 | if (r < 0) |
1629 | log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", arg_early_core_pattern); | |
1630 | } | |
1631 | ||
3c7878f9 LP |
1632 | static void do_reexecute( |
1633 | int argc, | |
1634 | char *argv[], | |
1635 | const struct rlimit *saved_rlimit_nofile, | |
1636 | const struct rlimit *saved_rlimit_memlock, | |
1637 | FDSet *fds, | |
1638 | const char *switch_root_dir, | |
1639 | const char *switch_root_init, | |
1640 | const char **ret_error_message) { | |
1641 | ||
1642 | unsigned i, j, args_size; | |
1643 | const char **args; | |
1644 | int r; | |
1645 | ||
1646 | assert(saved_rlimit_nofile); | |
1647 | assert(saved_rlimit_memlock); | |
1648 | assert(ret_error_message); | |
1649 | ||
1650 | /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get rebooted while | |
1651 | * we do that */ | |
1652 | watchdog_close(true); | |
1653 | ||
1654 | /* Reset the RLIMIT_NOFILE to the kernel default, so that the new systemd can pass the kernel default to its | |
1655 | * child processes */ | |
1656 | ||
1657 | if (saved_rlimit_nofile->rlim_cur > 0) | |
1658 | (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile); | |
1659 | if (saved_rlimit_memlock->rlim_cur != (rlim_t) -1) | |
1660 | (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock); | |
1661 | ||
1662 | if (switch_root_dir) { | |
1663 | /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the | |
1664 | * SIGCHLD for them after deserializing. */ | |
e73c54b8 | 1665 | broadcast_signal(SIGTERM, false, true, arg_default_timeout_stop_usec); |
3c7878f9 LP |
1666 | |
1667 | /* And switch root with MS_MOVE, because we remove the old directory afterwards and detach it. */ | |
1668 | r = switch_root(switch_root_dir, "/mnt", true, MS_MOVE); | |
1669 | if (r < 0) | |
1670 | log_error_errno(r, "Failed to switch root, trying to continue: %m"); | |
1671 | } | |
1672 | ||
1673 | args_size = MAX(6, argc+1); | |
1674 | args = newa(const char*, args_size); | |
1675 | ||
1676 | if (!switch_root_init) { | |
1677 | char sfd[DECIMAL_STR_MAX(int) + 1]; | |
1678 | ||
1679 | /* First try to spawn ourselves with the right path, and with full serialization. We do this only if | |
1680 | * the user didn't specify an explicit init to spawn. */ | |
1681 | ||
1682 | assert(arg_serialization); | |
1683 | assert(fds); | |
1684 | ||
1685 | xsprintf(sfd, "%i", fileno(arg_serialization)); | |
1686 | ||
1687 | i = 0; | |
1688 | args[i++] = SYSTEMD_BINARY_PATH; | |
1689 | if (switch_root_dir) | |
1690 | args[i++] = "--switched-root"; | |
1691 | args[i++] = arg_system ? "--system" : "--user"; | |
1692 | args[i++] = "--deserialize"; | |
1693 | args[i++] = sfd; | |
1694 | args[i++] = NULL; | |
1695 | ||
1696 | assert(i <= args_size); | |
1697 | ||
1698 | /* | |
1699 | * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do this is on | |
1700 | * its own on exec(), but it will do it on exit(). Hence, to ensure we get a summary here, fork() off | |
1701 | * a child, let it exit() cleanly, so that it prints the summary, and wait() for it in the parent, | |
1702 | * before proceeding into the exec(). | |
1703 | */ | |
1704 | valgrind_summary_hack(); | |
1705 | ||
1706 | (void) execv(args[0], (char* const*) args); | |
1707 | log_debug_errno(errno, "Failed to execute our own binary, trying fallback: %m"); | |
1708 | } | |
1709 | ||
1710 | /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and envp[]. (Well, | |
1711 | * modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], but let's hope that | |
1712 | * doesn't matter.) */ | |
1713 | ||
1714 | arg_serialization = safe_fclose(arg_serialization); | |
1715 | fds = fdset_free(fds); | |
1716 | ||
1717 | /* Reopen the console */ | |
1718 | (void) make_console_stdio(); | |
1719 | ||
1720 | for (j = 1, i = 1; j < (unsigned) argc; j++) | |
1721 | args[i++] = argv[j]; | |
1722 | args[i++] = NULL; | |
1723 | assert(i <= args_size); | |
1724 | ||
1725 | /* Reenable any blocked signals, especially important if we switch from initial ramdisk to init=... */ | |
1726 | (void) reset_all_signal_handlers(); | |
1727 | (void) reset_signal_mask(); | |
1728 | ||
1729 | if (switch_root_init) { | |
1730 | args[0] = switch_root_init; | |
1731 | (void) execv(args[0], (char* const*) args); | |
1732 | log_warning_errno(errno, "Failed to execute configured init, trying fallback: %m"); | |
1733 | } | |
1734 | ||
1735 | args[0] = "/sbin/init"; | |
1736 | (void) execv(args[0], (char* const*) args); | |
1737 | r = -errno; | |
1738 | ||
1739 | manager_status_printf(NULL, STATUS_TYPE_EMERGENCY, | |
1740 | ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL, | |
1741 | "Failed to execute /sbin/init"); | |
1742 | ||
1743 | if (r == -ENOENT) { | |
1744 | log_warning("No /sbin/init, trying fallback"); | |
1745 | ||
1746 | args[0] = "/bin/sh"; | |
1747 | args[1] = NULL; | |
1748 | (void) execv(args[0], (char* const*) args); | |
1749 | log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m"); | |
1750 | } else | |
1751 | log_warning_errno(r, "Failed to execute /sbin/init, giving up: %m"); | |
1752 | ||
1753 | *ret_error_message = "Failed to execute fallback shell"; | |
1754 | } | |
1755 | ||
7eb35049 LP |
1756 | static int invoke_main_loop( |
1757 | Manager *m, | |
1758 | bool *ret_reexecute, | |
1759 | int *ret_retval, /* Return parameters relevant for shutting down */ | |
1760 | const char **ret_shutdown_verb, /* … */ | |
1761 | FDSet **ret_fds, /* Return parameters for reexecuting */ | |
1762 | char **ret_switch_root_dir, /* … */ | |
1763 | char **ret_switch_root_init, /* … */ | |
1764 | const char **ret_error_message) { | |
1765 | ||
1766 | int r; | |
1767 | ||
1768 | assert(m); | |
1769 | assert(ret_reexecute); | |
1770 | assert(ret_retval); | |
1771 | assert(ret_shutdown_verb); | |
1772 | assert(ret_fds); | |
1773 | assert(ret_switch_root_dir); | |
1774 | assert(ret_switch_root_init); | |
1775 | assert(ret_error_message); | |
1776 | ||
1777 | for (;;) { | |
1778 | r = manager_loop(m); | |
1779 | if (r < 0) { | |
1780 | *ret_error_message = "Failed to run main loop"; | |
1781 | return log_emergency_errno(r, "Failed to run main loop: %m"); | |
1782 | } | |
1783 | ||
3ca4d0b3 | 1784 | switch ((ManagerObjective) r) { |
7eb35049 | 1785 | |
a6ecbf83 | 1786 | case MANAGER_RELOAD: { |
bda7d78b | 1787 | LogTarget saved_log_target; |
a6ecbf83 FB |
1788 | int saved_log_level; |
1789 | ||
7eb35049 LP |
1790 | log_info("Reloading."); |
1791 | ||
3fe91079 | 1792 | /* First, save any overridden log level/target, then parse the configuration file, which might |
bda7d78b FB |
1793 | * change the log level to new settings. */ |
1794 | ||
a6ecbf83 | 1795 | saved_log_level = m->log_level_overridden ? log_get_max_level() : -1; |
bda7d78b | 1796 | saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID; |
a6ecbf83 | 1797 | |
7eb35049 LP |
1798 | r = parse_config_file(); |
1799 | if (r < 0) | |
1800 | log_warning_errno(r, "Failed to parse config file, ignoring: %m"); | |
1801 | ||
1802 | set_manager_defaults(m); | |
1803 | ||
a6ecbf83 FB |
1804 | if (saved_log_level >= 0) |
1805 | manager_override_log_level(m, saved_log_level); | |
bda7d78b FB |
1806 | if (saved_log_target >= 0) |
1807 | manager_override_log_target(m, saved_log_target); | |
a6ecbf83 | 1808 | |
7eb35049 LP |
1809 | r = manager_reload(m); |
1810 | if (r < 0) | |
7a35fa24 LP |
1811 | /* Reloading failed before the point of no return. Let's continue running as if nothing happened. */ |
1812 | m->objective = MANAGER_OK; | |
7eb35049 LP |
1813 | |
1814 | break; | |
a6ecbf83 | 1815 | } |
7eb35049 LP |
1816 | |
1817 | case MANAGER_REEXECUTE: | |
1818 | ||
1819 | r = prepare_reexecute(m, &arg_serialization, ret_fds, false); | |
1820 | if (r < 0) { | |
1821 | *ret_error_message = "Failed to prepare for reexecution"; | |
1822 | return r; | |
1823 | } | |
1824 | ||
1825 | log_notice("Reexecuting."); | |
1826 | ||
1827 | *ret_reexecute = true; | |
1828 | *ret_retval = EXIT_SUCCESS; | |
1829 | *ret_shutdown_verb = NULL; | |
1830 | *ret_switch_root_dir = *ret_switch_root_init = NULL; | |
1831 | ||
1832 | return 0; | |
1833 | ||
1834 | case MANAGER_SWITCH_ROOT: | |
1835 | if (!m->switch_root_init) { | |
1836 | r = prepare_reexecute(m, &arg_serialization, ret_fds, true); | |
1837 | if (r < 0) { | |
1838 | *ret_error_message = "Failed to prepare for reexecution"; | |
1839 | return r; | |
1840 | } | |
1841 | } else | |
1842 | *ret_fds = NULL; | |
1843 | ||
1844 | log_notice("Switching root."); | |
1845 | ||
1846 | *ret_reexecute = true; | |
1847 | *ret_retval = EXIT_SUCCESS; | |
1848 | *ret_shutdown_verb = NULL; | |
1849 | ||
1850 | /* Steal the switch root parameters */ | |
1851 | *ret_switch_root_dir = m->switch_root; | |
1852 | *ret_switch_root_init = m->switch_root_init; | |
1853 | m->switch_root = m->switch_root_init = NULL; | |
1854 | ||
1855 | return 0; | |
1856 | ||
1857 | case MANAGER_EXIT: | |
1858 | ||
1859 | if (MANAGER_IS_USER(m)) { | |
1860 | log_debug("Exit."); | |
1861 | ||
1862 | *ret_reexecute = false; | |
1863 | *ret_retval = m->return_value; | |
1864 | *ret_shutdown_verb = NULL; | |
1865 | *ret_fds = NULL; | |
1866 | *ret_switch_root_dir = *ret_switch_root_init = NULL; | |
1867 | ||
1868 | return 0; | |
1869 | } | |
1870 | ||
1871 | _fallthrough_; | |
1872 | case MANAGER_REBOOT: | |
1873 | case MANAGER_POWEROFF: | |
1874 | case MANAGER_HALT: | |
1875 | case MANAGER_KEXEC: { | |
af41e508 LP |
1876 | static const char * const table[_MANAGER_OBJECTIVE_MAX] = { |
1877 | [MANAGER_EXIT] = "exit", | |
1878 | [MANAGER_REBOOT] = "reboot", | |
7eb35049 | 1879 | [MANAGER_POWEROFF] = "poweroff", |
af41e508 LP |
1880 | [MANAGER_HALT] = "halt", |
1881 | [MANAGER_KEXEC] = "kexec", | |
7eb35049 LP |
1882 | }; |
1883 | ||
1884 | log_notice("Shutting down."); | |
1885 | ||
1886 | *ret_reexecute = false; | |
1887 | *ret_retval = m->return_value; | |
af41e508 | 1888 | assert_se(*ret_shutdown_verb = table[m->objective]); |
7eb35049 LP |
1889 | *ret_fds = NULL; |
1890 | *ret_switch_root_dir = *ret_switch_root_init = NULL; | |
1891 | ||
1892 | return 0; | |
1893 | } | |
1894 | ||
1895 | default: | |
af41e508 | 1896 | assert_not_reached("Unknown or unexpected manager objective."); |
7eb35049 LP |
1897 | } |
1898 | } | |
1899 | } | |
1900 | ||
31aef7ff LP |
1901 | static void log_execution_mode(bool *ret_first_boot) { |
1902 | assert(ret_first_boot); | |
1903 | ||
1904 | if (arg_system) { | |
1905 | int v; | |
1906 | ||
1907 | log_info(PACKAGE_STRING " running in %ssystem mode. (" SYSTEMD_FEATURES ")", | |
1908 | arg_action == ACTION_TEST ? "test " : "" ); | |
1909 | ||
1910 | v = detect_virtualization(); | |
1911 | if (v > 0) | |
1912 | log_info("Detected virtualization %s.", virtualization_to_string(v)); | |
1913 | ||
1914 | log_info("Detected architecture %s.", architecture_to_string(uname_architecture())); | |
1915 | ||
1916 | if (in_initrd()) { | |
1917 | *ret_first_boot = false; | |
1918 | log_info("Running in initial RAM disk."); | |
1919 | } else { | |
1920 | /* Let's check whether we are in first boot, i.e. whether /etc is still unpopulated. We use | |
1921 | * /etc/machine-id as flag file, for this: if it exists we assume /etc is populated, if it | |
1922 | * doesn't it's unpopulated. This allows container managers and installers to provision a | |
1923 | * couple of files already. If the container manager wants to provision the machine ID itself | |
1924 | * it should pass $container_uuid to PID 1. */ | |
1925 | ||
1926 | *ret_first_boot = access("/etc/machine-id", F_OK) < 0; | |
1927 | if (*ret_first_boot) | |
1928 | log_info("Running with unpopulated /etc."); | |
1929 | } | |
1930 | } else { | |
b9e90f3a LP |
1931 | if (DEBUG_LOGGING) { |
1932 | _cleanup_free_ char *t; | |
31aef7ff | 1933 | |
b9e90f3a LP |
1934 | t = uid_to_name(getuid()); |
1935 | log_debug(PACKAGE_STRING " running in %suser mode for user " UID_FMT "/%s. (" SYSTEMD_FEATURES ")", | |
1936 | arg_action == ACTION_TEST ? " test" : "", getuid(), strna(t)); | |
1937 | } | |
31aef7ff LP |
1938 | |
1939 | *ret_first_boot = false; | |
1940 | } | |
1941 | } | |
1942 | ||
5afbaa36 LP |
1943 | static int initialize_runtime( |
1944 | bool skip_setup, | |
1945 | struct rlimit *saved_rlimit_nofile, | |
1946 | struct rlimit *saved_rlimit_memlock, | |
1947 | const char **ret_error_message) { | |
1948 | ||
1949 | int r; | |
1950 | ||
1951 | assert(ret_error_message); | |
1952 | ||
1953 | /* Sets up various runtime parameters. Many of these initializations are conditionalized: | |
1954 | * | |
1955 | * - Some only apply to --system instances | |
1956 | * - Some only apply to --user instances | |
1957 | * - Some only apply when we first start up, but not when we reexecute | |
1958 | */ | |
1959 | ||
2d776038 LP |
1960 | if (arg_action != ACTION_RUN) |
1961 | return 0; | |
1962 | ||
3c3c6cb9 LP |
1963 | if (arg_system) { |
1964 | /* Make sure we leave a core dump without panicing the kernel. */ | |
1965 | install_crash_handler(); | |
5afbaa36 | 1966 | |
3c3c6cb9 | 1967 | if (!skip_setup) { |
143fadf3 | 1968 | r = mount_cgroup_controllers(); |
3c3c6cb9 LP |
1969 | if (r < 0) { |
1970 | *ret_error_message = "Failed to mount cgroup hierarchies"; | |
1971 | return r; | |
1972 | } | |
1973 | ||
1974 | status_welcome(); | |
1975 | hostname_setup(); | |
1976 | machine_id_setup(NULL, arg_machine_id, NULL); | |
1977 | loopback_setup(); | |
1978 | bump_unix_max_dgram_qlen(); | |
a8b627aa | 1979 | bump_file_max_and_nr_open(); |
3c3c6cb9 LP |
1980 | test_usr(); |
1981 | write_container_id(); | |
1982 | } | |
8a2c1fbf | 1983 | |
3c3c6cb9 LP |
1984 | if (arg_watchdog_device) { |
1985 | r = watchdog_set_device(arg_watchdog_device); | |
1986 | if (r < 0) | |
1987 | log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device); | |
1988 | } | |
1989 | ||
1990 | if (arg_runtime_watchdog > 0 && arg_runtime_watchdog != USEC_INFINITY) | |
1991 | watchdog_set_timeout(&arg_runtime_watchdog); | |
1992 | } | |
5afbaa36 LP |
1993 | |
1994 | if (arg_timer_slack_nsec != NSEC_INFINITY) | |
1995 | if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0) | |
3a671cd1 | 1996 | log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m"); |
5afbaa36 LP |
1997 | |
1998 | if (arg_system && !cap_test_all(arg_capability_bounding_set)) { | |
1999 | r = capability_bounding_set_drop_usermode(arg_capability_bounding_set); | |
2000 | if (r < 0) { | |
2001 | *ret_error_message = "Failed to drop capability bounding set of usermode helpers"; | |
2002 | return log_emergency_errno(r, "Failed to drop capability bounding set of usermode helpers: %m"); | |
2003 | } | |
2004 | ||
2005 | r = capability_bounding_set_drop(arg_capability_bounding_set, true); | |
2006 | if (r < 0) { | |
2007 | *ret_error_message = "Failed to drop capability bounding set"; | |
2008 | return log_emergency_errno(r, "Failed to drop capability bounding set: %m"); | |
2009 | } | |
2010 | } | |
2011 | ||
39362f6f JB |
2012 | if (arg_system && arg_no_new_privs) { |
2013 | if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) { | |
2014 | *ret_error_message = "Failed to disable new privileges"; | |
2015 | return log_emergency_errno(errno, "Failed to disable new privileges: %m"); | |
2016 | } | |
2017 | } | |
2018 | ||
5afbaa36 LP |
2019 | if (arg_syscall_archs) { |
2020 | r = enforce_syscall_archs(arg_syscall_archs); | |
2021 | if (r < 0) { | |
2022 | *ret_error_message = "Failed to set syscall architectures"; | |
2023 | return r; | |
2024 | } | |
2025 | } | |
2026 | ||
2027 | if (!arg_system) | |
2028 | /* Become reaper of our children */ | |
2029 | if (prctl(PR_SET_CHILD_SUBREAPER, 1) < 0) | |
2030 | log_warning_errno(errno, "Failed to make us a subreaper: %m"); | |
2031 | ||
a17c1712 LP |
2032 | /* Bump up RLIMIT_NOFILE for systemd itself */ |
2033 | (void) bump_rlimit_nofile(saved_rlimit_nofile); | |
2034 | (void) bump_rlimit_memlock(saved_rlimit_memlock); | |
5afbaa36 LP |
2035 | |
2036 | return 0; | |
2037 | } | |
2038 | ||
6acca5fc LP |
2039 | static int do_queue_default_job( |
2040 | Manager *m, | |
2041 | const char **ret_error_message) { | |
2042 | ||
2043 | _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; | |
2044 | Job *default_unit_job; | |
2045 | Unit *target = NULL; | |
2046 | int r; | |
2047 | ||
2048 | log_debug("Activating default unit: %s", arg_default_unit); | |
2049 | ||
4109ede7 ZJS |
2050 | r = manager_load_startable_unit_or_warn(m, arg_default_unit, NULL, &target); |
2051 | if (r < 0) { | |
2052 | log_info("Falling back to rescue target: " SPECIAL_RESCUE_TARGET); | |
6acca5fc | 2053 | |
4109ede7 | 2054 | r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target); |
6acca5fc | 2055 | if (r < 0) { |
4109ede7 ZJS |
2056 | *ret_error_message = r == -ERFKILL ? "Rescue target masked" |
2057 | : "Failed to load rescue target"; | |
2058 | return r; | |
6acca5fc LP |
2059 | } |
2060 | } | |
2061 | ||
2062 | assert(target->load_state == UNIT_LOADED); | |
2063 | ||
2064 | r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, &error, &default_unit_job); | |
2065 | if (r == -EPERM) { | |
2066 | log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r)); | |
2067 | ||
2068 | sd_bus_error_free(&error); | |
2069 | ||
2070 | r = manager_add_job(m, JOB_START, target, JOB_REPLACE, &error, &default_unit_job); | |
2071 | if (r < 0) { | |
2072 | *ret_error_message = "Failed to start default target"; | |
2073 | return log_emergency_errno(r, "Failed to start default target: %s", bus_error_message(&error, r)); | |
2074 | } | |
2075 | ||
2076 | } else if (r < 0) { | |
2077 | *ret_error_message = "Failed to isolate default target"; | |
2078 | return log_emergency_errno(r, "Failed to isolate default target: %s", bus_error_message(&error, r)); | |
2079 | } | |
2080 | ||
2081 | m->default_unit_job_id = default_unit_job->id; | |
2082 | ||
2083 | return 0; | |
2084 | } | |
2085 | ||
970777b5 | 2086 | static void free_arguments(void) { |
970777b5 LP |
2087 | |
2088 | /* Frees all arg_* variables, with the exception of arg_serialization */ | |
ef59409f | 2089 | rlimit_free_all(arg_default_rlimit); |
970777b5 LP |
2090 | |
2091 | arg_default_unit = mfree(arg_default_unit); | |
2092 | arg_confirm_spawn = mfree(arg_confirm_spawn); | |
970777b5 LP |
2093 | arg_default_environment = strv_free(arg_default_environment); |
2094 | arg_syscall_archs = set_free(arg_syscall_archs); | |
2095 | } | |
2096 | ||
97d1fb94 LP |
2097 | static int load_configuration(int argc, char **argv, const char **ret_error_message) { |
2098 | int r; | |
2099 | ||
2100 | assert(ret_error_message); | |
2101 | ||
2102 | arg_default_tasks_max = system_tasks_max_scale(DEFAULT_TASKS_MAX_PERCENTAGE, 100U); | |
2103 | ||
2104 | r = parse_config_file(); | |
2105 | if (r < 0) { | |
2106 | *ret_error_message = "Failed to parse config file"; | |
2107 | return r; | |
2108 | } | |
2109 | ||
2110 | if (arg_system) { | |
2111 | r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0); | |
2112 | if (r < 0) | |
2113 | log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m"); | |
2114 | } | |
2115 | ||
2116 | /* Note that this also parses bits from the kernel command line, including "debug". */ | |
2117 | log_parse_environment(); | |
2118 | ||
2119 | r = parse_argv(argc, argv); | |
2120 | if (r < 0) { | |
2121 | *ret_error_message = "Failed to parse commandline arguments"; | |
2122 | return r; | |
2123 | } | |
2124 | ||
2125 | /* Initialize default unit */ | |
2126 | if (!arg_default_unit) { | |
2127 | arg_default_unit = strdup(SPECIAL_DEFAULT_TARGET); | |
2128 | if (!arg_default_unit) { | |
2129 | *ret_error_message = "Failed to set default unit"; | |
2130 | return log_oom(); | |
2131 | } | |
2132 | } | |
2133 | ||
db33214b | 2134 | /* Initialize the show status setting if it hasn't been set explicitly yet */ |
7a293242 | 2135 | if (arg_show_status == _SHOW_STATUS_INVALID) |
db33214b LP |
2136 | arg_show_status = SHOW_STATUS_YES; |
2137 | ||
97d1fb94 LP |
2138 | return 0; |
2139 | } | |
2140 | ||
b0d7c989 LP |
2141 | static int safety_checks(void) { |
2142 | ||
febf46a4 LP |
2143 | if (getpid_cached() == 1 && |
2144 | arg_action != ACTION_RUN) { | |
2145 | log_error("Unsupported execution mode while PID 1."); | |
2146 | return -EPERM; | |
2147 | } | |
2148 | ||
2149 | if (getpid_cached() == 1 && | |
2150 | !arg_system) { | |
2151 | log_error("Can't run --user mode as PID 1."); | |
2152 | return -EPERM; | |
2153 | } | |
2154 | ||
2155 | if (arg_action == ACTION_RUN && | |
2156 | arg_system && | |
2157 | getpid_cached() != 1) { | |
2158 | log_error("Can't run system mode unless PID 1."); | |
2159 | return -EPERM; | |
2160 | } | |
2161 | ||
b0d7c989 LP |
2162 | if (arg_action == ACTION_TEST && |
2163 | geteuid() == 0) { | |
2164 | log_error("Don't run test mode as root."); | |
2165 | return -EPERM; | |
2166 | } | |
2167 | ||
2168 | if (!arg_system && | |
2169 | arg_action == ACTION_RUN && | |
2170 | sd_booted() <= 0) { | |
2171 | log_error("Trying to run as user instance, but the system has not been booted with systemd."); | |
2172 | return -EOPNOTSUPP; | |
2173 | } | |
2174 | ||
2175 | if (!arg_system && | |
2176 | arg_action == ACTION_RUN && | |
2177 | !getenv("XDG_RUNTIME_DIR")) { | |
2178 | log_error("Trying to run as user instance, but $XDG_RUNTIME_DIR is not set."); | |
2179 | return -EUNATCH; | |
2180 | } | |
2181 | ||
2182 | if (arg_system && | |
2183 | arg_action == ACTION_RUN && | |
2184 | running_in_chroot() > 0) { | |
2185 | log_error("Cannot be run in a chroot() environment."); | |
2186 | return -EOPNOTSUPP; | |
2187 | } | |
2188 | ||
2189 | return 0; | |
2190 | } | |
2191 | ||
74da609f LP |
2192 | static int initialize_security( |
2193 | bool *loaded_policy, | |
2194 | dual_timestamp *security_start_timestamp, | |
2195 | dual_timestamp *security_finish_timestamp, | |
2196 | const char **ret_error_message) { | |
2197 | ||
2198 | int r; | |
2199 | ||
2200 | assert(loaded_policy); | |
2201 | assert(security_start_timestamp); | |
2202 | assert(security_finish_timestamp); | |
2203 | assert(ret_error_message); | |
2204 | ||
2205 | dual_timestamp_get(security_start_timestamp); | |
2206 | ||
97149f40 | 2207 | r = mac_selinux_setup(loaded_policy); |
74da609f LP |
2208 | if (r < 0) { |
2209 | *ret_error_message = "Failed to load SELinux policy"; | |
2210 | return r; | |
2211 | } | |
2212 | ||
2213 | r = mac_smack_setup(loaded_policy); | |
2214 | if (r < 0) { | |
2215 | *ret_error_message = "Failed to load SMACK policy"; | |
2216 | return r; | |
2217 | } | |
2218 | ||
2219 | r = ima_setup(); | |
2220 | if (r < 0) { | |
2221 | *ret_error_message = "Failed to load IMA policy"; | |
2222 | return r; | |
2223 | } | |
2224 | ||
2225 | dual_timestamp_get(security_finish_timestamp); | |
2226 | return 0; | |
2227 | } | |
2228 | ||
263162da LP |
2229 | static void test_summary(Manager *m) { |
2230 | assert(m); | |
2231 | ||
2232 | printf("-> By units:\n"); | |
2233 | manager_dump_units(m, stdout, "\t"); | |
2234 | ||
2235 | printf("-> By jobs:\n"); | |
2236 | manager_dump_jobs(m, stdout, "\t"); | |
2237 | } | |
2238 | ||
efeb853f LP |
2239 | static int collect_fds(FDSet **ret_fds, const char **ret_error_message) { |
2240 | int r; | |
2241 | ||
2242 | assert(ret_fds); | |
2243 | assert(ret_error_message); | |
2244 | ||
2245 | r = fdset_new_fill(ret_fds); | |
2246 | if (r < 0) { | |
2247 | *ret_error_message = "Failed to allocate fd set"; | |
2248 | return log_emergency_errno(r, "Failed to allocate fd set: %m"); | |
2249 | } | |
2250 | ||
2251 | fdset_cloexec(*ret_fds, true); | |
2252 | ||
2253 | if (arg_serialization) | |
2254 | assert_se(fdset_remove(*ret_fds, fileno(arg_serialization)) >= 0); | |
2255 | ||
2256 | return 0; | |
2257 | } | |
2258 | ||
2e51b31c LP |
2259 | static void setup_console_terminal(bool skip_setup) { |
2260 | ||
2261 | if (!arg_system) | |
2262 | return; | |
2263 | ||
2264 | /* Become a session leader if we aren't one yet. */ | |
2265 | (void) setsid(); | |
2266 | ||
2267 | /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a controlling | |
2268 | * tty. */ | |
2269 | (void) release_terminal(); | |
2270 | ||
2271 | /* Reset the console, but only if this is really init and we are freshly booted */ | |
2272 | if (getpid_cached() == 1 && !skip_setup) | |
2273 | (void) console_setup(); | |
2274 | } | |
2275 | ||
aa40ff07 LP |
2276 | static bool early_skip_setup_check(int argc, char *argv[]) { |
2277 | bool found_deserialize = false; | |
2278 | int i; | |
2279 | ||
2280 | /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much later, so | |
2281 | * let's just have a quick peek here. Note that if we have switched root, do all the special setup things | |
2282 | * anyway, even if in that case we also do deserialization. */ | |
2283 | ||
2284 | for (i = 1; i < argc; i++) { | |
aa40ff07 LP |
2285 | if (streq(argv[i], "--switched-root")) |
2286 | return false; /* If we switched root, don't skip the setup. */ | |
2287 | else if (streq(argv[i], "--deserialize")) | |
2288 | found_deserialize = true; | |
2289 | } | |
2290 | ||
2291 | return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */ | |
2292 | } | |
2293 | ||
60918275 | 2294 | int main(int argc, char *argv[]) { |
625e8690 LP |
2295 | |
2296 | dual_timestamp initrd_timestamp = DUAL_TIMESTAMP_NULL, userspace_timestamp = DUAL_TIMESTAMP_NULL, kernel_timestamp = DUAL_TIMESTAMP_NULL, | |
2297 | security_start_timestamp = DUAL_TIMESTAMP_NULL, security_finish_timestamp = DUAL_TIMESTAMP_NULL; | |
2298 | struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), saved_rlimit_memlock = RLIMIT_MAKE_CONST((rlim_t) -1); | |
2299 | bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false, reexecute = false; | |
2300 | char *switch_root_dir = NULL, *switch_root_init = NULL; | |
9d76d730 | 2301 | usec_t before_startup, after_startup; |
625e8690 | 2302 | static char systemd[] = "systemd"; |
9d76d730 | 2303 | char timespan[FORMAT_TIMESPAN_MAX]; |
625e8690 LP |
2304 | const char *shutdown_verb = NULL, *error_message = NULL; |
2305 | int r, retval = EXIT_FAILURE; | |
2306 | Manager *m = NULL; | |
a16e1123 | 2307 | FDSet *fds = NULL; |
27b14a22 | 2308 | |
d72a8f10 | 2309 | /* SysV compatibility: redirect init → telinit */ |
6808a0bc | 2310 | redirect_telinit(argc, argv); |
2cb1a60d | 2311 | |
d72a8f10 | 2312 | /* Take timestamps early on */ |
c3a170f3 HH |
2313 | dual_timestamp_from_monotonic(&kernel_timestamp, 0); |
2314 | dual_timestamp_get(&userspace_timestamp); | |
2315 | ||
d72a8f10 LP |
2316 | /* Figure out whether we need to do initialize the system, or if we already did that because we are |
2317 | * reexecuting */ | |
aa40ff07 | 2318 | skip_setup = early_skip_setup_check(argc, argv); |
d03bc1b8 | 2319 | |
d72a8f10 LP |
2320 | /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent reexecution we |
2321 | * are then called 'systemd'. That is confusing, hence let's call us systemd right-away. */ | |
f3b6a3ed | 2322 | program_invocation_short_name = systemd; |
eee8b7ab | 2323 | (void) prctl(PR_SET_NAME, systemd); |
5d6b1584 | 2324 | |
d72a8f10 | 2325 | /* Save the original command line */ |
9a0e6896 LP |
2326 | saved_argv = argv; |
2327 | saved_argc = argc; | |
f3b6a3ed | 2328 | |
6fdb8de4 | 2329 | /* Make sure that if the user says "syslog" we actually log to the journal. */ |
c1dc6153 | 2330 | log_set_upgrade_syslog_to_journal(true); |
bbe63281 | 2331 | |
df0ff127 | 2332 | if (getpid_cached() == 1) { |
b5752d23 LP |
2333 | /* When we run as PID 1 force system mode */ |
2334 | arg_system = true; | |
2335 | ||
48a601fe | 2336 | /* Disable the umask logic */ |
90dc8c2e MG |
2337 | umask(0); |
2338 | ||
92890452 LP |
2339 | /* Make sure that at least initially we do not ever log to journald/syslogd, because it might not be |
2340 | * activated yet (even though the log socket for it exists). */ | |
d075092f LP |
2341 | log_set_prohibit_ipc(true); |
2342 | ||
48a601fe LP |
2343 | /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This is |
2344 | * important so that we never end up logging to any foreign stderr, for example if we have to log in a | |
2345 | * child process right before execve()'ing the actual binary, at a point in time where socket | |
2346 | * activation stderr/stdout area already set up. */ | |
2347 | log_set_always_reopen_console(true); | |
48a601fe | 2348 | |
92890452 | 2349 | if (detect_container() <= 0) { |
4f8d551f | 2350 | |
92890452 | 2351 | /* Running outside of a container as PID 1 */ |
92890452 LP |
2352 | log_set_target(LOG_TARGET_KMSG); |
2353 | log_open(); | |
a866073d | 2354 | |
92890452 LP |
2355 | if (in_initrd()) |
2356 | initrd_timestamp = userspace_timestamp; | |
c3ba6250 | 2357 | |
92890452 LP |
2358 | if (!skip_setup) { |
2359 | r = mount_setup_early(); | |
2360 | if (r < 0) { | |
2361 | error_message = "Failed to mount early API filesystems"; | |
2362 | goto finish; | |
2363 | } | |
2364 | ||
2365 | r = initialize_security( | |
2366 | &loaded_policy, | |
2367 | &security_start_timestamp, | |
2368 | &security_finish_timestamp, | |
2369 | &error_message); | |
2370 | if (r < 0) | |
2371 | goto finish; | |
d723cd65 | 2372 | } |
eee8b7ab | 2373 | |
92890452 LP |
2374 | if (mac_selinux_init() < 0) { |
2375 | error_message = "Failed to initialize SELinux policy"; | |
96694e99 | 2376 | goto finish; |
92890452 | 2377 | } |
0b3325e7 | 2378 | |
92890452 LP |
2379 | if (!skip_setup) |
2380 | initialize_clock(); | |
2381 | ||
2382 | /* Set the default for later on, but don't actually open the logs like this for now. Note that | |
2383 | * if we are transitioning from the initrd there might still be journal fd open, and we | |
2384 | * shouldn't attempt opening that before we parsed /proc/cmdline which might redirect output | |
2385 | * elsewhere. */ | |
2386 | log_set_target(LOG_TARGET_JOURNAL_OR_KMSG); | |
2387 | ||
2388 | } else { | |
2389 | /* Running inside a container, as PID 1 */ | |
92890452 LP |
2390 | log_set_target(LOG_TARGET_CONSOLE); |
2391 | log_open(); | |
2392 | ||
2393 | /* For later on, see above... */ | |
2394 | log_set_target(LOG_TARGET_JOURNAL); | |
2395 | ||
2396 | /* clear the kernel timestamp, | |
2397 | * because we are in a container */ | |
2398 | kernel_timestamp = DUAL_TIMESTAMP_NULL; | |
cb6531be | 2399 | } |
7948c4df | 2400 | |
92890452 | 2401 | initialize_coredump(skip_setup); |
a866073d | 2402 | |
92890452 LP |
2403 | r = fixup_environment(); |
2404 | if (r < 0) { | |
2405 | log_emergency_errno(r, "Failed to fix up PID 1 environment: %m"); | |
2406 | error_message = "Failed to fix up PID1 environment"; | |
2407 | goto finish; | |
2408 | } | |
a866073d | 2409 | |
c3a170f3 | 2410 | } else { |
a866073d | 2411 | /* Running as user instance */ |
463d0d15 | 2412 | arg_system = false; |
eeecf6e6 | 2413 | log_set_target(LOG_TARGET_AUTO); |
871e5809 | 2414 | log_open(); |
c3a170f3 HH |
2415 | |
2416 | /* clear the kernel timestamp, | |
2417 | * because we are not PID 1 */ | |
6513d561 | 2418 | kernel_timestamp = DUAL_TIMESTAMP_NULL; |
bbe63281 | 2419 | } |
a5dab5ce | 2420 | |
6edefe0b | 2421 | if (arg_system) { |
92890452 LP |
2422 | /* Try to figure out if we can use colors with the console. No need to do that for user instances since |
2423 | * they never log into the console. */ | |
3a18b604 | 2424 | log_show_color(colors_enabled()); |
92890452 | 2425 | |
c76cf844 AK |
2426 | r = make_null_stdio(); |
2427 | if (r < 0) | |
92890452 | 2428 | log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m"); |
3a18b604 FB |
2429 | } |
2430 | ||
f170852a LP |
2431 | /* Mount /proc, /sys and friends, so that /proc/cmdline and |
2432 | * /proc/$PID/fd is available. */ | |
df0ff127 | 2433 | if (getpid_cached() == 1) { |
f84f9974 | 2434 | |
a132bef0 | 2435 | /* Load the kernel modules early. */ |
2e75e2a8 DM |
2436 | if (!skip_setup) |
2437 | kmod_setup(); | |
2e75e2a8 | 2438 | |
0c85a4f3 | 2439 | r = mount_setup(loaded_policy); |
cb6531be ZJS |
2440 | if (r < 0) { |
2441 | error_message = "Failed to mount API filesystems"; | |
8efe3c01 | 2442 | goto finish; |
cb6531be | 2443 | } |
0c85a4f3 | 2444 | } |
4ade7963 LP |
2445 | |
2446 | /* Reset all signal handlers. */ | |
ce30c8dc LP |
2447 | (void) reset_all_signal_handlers(); |
2448 | (void) ignore_signals(SIGNALS_IGNORE, -1); | |
078e4539 | 2449 | |
97d1fb94 LP |
2450 | r = load_configuration(argc, argv, &error_message); |
2451 | if (r < 0) | |
f170852a | 2452 | goto finish; |
10c961b9 | 2453 | |
b0d7c989 LP |
2454 | r = safety_checks(); |
2455 | if (r < 0) | |
fe783b03 | 2456 | goto finish; |
fe783b03 | 2457 | |
bbc1acab | 2458 | if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES)) |
0221d68a | 2459 | (void) pager_open(arg_pager_flags); |
b0d7c989 LP |
2460 | |
2461 | if (arg_action != ACTION_RUN) | |
74e7579c | 2462 | skip_setup = true; |
b87c2aa6 | 2463 | |
fa0f4d8a | 2464 | if (arg_action == ACTION_HELP) { |
37ec0fdd | 2465 | retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS; |
f170852a | 2466 | goto finish; |
9ba0bc4e ZJS |
2467 | } else if (arg_action == ACTION_VERSION) { |
2468 | retval = version(); | |
2469 | goto finish; | |
fa0f4d8a | 2470 | } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) { |
e537352b | 2471 | unit_dump_config_items(stdout); |
22f4096c | 2472 | retval = EXIT_SUCCESS; |
e537352b | 2473 | goto finish; |
bbc1acab YW |
2474 | } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) { |
2475 | dump_bus_properties(stdout); | |
2476 | retval = EXIT_SUCCESS; | |
2477 | goto finish; | |
f170852a LP |
2478 | } |
2479 | ||
4c701096 | 2480 | assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST)); |
f170852a | 2481 | |
5a2e0c62 LP |
2482 | /* Move out of the way, so that we won't block unmounts */ |
2483 | assert_se(chdir("/") == 0); | |
2484 | ||
dea374e8 | 2485 | if (arg_action == ACTION_RUN) { |
a70c72a0 | 2486 | |
c6885f5f FB |
2487 | /* A core pattern might have been specified via the cmdline. */ |
2488 | initialize_core_pattern(skip_setup); | |
2489 | ||
efeb853f | 2490 | /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */ |
a70c72a0 LP |
2491 | log_close(); |
2492 | ||
2493 | /* Remember open file descriptors for later deserialization */ | |
efeb853f LP |
2494 | r = collect_fds(&fds, &error_message); |
2495 | if (r < 0) | |
dea374e8 | 2496 | goto finish; |
a16e1123 | 2497 | |
2e51b31c LP |
2498 | /* Give up any control of the console, but make sure its initialized. */ |
2499 | setup_console_terminal(skip_setup); | |
56d96fc0 | 2500 | |
a70c72a0 LP |
2501 | /* Open the logging devices, if possible and necessary */ |
2502 | log_open(); | |
56d96fc0 | 2503 | } |
4ade7963 | 2504 | |
31aef7ff | 2505 | log_execution_mode(&first_boot); |
a5dab5ce | 2506 | |
2d776038 LP |
2507 | r = initialize_runtime(skip_setup, |
2508 | &saved_rlimit_nofile, | |
2509 | &saved_rlimit_memlock, | |
2510 | &error_message); | |
2511 | if (r < 0) | |
2512 | goto finish; | |
4096d6f5 | 2513 | |
e0a3da1f ZJS |
2514 | r = manager_new(arg_system ? UNIT_FILE_SYSTEM : UNIT_FILE_USER, |
2515 | arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0, | |
2516 | &m); | |
e96d6be7 | 2517 | if (r < 0) { |
da927ba9 | 2518 | log_emergency_errno(r, "Failed to allocate manager object: %m"); |
cb6531be | 2519 | error_message = "Failed to allocate manager object"; |
60918275 LP |
2520 | goto finish; |
2521 | } | |
2522 | ||
9f9f0342 LP |
2523 | m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp; |
2524 | m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp; | |
2525 | m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp; | |
d4ee7bd8 YW |
2526 | m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp; |
2527 | m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp; | |
9e58ff9c | 2528 | |
85cb4151 | 2529 | set_manager_defaults(m); |
7b46fc6a | 2530 | set_manager_settings(m); |
fd130612 | 2531 | manager_set_first_boot(m, first_boot); |
27d340c7 | 2532 | |
bf4df7c3 | 2533 | /* Remember whether we should queue the default job */ |
d3b1c508 | 2534 | queue_default_job = !arg_serialization || arg_switched_root; |
bf4df7c3 | 2535 | |
9d76d730 LP |
2536 | before_startup = now(CLOCK_MONOTONIC); |
2537 | ||
d3b1c508 | 2538 | r = manager_startup(m, arg_serialization, fds); |
58f88d92 | 2539 | if (r < 0) { |
cefb3eda | 2540 | error_message = "Failed to start up manager"; |
58f88d92 ZJS |
2541 | goto finish; |
2542 | } | |
a16e1123 | 2543 | |
6acca5fc | 2544 | /* This will close all file descriptors that were opened, but not claimed by any unit. */ |
2feceb5e | 2545 | fds = fdset_free(fds); |
74ca738f | 2546 | arg_serialization = safe_fclose(arg_serialization); |
bf4df7c3 LP |
2547 | |
2548 | if (queue_default_job) { | |
6acca5fc | 2549 | r = do_queue_default_job(m, &error_message); |
718db961 | 2550 | if (r < 0) |
37d88da7 | 2551 | goto finish; |
6acca5fc | 2552 | } |
ab17a050 | 2553 | |
6acca5fc | 2554 | after_startup = now(CLOCK_MONOTONIC); |
60918275 | 2555 | |
6acca5fc LP |
2556 | log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG, |
2557 | "Loaded units and determined initial transaction in %s.", | |
2558 | format_timespan(timespan, sizeof(timespan), after_startup - before_startup, 100 * USEC_PER_MSEC)); | |
07672f49 | 2559 | |
6acca5fc | 2560 | if (arg_action == ACTION_TEST) { |
263162da | 2561 | test_summary(m); |
6acca5fc LP |
2562 | retval = EXIT_SUCCESS; |
2563 | goto finish; | |
e965d56d | 2564 | } |
d46de8a1 | 2565 | |
3046b6db LP |
2566 | (void) invoke_main_loop(m, |
2567 | &reexecute, | |
2568 | &retval, | |
2569 | &shutdown_verb, | |
2570 | &fds, | |
2571 | &switch_root_dir, | |
2572 | &switch_root_init, | |
2573 | &error_message); | |
f170852a | 2574 | |
60918275 | 2575 | finish: |
b87c2aa6 ZJS |
2576 | pager_close(); |
2577 | ||
92890452 | 2578 | if (m) { |
81f5fc2d | 2579 | arg_shutdown_watchdog = m->shutdown_watchdog; |
92890452 LP |
2580 | m = manager_free(m); |
2581 | } | |
60918275 | 2582 | |
970777b5 | 2583 | free_arguments(); |
cc56fafe | 2584 | mac_selinux_finish(); |
b2bb3dbe | 2585 | |
3c7878f9 LP |
2586 | if (reexecute) |
2587 | do_reexecute(argc, argv, | |
2588 | &saved_rlimit_nofile, | |
2589 | &saved_rlimit_memlock, | |
2590 | fds, | |
2591 | switch_root_dir, | |
2592 | switch_root_init, | |
2593 | &error_message); /* This only returns if reexecution failed */ | |
a16e1123 | 2594 | |
74ca738f | 2595 | arg_serialization = safe_fclose(arg_serialization); |
2feceb5e | 2596 | fds = fdset_free(fds); |
a16e1123 | 2597 | |
349cc4a5 | 2598 | #if HAVE_VALGRIND_VALGRIND_H |
54b434b1 LP |
2599 | /* If we are PID 1 and running under valgrind, then let's exit |
2600 | * here explicitly. valgrind will only generate nice output on | |
2601 | * exit(), not on exec(), hence let's do the former not the | |
2602 | * latter here. */ | |
8a2c1fbf EJ |
2603 | if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) { |
2604 | /* Cleanup watchdog_device strings for valgrind. We need them | |
2605 | * in become_shutdown() so normally we cannot free them yet. */ | |
2606 | watchdog_free_device(); | |
2607 | arg_watchdog_device = mfree(arg_watchdog_device); | |
27fe58b7 | 2608 | return retval; |
8a2c1fbf | 2609 | } |
54b434b1 LP |
2610 | #endif |
2611 | ||
b9080b03 | 2612 | if (shutdown_verb) { |
7eb35049 | 2613 | r = become_shutdown(shutdown_verb, retval); |
4a36297c | 2614 | log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting"); |
9b9881d7 | 2615 | error_message = "Failed to execute shutdown binary"; |
b9080b03 FF |
2616 | } |
2617 | ||
8a2c1fbf EJ |
2618 | watchdog_free_device(); |
2619 | arg_watchdog_device = mfree(arg_watchdog_device); | |
2620 | ||
df0ff127 | 2621 | if (getpid_cached() == 1) { |
cb6531be ZJS |
2622 | if (error_message) |
2623 | manager_status_printf(NULL, STATUS_TYPE_EMERGENCY, | |
1fc464f6 | 2624 | ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL, |
cb6531be | 2625 | "%s, freezing.", error_message); |
b9e74c39 | 2626 | freeze_or_reboot(); |
cb6531be | 2627 | } |
c3b3c274 | 2628 | |
60918275 LP |
2629 | return retval; |
2630 | } |