]>
Commit | Line | Data |
---|---|---|
db9ecf05 | 1 | /* SPDX-License-Identifier: LGPL-2.1-or-later */ |
a7334b09 | 2 | |
60918275 | 3 | #include <errno.h> |
3dfc9763 | 4 | #include <fcntl.h> |
f170852a | 5 | #include <getopt.h> |
d4a402e4 | 6 | #include <linux/oom.h> |
664f88a7 | 7 | #include <sys/mount.h> |
3dfc9763 | 8 | #include <sys/prctl.h> |
40efaaed | 9 | #include <sys/utsname.h> |
3dfc9763 | 10 | #include <unistd.h> |
349cc4a5 | 11 | #if HAVE_VALGRIND_VALGRIND_H |
50b35193 | 12 | # include <valgrind/valgrind.h> |
3dfc9763 | 13 | #endif |
54b434b1 | 14 | |
718db961 | 15 | #include "sd-bus.h" |
cf0fbc49 | 16 | #include "sd-daemon.h" |
b2e7486c | 17 | #include "sd-messages.h" |
3dfc9763 | 18 | |
b5efdb8a | 19 | #include "alloc-util.h" |
2ffadd3c | 20 | #include "apparmor-setup.h" |
d9d93745 | 21 | #include "architecture.h" |
ee617a4e | 22 | #include "argv-util.h" |
b1994387 ILG |
23 | #if HAVE_LIBBPF |
24 | #include "bpf-lsm.h" | |
25 | #endif | |
3dfc9763 LP |
26 | #include "build.h" |
27 | #include "bus-error.h" | |
28 | #include "bus-util.h" | |
430f0182 | 29 | #include "capability-util.h" |
a88c5b8a | 30 | #include "cgroup-util.h" |
aaa27e2e | 31 | #include "chase.h" |
24efb112 | 32 | #include "clock-util.h" |
3dfc9763 | 33 | #include "conf-parser.h" |
024469dd | 34 | #include "confidential-virt.h" |
3f37a825 | 35 | #include "copy.h" |
618234a5 | 36 | #include "cpu-set-util.h" |
898c9a6f | 37 | #include "crash-handler.h" |
3dfc9763 | 38 | #include "dbus-manager.h" |
c18ecf03 | 39 | #include "dbus.h" |
28db6fbf | 40 | #include "constants.h" |
32429805 | 41 | #include "dev-setup.h" |
c18ecf03 | 42 | #include "efi-random.h" |
209b2592 | 43 | #include "efivars.h" |
eee8b7ab | 44 | #include "emergency-action.h" |
3dfc9763 | 45 | #include "env-util.h" |
57b7a260 | 46 | #include "exit-status.h" |
3ffd4af2 | 47 | #include "fd-util.h" |
3dfc9763 | 48 | #include "fdset.h" |
718db961 | 49 | #include "fileio.h" |
f97b34a6 | 50 | #include "format-util.h" |
f4f15635 | 51 | #include "fs-util.h" |
6339d3e6 | 52 | #include "getopt-defs.h" |
d247f232 | 53 | #include "hexdecoct.h" |
3dfc9763 LP |
54 | #include "hostname-setup.h" |
55 | #include "ima-setup.h" | |
4b9a4b01 | 56 | #include "import-creds.h" |
baa6a42d | 57 | #include "initrd-util.h" |
3dfc9763 LP |
58 | #include "killall.h" |
59 | #include "kmod-setup.h" | |
eefc66aa | 60 | #include "limits-util.h" |
d7b8eec7 | 61 | #include "load-fragment.h" |
3dfc9763 | 62 | #include "log.h" |
b6e66135 | 63 | #include "loopback-setup.h" |
b6e66135 | 64 | #include "machine-id-setup.h" |
898c9a6f | 65 | #include "main.h" |
3dfc9763 | 66 | #include "manager.h" |
2a341bb9 | 67 | #include "manager-dump.h" |
a01ba4b2 | 68 | #include "manager-serialize.h" |
35cd0ba5 | 69 | #include "mkdir-label.h" |
3dfc9763 | 70 | #include "mount-setup.h" |
ffc1ec73 | 71 | #include "mount-util.h" |
d58ad743 | 72 | #include "os-util.h" |
3dfc9763 | 73 | #include "pager.h" |
614b022c | 74 | #include "parse-argument.h" |
6bedfcbb | 75 | #include "parse-util.h" |
7d5ceb64 | 76 | #include "path-util.h" |
294bf0c3 | 77 | #include "pretty-print.h" |
4e731273 | 78 | #include "proc-cmdline.h" |
3dfc9763 | 79 | #include "process-util.h" |
6bb00842 | 80 | #include "psi-util.h" |
d247f232 | 81 | #include "random-util.h" |
78f22b97 | 82 | #include "rlimit-util.h" |
83f12b27 | 83 | #include "seccomp-util.h" |
b6e66135 | 84 | #include "selinux-setup.h" |
3dfc9763 LP |
85 | #include "selinux-util.h" |
86 | #include "signal-util.h" | |
ffbd2c4d | 87 | #include "smack-setup.h" |
3dfc9763 | 88 | #include "special.h" |
8fcde012 | 89 | #include "stat-util.h" |
15a5e950 | 90 | #include "stdio-util.h" |
3dfc9763 LP |
91 | #include "strv.h" |
92 | #include "switch-root.h" | |
a8b627aa | 93 | #include "sysctl-util.h" |
3dfc9763 | 94 | #include "terminal-util.h" |
b10abe4b | 95 | #include "time-util.h" |
8612da97 | 96 | #include "umask-util.h" |
b1d4f8e1 | 97 | #include "user-util.h" |
bdb577f5 | 98 | #include "version.h" |
3dfc9763 LP |
99 | #include "virt.h" |
100 | #include "watchdog.h" | |
b6e66135 | 101 | |
7e11a95e EV |
102 | #if HAS_FEATURE_ADDRESS_SANITIZER |
103 | #include <sanitizer/lsan_interface.h> | |
104 | #endif | |
105 | ||
f170852a LP |
106 | static enum { |
107 | ACTION_RUN, | |
e965d56d | 108 | ACTION_HELP, |
9ba0bc4e | 109 | ACTION_VERSION, |
e537352b | 110 | ACTION_TEST, |
bbc1acab YW |
111 | ACTION_DUMP_CONFIGURATION_ITEMS, |
112 | ACTION_DUMP_BUS_PROPERTIES, | |
5c08257b | 113 | ACTION_BUS_INTROSPECT, |
fa0f4d8a | 114 | } arg_action = ACTION_RUN; |
fb39af4c | 115 | |
5c08257b ZJS |
116 | static const char *arg_bus_introspect = NULL; |
117 | ||
45250e66 LP |
118 | /* Those variables are initialized to 0 automatically, so we avoid uninitialized memory access. Real |
119 | * defaults are assigned in reset_arguments() below. */ | |
fb39af4c | 120 | static char *arg_default_unit; |
4870133b | 121 | static RuntimeScope arg_runtime_scope; |
898c9a6f LP |
122 | bool arg_dump_core; |
123 | int arg_crash_chvt; | |
124 | bool arg_crash_shell; | |
125 | bool arg_crash_reboot; | |
fb39af4c ZJS |
126 | static char *arg_confirm_spawn; |
127 | static ShowStatus arg_show_status; | |
36cf4507 | 128 | static StatusUnitFormat arg_status_unit_format; |
fb39af4c ZJS |
129 | static bool arg_switched_root; |
130 | static PagerFlags arg_pager_flags; | |
131 | static bool arg_service_watchdogs; | |
c9e120e0 | 132 | static UnitDefaults arg_defaults; |
fb39af4c | 133 | static usec_t arg_runtime_watchdog; |
65224c1d | 134 | static usec_t arg_reboot_watchdog; |
acafd7d8 | 135 | static usec_t arg_kexec_watchdog; |
5717062e | 136 | static usec_t arg_pretimeout_watchdog; |
fb39af4c | 137 | static char *arg_early_core_pattern; |
aff3a9e1 | 138 | static char *arg_watchdog_pretimeout_governor; |
fb39af4c ZJS |
139 | static char *arg_watchdog_device; |
140 | static char **arg_default_environment; | |
d55ed7de | 141 | static char **arg_manager_environment; |
fb39af4c ZJS |
142 | static uint64_t arg_capability_bounding_set; |
143 | static bool arg_no_new_privs; | |
ffc1ec73 | 144 | static int arg_protect_system; |
fb39af4c | 145 | static nsec_t arg_timer_slack_nsec; |
fb39af4c ZJS |
146 | static Set* arg_syscall_archs; |
147 | static FILE* arg_serialization; | |
fb39af4c ZJS |
148 | static sd_id128_t arg_machine_id; |
149 | static EmergencyAction arg_cad_burst_action; | |
fb39af4c | 150 | static CPUSet arg_cpu_affinity; |
b070c7c0 | 151 | static NUMAPolicy arg_numa_policy; |
3753325b | 152 | static usec_t arg_clock_usec; |
d247f232 LP |
153 | static void *arg_random_seed; |
154 | static size_t arg_random_seed_size; | |
856bfaeb LB |
155 | static usec_t arg_reload_limit_interval_sec; |
156 | static unsigned arg_reload_limit_burst; | |
61fbbac1 | 157 | |
0e06a031 LP |
158 | /* A copy of the original environment block */ |
159 | static char **saved_env = NULL; | |
160 | ||
a9fd4cd1 FB |
161 | static int parse_configuration(const struct rlimit *saved_rlimit_nofile, |
162 | const struct rlimit *saved_rlimit_memlock); | |
4fc935ca | 163 | |
f70e6fb4 ZJS |
164 | static int manager_find_user_config_paths(char ***ret_files, char ***ret_dirs) { |
165 | _cleanup_free_ char *base = NULL; | |
166 | _cleanup_strv_free_ char **files = NULL, **dirs = NULL; | |
167 | int r; | |
168 | ||
169 | r = xdg_user_config_dir(&base, "/systemd"); | |
170 | if (r < 0) | |
171 | return r; | |
172 | ||
173 | r = strv_extendf(&files, "%s/user.conf", base); | |
174 | if (r < 0) | |
175 | return r; | |
176 | ||
177 | r = strv_extend(&files, PKGSYSCONFDIR "/user.conf"); | |
178 | if (r < 0) | |
179 | return r; | |
180 | ||
181 | r = strv_consume(&dirs, TAKE_PTR(base)); | |
182 | if (r < 0) | |
183 | return r; | |
184 | ||
185 | r = strv_extend_strv(&dirs, CONF_PATHS_STRV("systemd"), false); | |
186 | if (r < 0) | |
187 | return r; | |
188 | ||
189 | *ret_files = TAKE_PTR(files); | |
190 | *ret_dirs = TAKE_PTR(dirs); | |
191 | return 0; | |
192 | } | |
193 | ||
56d96fc0 | 194 | static int console_setup(void) { |
254d1313 | 195 | _cleanup_close_ int tty_fd = -EBADF; |
29f5a5ae | 196 | unsigned rows, cols; |
56d96fc0 | 197 | int r; |
80876c20 | 198 | |
512947d4 | 199 | tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); |
23bbb0de MS |
200 | if (tty_fd < 0) |
201 | return log_error_errno(tty_fd, "Failed to open /dev/console: %m"); | |
80876c20 | 202 | |
56d96fc0 LP |
203 | /* We don't want to force text mode. plymouth may be showing |
204 | * pictures already from initrd. */ | |
512947d4 | 205 | r = reset_terminal_fd(tty_fd, false); |
23bbb0de MS |
206 | if (r < 0) |
207 | return log_error_errno(r, "Failed to reset /dev/console: %m"); | |
843d2643 | 208 | |
29f5a5ae DDM |
209 | r = proc_cmdline_tty_size("/dev/console", &rows, &cols); |
210 | if (r < 0) | |
211 | log_warning_errno(r, "Failed to get terminal size, ignoring: %m"); | |
212 | else { | |
213 | r = terminal_set_size_fd(tty_fd, NULL, rows, cols); | |
214 | if (r < 0) | |
215 | log_warning_errno(r, "Failed to set terminal size, ignoring: %m"); | |
216 | } | |
217 | ||
56d96fc0 | 218 | return 0; |
80876c20 LP |
219 | } |
220 | ||
96287a49 | 221 | static int parse_proc_cmdline_item(const char *key, const char *value, void *data) { |
059cb385 | 222 | int r; |
f170852a | 223 | |
059cb385 | 224 | assert(key); |
5192bd19 | 225 | |
1d84ad94 | 226 | if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) { |
bf4df7c3 | 227 | |
1d84ad94 LP |
228 | if (proc_cmdline_value_missing(key, value)) |
229 | return 0; | |
bf4df7c3 | 230 | |
1d84ad94 LP |
231 | if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE)) |
232 | log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value); | |
cd57038a ZJS |
233 | else if (in_initrd() == !!startswith(key, "rd.")) |
234 | return free_and_strdup_warn(&arg_default_unit, value); | |
f170852a | 235 | |
1d84ad94 | 236 | } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) { |
4fc935ca | 237 | |
1d84ad94 | 238 | r = value ? parse_boolean(value) : true; |
fb472900 | 239 | if (r < 0) |
5e1ee764 | 240 | log_warning_errno(r, "Failed to parse dump core switch %s, ignoring: %m", value); |
4fc935ca | 241 | else |
fa0f4d8a | 242 | arg_dump_core = r; |
4fc935ca | 243 | |
c6885f5f FB |
244 | } else if (proc_cmdline_key_streq(key, "systemd.early_core_pattern")) { |
245 | ||
246 | if (proc_cmdline_value_missing(key, value)) | |
247 | return 0; | |
248 | ||
249 | if (path_is_absolute(value)) | |
614b022c | 250 | (void) parse_path_argument(value, false, &arg_early_core_pattern); |
c6885f5f FB |
251 | else |
252 | log_warning("Specified core pattern '%s' is not an absolute path, ignoring.", value); | |
253 | ||
1d84ad94 | 254 | } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) { |
b9e74c39 | 255 | |
1d84ad94 LP |
256 | if (!value) |
257 | arg_crash_chvt = 0; /* turn on */ | |
5e1ee764 | 258 | else { |
a07a7324 | 259 | r = parse_crash_chvt(value, &arg_crash_chvt); |
5e1ee764 YW |
260 | if (r < 0) |
261 | log_warning_errno(r, "Failed to parse crash chvt switch %s, ignoring: %m", value); | |
262 | } | |
b9e74c39 | 263 | |
1d84ad94 | 264 | } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) { |
4fc935ca | 265 | |
1d84ad94 | 266 | r = value ? parse_boolean(value) : true; |
fb472900 | 267 | if (r < 0) |
5e1ee764 | 268 | log_warning_errno(r, "Failed to parse crash shell switch %s, ignoring: %m", value); |
4fc935ca | 269 | else |
fa0f4d8a | 270 | arg_crash_shell = r; |
5e7ee61c | 271 | |
1d84ad94 | 272 | } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) { |
5e7ee61c | 273 | |
1d84ad94 | 274 | r = value ? parse_boolean(value) : true; |
b9e74c39 | 275 | if (r < 0) |
5e1ee764 | 276 | log_warning_errno(r, "Failed to parse crash reboot switch %s, ignoring: %m", value); |
5e7ee61c | 277 | else |
b9e74c39 | 278 | arg_crash_reboot = r; |
5e7ee61c | 279 | |
1d84ad94 LP |
280 | } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) { |
281 | char *s; | |
7d5ceb64 | 282 | |
1d84ad94 | 283 | r = parse_confirm_spawn(value, &s); |
059cb385 | 284 | if (r < 0) |
5e1ee764 YW |
285 | log_warning_errno(r, "Failed to parse confirm_spawn switch %s, ignoring: %m", value); |
286 | else | |
287 | free_and_replace(arg_confirm_spawn, s); | |
601f6a1e | 288 | |
2a12e32e JK |
289 | } else if (proc_cmdline_key_streq(key, "systemd.service_watchdogs")) { |
290 | ||
291 | r = value ? parse_boolean(value) : true; | |
292 | if (r < 0) | |
5e1ee764 | 293 | log_warning_errno(r, "Failed to parse service watchdog switch %s, ignoring: %m", value); |
2a12e32e JK |
294 | else |
295 | arg_service_watchdogs = r; | |
296 | ||
1d84ad94 | 297 | } else if (proc_cmdline_key_streq(key, "systemd.show_status")) { |
9e58ff9c | 298 | |
1d84ad94 LP |
299 | if (value) { |
300 | r = parse_show_status(value, &arg_show_status); | |
301 | if (r < 0) | |
5e1ee764 | 302 | log_warning_errno(r, "Failed to parse show status switch %s, ignoring: %m", value); |
1d84ad94 LP |
303 | } else |
304 | arg_show_status = SHOW_STATUS_YES; | |
059cb385 | 305 | |
36cf4507 ZJS |
306 | } else if (proc_cmdline_key_streq(key, "systemd.status_unit_format")) { |
307 | ||
308 | if (proc_cmdline_value_missing(key, value)) | |
309 | return 0; | |
310 | ||
311 | r = status_unit_format_from_string(value); | |
312 | if (r < 0) | |
313 | log_warning_errno(r, "Failed to parse %s=%s, ignoring: %m", key, value); | |
314 | else | |
315 | arg_status_unit_format = r; | |
316 | ||
1d84ad94 LP |
317 | } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) { |
318 | ||
319 | if (proc_cmdline_value_missing(key, value)) | |
320 | return 0; | |
0a494f1f | 321 | |
059cb385 | 322 | r = exec_output_from_string(value); |
fb472900 | 323 | if (r < 0) |
5e1ee764 | 324 | log_warning_errno(r, "Failed to parse default standard output switch %s, ignoring: %m", value); |
0a494f1f | 325 | else |
c9e120e0 | 326 | arg_defaults.std_output = r; |
0a494f1f | 327 | |
1d84ad94 LP |
328 | } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) { |
329 | ||
330 | if (proc_cmdline_value_missing(key, value)) | |
331 | return 0; | |
059cb385 LP |
332 | |
333 | r = exec_output_from_string(value); | |
fb472900 | 334 | if (r < 0) |
5e1ee764 | 335 | log_warning_errno(r, "Failed to parse default standard error switch %s, ignoring: %m", value); |
0a494f1f | 336 | else |
c9e120e0 | 337 | arg_defaults.std_error = r; |
9e7c5357 | 338 | |
1d84ad94 LP |
339 | } else if (streq(key, "systemd.setenv")) { |
340 | ||
341 | if (proc_cmdline_value_missing(key, value)) | |
342 | return 0; | |
059cb385 | 343 | |
b70935ac ZJS |
344 | if (!env_assignment_is_valid(value)) |
345 | log_warning("Environment variable assignment '%s' is not valid. Ignoring.", value); | |
346 | else { | |
347 | r = strv_env_replace_strdup(&arg_default_environment, value); | |
348 | if (r < 0) | |
1d84ad94 | 349 | return log_oom(); |
b70935ac | 350 | } |
9e58ff9c | 351 | |
1d84ad94 LP |
352 | } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) { |
353 | ||
354 | if (proc_cmdline_value_missing(key, value)) | |
355 | return 0; | |
356 | ||
aea3f594 | 357 | r = id128_from_string_nonzero(value, &arg_machine_id); |
1d84ad94 | 358 | if (r < 0) |
5e1ee764 | 359 | log_warning_errno(r, "MachineID '%s' is not valid, ignoring: %m", value); |
ee48dbd5 | 360 | |
1d84ad94 LP |
361 | } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) { |
362 | ||
363 | if (proc_cmdline_value_missing(key, value)) | |
364 | return 0; | |
365 | ||
c9e120e0 | 366 | r = parse_sec(value, &arg_defaults.timeout_start_usec); |
1d84ad94 | 367 | if (r < 0) |
5e1ee764 | 368 | log_warning_errno(r, "Failed to parse default start timeout '%s', ignoring: %m", value); |
1d84ad94 | 369 | |
c9e120e0 LP |
370 | if (arg_defaults.timeout_start_usec <= 0) |
371 | arg_defaults.timeout_start_usec = USEC_INFINITY; | |
ee48dbd5 | 372 | |
6b818cd7 DDM |
373 | } else if (proc_cmdline_key_streq(key, "systemd.default_device_timeout_sec")) { |
374 | ||
375 | if (proc_cmdline_value_missing(key, value)) | |
376 | return 0; | |
377 | ||
c9e120e0 | 378 | r = parse_sec(value, &arg_defaults.device_timeout_usec); |
6b818cd7 DDM |
379 | if (r < 0) |
380 | log_warning_errno(r, "Failed to parse default device timeout '%s', ignoring: %m", value); | |
381 | ||
c9e120e0 LP |
382 | if (arg_defaults.device_timeout_usec <= 0) |
383 | arg_defaults.device_timeout_usec = USEC_INFINITY; | |
6b818cd7 | 384 | |
68d58f38 LP |
385 | } else if (proc_cmdline_key_streq(key, "systemd.cpu_affinity")) { |
386 | ||
387 | if (proc_cmdline_value_missing(key, value)) | |
388 | return 0; | |
389 | ||
390 | r = parse_cpu_set(value, &arg_cpu_affinity); | |
391 | if (r < 0) | |
162392b7 | 392 | log_warning_errno(r, "Failed to parse CPU affinity mask '%s', ignoring: %m", value); |
68d58f38 | 393 | |
8a2c1fbf EJ |
394 | } else if (proc_cmdline_key_streq(key, "systemd.watchdog_device")) { |
395 | ||
396 | if (proc_cmdline_value_missing(key, value)) | |
397 | return 0; | |
398 | ||
614b022c | 399 | (void) parse_path_argument(value, false, &arg_watchdog_device); |
8a2c1fbf | 400 | |
b3aa73e4 FB |
401 | } else if (proc_cmdline_key_streq(key, "systemd.watchdog_sec")) { |
402 | ||
403 | if (proc_cmdline_value_missing(key, value)) | |
404 | return 0; | |
405 | ||
8a85c5b6 FB |
406 | if (streq(value, "default")) |
407 | arg_runtime_watchdog = USEC_INFINITY; | |
c91c95e6 LP |
408 | else if (streq(value, "off")) |
409 | arg_runtime_watchdog = 0; | |
8a85c5b6 FB |
410 | else { |
411 | r = parse_sec(value, &arg_runtime_watchdog); | |
412 | if (r < 0) { | |
413 | log_warning_errno(r, "Failed to parse systemd.watchdog_sec= argument '%s', ignoring: %m", value); | |
414 | return 0; | |
415 | } | |
416 | } | |
417 | ||
418 | arg_kexec_watchdog = arg_reboot_watchdog = arg_runtime_watchdog; | |
b3aa73e4 | 419 | |
5717062e CK |
420 | } else if (proc_cmdline_key_streq(key, "systemd.watchdog_pre_sec")) { |
421 | ||
422 | if (proc_cmdline_value_missing(key, value)) | |
423 | return 0; | |
424 | ||
425 | if (streq(value, "default")) | |
426 | arg_pretimeout_watchdog = USEC_INFINITY; | |
427 | else if (streq(value, "off")) | |
428 | arg_pretimeout_watchdog = 0; | |
429 | else { | |
430 | r = parse_sec(value, &arg_pretimeout_watchdog); | |
431 | if (r < 0) { | |
432 | log_warning_errno(r, "Failed to parse systemd.watchdog_pre_sec= argument '%s', ignoring: %m", value); | |
433 | return 0; | |
434 | } | |
435 | } | |
436 | ||
aff3a9e1 LB |
437 | } else if (proc_cmdline_key_streq(key, "systemd.watchdog_pretimeout_governor")) { |
438 | ||
439 | if (proc_cmdline_value_missing(key, value) || isempty(value)) { | |
440 | arg_watchdog_pretimeout_governor = mfree(arg_watchdog_pretimeout_governor); | |
441 | return 0; | |
442 | } | |
443 | ||
444 | if (!string_is_safe(value)) { | |
445 | log_warning("Watchdog pretimeout governor '%s' is not valid, ignoring.", value); | |
446 | return 0; | |
447 | } | |
448 | ||
449 | return free_and_strdup_warn(&arg_watchdog_pretimeout_governor, value); | |
450 | ||
3753325b LP |
451 | } else if (proc_cmdline_key_streq(key, "systemd.clock_usec")) { |
452 | ||
453 | if (proc_cmdline_value_missing(key, value)) | |
454 | return 0; | |
455 | ||
456 | r = safe_atou64(value, &arg_clock_usec); | |
457 | if (r < 0) | |
458 | log_warning_errno(r, "Failed to parse systemd.clock_usec= argument, ignoring: %s", value); | |
459 | ||
d247f232 LP |
460 | } else if (proc_cmdline_key_streq(key, "systemd.random_seed")) { |
461 | void *p; | |
462 | size_t sz; | |
463 | ||
464 | if (proc_cmdline_value_missing(key, value)) | |
465 | return 0; | |
466 | ||
bdd2036e | 467 | r = unbase64mem(value, &p, &sz); |
d247f232 LP |
468 | if (r < 0) |
469 | log_warning_errno(r, "Failed to parse systemd.random_seed= argument, ignoring: %s", value); | |
470 | ||
471 | free(arg_random_seed); | |
472 | arg_random_seed = sz > 0 ? p : mfree(p); | |
473 | arg_random_seed_size = sz; | |
474 | ||
856bfaeb LB |
475 | } else if (proc_cmdline_key_streq(key, "systemd.reload_limit_interval_sec")) { |
476 | ||
477 | if (proc_cmdline_value_missing(key, value)) | |
478 | return 0; | |
479 | ||
480 | r = parse_sec(value, &arg_reload_limit_interval_sec); | |
481 | if (r < 0) { | |
482 | log_warning_errno(r, "Failed to parse systemd.reload_limit_interval_sec= argument '%s', ignoring: %m", value); | |
483 | return 0; | |
484 | } | |
485 | ||
486 | } else if (proc_cmdline_key_streq(key, "systemd.reload_limit_burst")) { | |
487 | ||
488 | if (proc_cmdline_value_missing(key, value)) | |
489 | return 0; | |
490 | ||
491 | r = safe_atou(value, &arg_reload_limit_burst); | |
492 | if (r < 0) { | |
493 | log_warning_errno(r, "Failed to parse systemd.reload_limit_burst= argument '%s', ignoring: %m", value); | |
494 | return 0; | |
495 | } | |
496 | ||
059cb385 | 497 | } else if (streq(key, "quiet") && !value) { |
d7b15e0a | 498 | |
7a293242 | 499 | if (arg_show_status == _SHOW_STATUS_INVALID) |
0d066dd1 | 500 | arg_show_status = SHOW_STATUS_ERROR; |
059cb385 LP |
501 | |
502 | } else if (streq(key, "debug") && !value) { | |
d7b15e0a | 503 | |
1de1c9c3 LP |
504 | /* Note that log_parse_environment() handles 'debug' |
505 | * too, and sets the log level to LOG_DEBUG. */ | |
d7b15e0a | 506 | |
75f86906 | 507 | if (detect_container() > 0) |
b2103dcc | 508 | log_set_target(LOG_TARGET_CONSOLE); |
059cb385 | 509 | |
dcd61450 | 510 | } else if (!value) { |
e2c9a131 | 511 | const char *target; |
f170852a | 512 | |
ceae6295 | 513 | /* Compatible with SysV, but supported independently even if SysV compatibility is disabled. */ |
e2c9a131 EV |
514 | target = runlevel_to_target(key); |
515 | if (target) | |
cd57038a | 516 | return free_and_strdup_warn(&arg_default_unit, target); |
f170852a LP |
517 | } |
518 | ||
519 | return 0; | |
520 | } | |
521 | ||
e8e581bf ZJS |
522 | #define DEFINE_SETTER(name, func, descr) \ |
523 | static int name(const char *unit, \ | |
524 | const char *filename, \ | |
525 | unsigned line, \ | |
526 | const char *section, \ | |
71a61510 | 527 | unsigned section_line, \ |
e8e581bf ZJS |
528 | const char *lvalue, \ |
529 | int ltype, \ | |
530 | const char *rvalue, \ | |
531 | void *data, \ | |
532 | void *userdata) { \ | |
533 | \ | |
534 | int r; \ | |
535 | \ | |
536 | assert(filename); \ | |
537 | assert(lvalue); \ | |
538 | assert(rvalue); \ | |
539 | \ | |
540 | r = func(rvalue); \ | |
541 | if (r < 0) \ | |
d1cefe0a LP |
542 | log_syntax(unit, LOG_ERR, filename, line, r, \ |
543 | "Invalid " descr "'%s': %m", \ | |
544 | rvalue); \ | |
e8e581bf ZJS |
545 | \ |
546 | return 0; \ | |
547 | } | |
487393e9 | 548 | |
a6ecbf83 FB |
549 | DEFINE_SETTER(config_parse_level2, log_set_max_level_from_string, "log level"); |
550 | DEFINE_SETTER(config_parse_target, log_set_target_from_string, "target"); | |
c5673ed0 | 551 | DEFINE_SETTER(config_parse_color, log_show_color_from_string, "color"); |
a6ecbf83 | 552 | DEFINE_SETTER(config_parse_location, log_show_location_from_string, "location"); |
c5673ed0 | 553 | DEFINE_SETTER(config_parse_time, log_show_time_from_string, "time"); |
487393e9 | 554 | |
a61d6874 ZJS |
555 | static int config_parse_default_timeout_abort( |
556 | const char *unit, | |
557 | const char *filename, | |
558 | unsigned line, | |
559 | const char *section, | |
560 | unsigned section_line, | |
561 | const char *lvalue, | |
562 | int ltype, | |
563 | const char *rvalue, | |
564 | void *data, | |
565 | void *userdata) { | |
566 | int r; | |
567 | ||
c9e120e0 LP |
568 | r = config_parse_timeout_abort( |
569 | unit, | |
570 | filename, | |
571 | line, | |
572 | section, | |
573 | section_line, | |
574 | lvalue, | |
575 | ltype, | |
576 | rvalue, | |
577 | &arg_defaults.timeout_abort_usec, | |
578 | userdata); | |
a61d6874 | 579 | if (r >= 0) |
c9e120e0 | 580 | arg_defaults.timeout_abort_set = r; |
a61d6874 ZJS |
581 | return 0; |
582 | } | |
487393e9 | 583 | |
d4a402e4 LP |
584 | static int config_parse_oom_score_adjust( |
585 | const char *unit, | |
586 | const char *filename, | |
587 | unsigned line, | |
588 | const char *section, | |
589 | unsigned section_line, | |
590 | const char *lvalue, | |
591 | int ltype, | |
592 | const char *rvalue, | |
593 | void *data, | |
594 | void *userdata) { | |
595 | ||
596 | int oa, r; | |
597 | ||
598 | if (isempty(rvalue)) { | |
c9e120e0 | 599 | arg_defaults.oom_score_adjust_set = false; |
d4a402e4 LP |
600 | return 0; |
601 | } | |
602 | ||
603 | r = parse_oom_score_adjust(rvalue, &oa); | |
604 | if (r < 0) { | |
605 | log_syntax(unit, LOG_WARNING, filename, line, r, "Failed to parse the OOM score adjust value '%s', ignoring: %m", rvalue); | |
606 | return 0; | |
607 | } | |
608 | ||
c9e120e0 LP |
609 | arg_defaults.oom_score_adjust = oa; |
610 | arg_defaults.oom_score_adjust_set = true; | |
d4a402e4 LP |
611 | |
612 | return 0; | |
613 | } | |
614 | ||
ffc1ec73 LP |
615 | static int config_parse_protect_system_pid1( |
616 | const char *unit, | |
617 | const char *filename, | |
618 | unsigned line, | |
619 | const char *section, | |
620 | unsigned section_line, | |
621 | const char *lvalue, | |
622 | int ltype, | |
623 | const char *rvalue, | |
624 | void *data, | |
625 | void *userdata) { | |
626 | ||
627 | int *v = ASSERT_PTR(data), r; | |
628 | ||
629 | /* This is modelled after the per-service ProtectSystem= setting, but a bit more restricted on one | |
630 | * hand, and more automatic in another. i.e. we currently only support yes/no (not "strict" or | |
631 | * "full"). And we will enable this automatically for the initrd unless configured otherwise. | |
632 | * | |
633 | * We might extend this later to match more closely what the per-service ProtectSystem= can do, but | |
634 | * this is not trivial, due to ordering constraints: besides /usr/ we don't really have much mounted | |
635 | * at the moment we enable this logic. */ | |
636 | ||
637 | if (isempty(rvalue) || streq(rvalue, "auto")) { | |
638 | *v = -1; | |
639 | return 0; | |
640 | } | |
641 | ||
642 | r = parse_boolean(rvalue); | |
643 | if (r < 0) { | |
644 | log_syntax(unit, LOG_WARNING, filename, line, r, "Failed to parse ProtectSystem= argument '%s', ignoring: %m", rvalue); | |
645 | return 0; | |
646 | } | |
647 | ||
648 | *v = r; | |
649 | return 0; | |
650 | } | |
651 | ||
a61d6874 | 652 | static int parse_config_file(void) { |
f975e971 | 653 | const ConfigTableItem items[] = { |
3f87eaa5 YW |
654 | { "Manager", "LogLevel", config_parse_level2, 0, NULL }, |
655 | { "Manager", "LogTarget", config_parse_target, 0, NULL }, | |
656 | { "Manager", "LogColor", config_parse_color, 0, NULL }, | |
657 | { "Manager", "LogLocation", config_parse_location, 0, NULL }, | |
658 | { "Manager", "LogTime", config_parse_time, 0, NULL }, | |
659 | { "Manager", "DumpCore", config_parse_bool, 0, &arg_dump_core }, | |
660 | { "Manager", "CrashChVT", /* legacy */ config_parse_crash_chvt, 0, &arg_crash_chvt }, | |
661 | { "Manager", "CrashChangeVT", config_parse_crash_chvt, 0, &arg_crash_chvt }, | |
662 | { "Manager", "CrashShell", config_parse_bool, 0, &arg_crash_shell }, | |
663 | { "Manager", "CrashReboot", config_parse_bool, 0, &arg_crash_reboot }, | |
664 | { "Manager", "ShowStatus", config_parse_show_status, 0, &arg_show_status }, | |
665 | { "Manager", "StatusUnitFormat", config_parse_status_unit_format, 0, &arg_status_unit_format }, | |
666 | { "Manager", "CPUAffinity", config_parse_cpu_affinity2, 0, &arg_cpu_affinity }, | |
667 | { "Manager", "NUMAPolicy", config_parse_numa_policy, 0, &arg_numa_policy.type }, | |
668 | { "Manager", "NUMAMask", config_parse_numa_mask, 0, &arg_numa_policy }, | |
669 | { "Manager", "JoinControllers", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, | |
670 | { "Manager", "RuntimeWatchdogSec", config_parse_watchdog_sec, 0, &arg_runtime_watchdog }, | |
671 | { "Manager", "RuntimeWatchdogPreSec", config_parse_watchdog_sec, 0, &arg_pretimeout_watchdog }, | |
672 | { "Manager", "RebootWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, | |
673 | { "Manager", "ShutdownWatchdogSec", config_parse_watchdog_sec, 0, &arg_reboot_watchdog }, /* obsolete alias */ | |
674 | { "Manager", "KExecWatchdogSec", config_parse_watchdog_sec, 0, &arg_kexec_watchdog }, | |
675 | { "Manager", "WatchdogDevice", config_parse_path, 0, &arg_watchdog_device }, | |
676 | { "Manager", "RuntimeWatchdogPreGovernor", config_parse_string, CONFIG_PARSE_STRING_SAFE, &arg_watchdog_pretimeout_governor }, | |
677 | { "Manager", "CapabilityBoundingSet", config_parse_capability_set, 0, &arg_capability_bounding_set }, | |
678 | { "Manager", "NoNewPrivileges", config_parse_bool, 0, &arg_no_new_privs }, | |
ffc1ec73 | 679 | { "Manager", "ProtectSystem", config_parse_protect_system_pid1, 0, &arg_protect_system }, |
349cc4a5 | 680 | #if HAVE_SECCOMP |
3f87eaa5 | 681 | { "Manager", "SystemCallArchitectures", config_parse_syscall_archs, 0, &arg_syscall_archs }, |
6aa2c555 LP |
682 | #else |
683 | { "Manager", "SystemCallArchitectures", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, | |
684 | ||
89fffa27 | 685 | #endif |
3f87eaa5 | 686 | { "Manager", "TimerSlackNSec", config_parse_nsec, 0, &arg_timer_slack_nsec }, |
c9e120e0 LP |
687 | { "Manager", "DefaultTimerAccuracySec", config_parse_sec, 0, &arg_defaults.timer_accuracy_usec }, |
688 | { "Manager", "DefaultStandardOutput", config_parse_output_restricted, 0, &arg_defaults.std_output }, | |
689 | { "Manager", "DefaultStandardError", config_parse_output_restricted, 0, &arg_defaults.std_error }, | |
690 | { "Manager", "DefaultTimeoutStartSec", config_parse_sec, 0, &arg_defaults.timeout_start_usec }, | |
691 | { "Manager", "DefaultTimeoutStopSec", config_parse_sec, 0, &arg_defaults.timeout_stop_usec }, | |
3f87eaa5 | 692 | { "Manager", "DefaultTimeoutAbortSec", config_parse_default_timeout_abort, 0, NULL }, |
c9e120e0 LP |
693 | { "Manager", "DefaultDeviceTimeoutSec", config_parse_sec, 0, &arg_defaults.device_timeout_usec }, |
694 | { "Manager", "DefaultRestartSec", config_parse_sec, 0, &arg_defaults.restart_usec }, | |
695 | { "Manager", "DefaultStartLimitInterval", config_parse_sec, 0, &arg_defaults.start_limit_interval}, /* obsolete alias */ | |
696 | { "Manager", "DefaultStartLimitIntervalSec", config_parse_sec, 0, &arg_defaults.start_limit_interval}, | |
697 | { "Manager", "DefaultStartLimitBurst", config_parse_unsigned, 0, &arg_defaults.start_limit_burst }, | |
4870133b LP |
698 | { "Manager", "DefaultEnvironment", config_parse_environ, arg_runtime_scope, &arg_default_environment }, |
699 | { "Manager", "ManagerEnvironment", config_parse_environ, arg_runtime_scope, &arg_manager_environment }, | |
c9e120e0 LP |
700 | { "Manager", "DefaultLimitCPU", config_parse_rlimit, RLIMIT_CPU, arg_defaults.rlimit }, |
701 | { "Manager", "DefaultLimitFSIZE", config_parse_rlimit, RLIMIT_FSIZE, arg_defaults.rlimit }, | |
702 | { "Manager", "DefaultLimitDATA", config_parse_rlimit, RLIMIT_DATA, arg_defaults.rlimit }, | |
703 | { "Manager", "DefaultLimitSTACK", config_parse_rlimit, RLIMIT_STACK, arg_defaults.rlimit }, | |
704 | { "Manager", "DefaultLimitCORE", config_parse_rlimit, RLIMIT_CORE, arg_defaults.rlimit }, | |
705 | { "Manager", "DefaultLimitRSS", config_parse_rlimit, RLIMIT_RSS, arg_defaults.rlimit }, | |
706 | { "Manager", "DefaultLimitNOFILE", config_parse_rlimit, RLIMIT_NOFILE, arg_defaults.rlimit }, | |
707 | { "Manager", "DefaultLimitAS", config_parse_rlimit, RLIMIT_AS, arg_defaults.rlimit }, | |
708 | { "Manager", "DefaultLimitNPROC", config_parse_rlimit, RLIMIT_NPROC, arg_defaults.rlimit }, | |
709 | { "Manager", "DefaultLimitMEMLOCK", config_parse_rlimit, RLIMIT_MEMLOCK, arg_defaults.rlimit }, | |
710 | { "Manager", "DefaultLimitLOCKS", config_parse_rlimit, RLIMIT_LOCKS, arg_defaults.rlimit }, | |
711 | { "Manager", "DefaultLimitSIGPENDING", config_parse_rlimit, RLIMIT_SIGPENDING, arg_defaults.rlimit }, | |
712 | { "Manager", "DefaultLimitMSGQUEUE", config_parse_rlimit, RLIMIT_MSGQUEUE, arg_defaults.rlimit }, | |
713 | { "Manager", "DefaultLimitNICE", config_parse_rlimit, RLIMIT_NICE, arg_defaults.rlimit }, | |
714 | { "Manager", "DefaultLimitRTPRIO", config_parse_rlimit, RLIMIT_RTPRIO, arg_defaults.rlimit }, | |
715 | { "Manager", "DefaultLimitRTTIME", config_parse_rlimit, RLIMIT_RTTIME, arg_defaults.rlimit }, | |
716 | { "Manager", "DefaultCPUAccounting", config_parse_bool, 0, &arg_defaults.cpu_accounting }, | |
717 | { "Manager", "DefaultIOAccounting", config_parse_bool, 0, &arg_defaults.io_accounting }, | |
718 | { "Manager", "DefaultIPAccounting", config_parse_bool, 0, &arg_defaults.ip_accounting }, | |
719 | { "Manager", "DefaultBlockIOAccounting", config_parse_bool, 0, &arg_defaults.blockio_accounting }, | |
720 | { "Manager", "DefaultMemoryAccounting", config_parse_bool, 0, &arg_defaults.memory_accounting }, | |
721 | { "Manager", "DefaultTasksAccounting", config_parse_bool, 0, &arg_defaults.tasks_accounting }, | |
722 | { "Manager", "DefaultTasksMax", config_parse_tasks_max, 0, &arg_defaults.tasks_max }, | |
723 | { "Manager", "DefaultMemoryPressureThresholdSec", config_parse_sec, 0, &arg_defaults.memory_pressure_threshold_usec }, | |
724 | { "Manager", "DefaultMemoryPressureWatch", config_parse_memory_pressure_watch, 0, &arg_defaults.memory_pressure_watch }, | |
4870133b | 725 | { "Manager", "CtrlAltDelBurstAction", config_parse_emergency_action, arg_runtime_scope, &arg_cad_burst_action }, |
c9e120e0 | 726 | { "Manager", "DefaultOOMPolicy", config_parse_oom_policy, 0, &arg_defaults.oom_policy }, |
3f87eaa5 | 727 | { "Manager", "DefaultOOMScoreAdjust", config_parse_oom_score_adjust, 0, NULL }, |
856bfaeb LB |
728 | { "Manager", "ReloadLimitIntervalSec", config_parse_sec, 0, &arg_reload_limit_interval_sec }, |
729 | { "Manager", "ReloadLimitBurst", config_parse_unsigned, 0, &arg_reload_limit_burst }, | |
aa5ae971 | 730 | #if ENABLE_SMACK |
c9e120e0 | 731 | { "Manager", "DefaultSmackProcessLabel", config_parse_string, 0, &arg_defaults.smack_process_label }, |
aa5ae971 ŁS |
732 | #else |
733 | { "Manager", "DefaultSmackProcessLabel", config_parse_warn_compat, DISABLED_CONFIGURATION, NULL }, | |
734 | #endif | |
d3b1c508 | 735 | {} |
487393e9 LP |
736 | }; |
737 | ||
4870133b | 738 | if (arg_runtime_scope == RUNTIME_SCOPE_SYSTEM) |
07e0ffc8 FB |
739 | (void) config_parse_config_file("system.conf", |
740 | "Manager\0", | |
741 | config_item_table_lookup, items, | |
742 | CONFIG_PARSE_WARN, | |
743 | NULL); | |
e94a009c | 744 | else { |
07e0ffc8 FB |
745 | _cleanup_strv_free_ char **files = NULL, **dirs = NULL; |
746 | int r; | |
747 | ||
4870133b LP |
748 | assert(arg_runtime_scope == RUNTIME_SCOPE_USER); |
749 | ||
e94a009c | 750 | r = manager_find_user_config_paths(&files, &dirs); |
f70e6fb4 ZJS |
751 | if (r < 0) |
752 | return log_error_errno(r, "Failed to determine config file paths: %m"); | |
e94a009c | 753 | |
07e0ffc8 FB |
754 | (void) config_parse_many( |
755 | (const char* const*) files, | |
756 | (const char* const*) dirs, | |
757 | "user.conf.d", | |
947f59ba | 758 | /* root = */ NULL, |
07e0ffc8 FB |
759 | "Manager\0", |
760 | config_item_table_lookup, items, | |
761 | CONFIG_PARSE_WARN, | |
762 | NULL, NULL, NULL); | |
f70e6fb4 | 763 | } |
75eb6154 | 764 | |
f70e6fb4 ZJS |
765 | /* Traditionally "0" was used to turn off the default unit timeouts. Fix this up so that we use |
766 | * USEC_INFINITY like everywhere else. */ | |
c9e120e0 LP |
767 | if (arg_defaults.timeout_start_usec <= 0) |
768 | arg_defaults.timeout_start_usec = USEC_INFINITY; | |
769 | if (arg_defaults.timeout_stop_usec <= 0) | |
770 | arg_defaults.timeout_stop_usec = USEC_INFINITY; | |
487393e9 | 771 | |
487393e9 LP |
772 | return 0; |
773 | } | |
774 | ||
85cb4151 | 775 | static void set_manager_defaults(Manager *m) { |
bfb27b06 | 776 | int r; |
06af2a04 TB |
777 | |
778 | assert(m); | |
779 | ||
bfb27b06 LP |
780 | /* Propagates the various default unit property settings into the manager object, i.e. properties |
781 | * that do not affect the manager itself, but are just what newly allocated units will have set if | |
782 | * they haven't set anything else. (Also see set_manager_settings() for the settings that affect the | |
783 | * manager's own behaviour) */ | |
784 | ||
785 | r = manager_set_unit_defaults(m, &arg_defaults); | |
786 | if (r < 0) | |
787 | log_warning_errno(r, "Failed to set manager defaults, ignoring: %m"); | |
788 | ||
789 | r = manager_default_environment(m); | |
790 | if (r < 0) | |
791 | log_warning_errno(r, "Failed to set manager default environment, ignoring: %m"); | |
792 | ||
793 | r = manager_transient_environment_add(m, arg_default_environment); | |
794 | if (r < 0) | |
795 | log_warning_errno(r, "Failed to add to transient environment, ignoring: %m"); | |
06af2a04 TB |
796 | } |
797 | ||
7b46fc6a | 798 | static void set_manager_settings(Manager *m) { |
aff3a9e1 | 799 | int r; |
7b46fc6a LP |
800 | |
801 | assert(m); | |
802 | ||
986935cf FB |
803 | /* Propagates the various manager settings into the manager object, i.e. properties that |
804 | * effect the manager itself (as opposed to just being inherited into newly allocated | |
805 | * units, see set_manager_defaults() above). */ | |
5b65ae15 | 806 | |
7b46fc6a | 807 | m->confirm_spawn = arg_confirm_spawn; |
2a12e32e | 808 | m->service_watchdogs = arg_service_watchdogs; |
7b46fc6a | 809 | m->cad_burst_action = arg_cad_burst_action; |
856bfaeb LB |
810 | /* Note that we don't do structured initialization here, otherwise it will reset the rate limit |
811 | * counter on every daemon-reload. */ | |
812 | m->reload_ratelimit.interval = arg_reload_limit_interval_sec; | |
813 | m->reload_ratelimit.burst = arg_reload_limit_burst; | |
7b46fc6a | 814 | |
986935cf FB |
815 | manager_set_watchdog(m, WATCHDOG_RUNTIME, arg_runtime_watchdog); |
816 | manager_set_watchdog(m, WATCHDOG_REBOOT, arg_reboot_watchdog); | |
817 | manager_set_watchdog(m, WATCHDOG_KEXEC, arg_kexec_watchdog); | |
5717062e | 818 | manager_set_watchdog(m, WATCHDOG_PRETIMEOUT, arg_pretimeout_watchdog); |
aff3a9e1 LB |
819 | r = manager_set_watchdog_pretimeout_governor(m, arg_watchdog_pretimeout_governor); |
820 | if (r < 0) | |
821 | log_warning_errno(r, "Failed to set watchdog pretimeout governor to '%s', ignoring: %m", arg_watchdog_pretimeout_governor); | |
986935cf | 822 | |
7c52d523 | 823 | manager_set_show_status(m, arg_show_status, "command line"); |
36cf4507 | 824 | m->status_unit_format = arg_status_unit_format; |
7b46fc6a LP |
825 | } |
826 | ||
f170852a | 827 | static int parse_argv(int argc, char *argv[]) { |
f170852a | 828 | enum { |
6339d3e6 YW |
829 | COMMON_GETOPT_ARGS, |
830 | SYSTEMD_GETOPT_ARGS, | |
f170852a LP |
831 | }; |
832 | ||
833 | static const struct option options[] = { | |
6339d3e6 YW |
834 | COMMON_GETOPT_OPTIONS, |
835 | SYSTEMD_GETOPT_OPTIONS, | |
fb472900 | 836 | {} |
f170852a LP |
837 | }; |
838 | ||
839 | int c, r; | |
9a9ca408 | 840 | bool user_arg_seen = false; |
f170852a LP |
841 | |
842 | assert(argc >= 1); | |
843 | assert(argv); | |
844 | ||
df0ff127 | 845 | if (getpid_cached() == 1) |
b770165a LP |
846 | opterr = 0; |
847 | ||
6339d3e6 | 848 | while ((c = getopt_long(argc, argv, SYSTEMD_GETOPT_SHORT_OPTIONS, options, NULL)) >= 0) |
f170852a LP |
849 | |
850 | switch (c) { | |
851 | ||
852 | case ARG_LOG_LEVEL: | |
fb472900 | 853 | r = log_set_max_level_from_string(optarg); |
2b5107e1 ZJS |
854 | if (r < 0) |
855 | return log_error_errno(r, "Failed to parse log level \"%s\": %m", optarg); | |
f170852a LP |
856 | |
857 | break; | |
858 | ||
859 | case ARG_LOG_TARGET: | |
fb472900 | 860 | r = log_set_target_from_string(optarg); |
2b5107e1 ZJS |
861 | if (r < 0) |
862 | return log_error_errno(r, "Failed to parse log target \"%s\": %m", optarg); | |
f170852a LP |
863 | |
864 | break; | |
865 | ||
bbe63281 LP |
866 | case ARG_LOG_COLOR: |
867 | ||
d0b170c8 | 868 | if (optarg) { |
fb472900 | 869 | r = log_show_color_from_string(optarg); |
2b5107e1 ZJS |
870 | if (r < 0) |
871 | return log_error_errno(r, "Failed to parse log color setting \"%s\": %m", | |
872 | optarg); | |
d0b170c8 LP |
873 | } else |
874 | log_show_color(true); | |
bbe63281 LP |
875 | |
876 | break; | |
877 | ||
878 | case ARG_LOG_LOCATION: | |
d0b170c8 | 879 | if (optarg) { |
fb472900 | 880 | r = log_show_location_from_string(optarg); |
2b5107e1 ZJS |
881 | if (r < 0) |
882 | return log_error_errno(r, "Failed to parse log location setting \"%s\": %m", | |
883 | optarg); | |
d0b170c8 LP |
884 | } else |
885 | log_show_location(true); | |
bbe63281 LP |
886 | |
887 | break; | |
888 | ||
c5673ed0 DS |
889 | case ARG_LOG_TIME: |
890 | ||
891 | if (optarg) { | |
892 | r = log_show_time_from_string(optarg); | |
893 | if (r < 0) | |
894 | return log_error_errno(r, "Failed to parse log time setting \"%s\": %m", | |
895 | optarg); | |
896 | } else | |
897 | log_show_time(true); | |
898 | ||
899 | break; | |
900 | ||
0a494f1f | 901 | case ARG_DEFAULT_STD_OUTPUT: |
fb472900 | 902 | r = exec_output_from_string(optarg); |
2b5107e1 ZJS |
903 | if (r < 0) |
904 | return log_error_errno(r, "Failed to parse default standard output setting \"%s\": %m", | |
905 | optarg); | |
c9e120e0 | 906 | arg_defaults.std_output = r; |
0a494f1f LP |
907 | break; |
908 | ||
909 | case ARG_DEFAULT_STD_ERROR: | |
fb472900 | 910 | r = exec_output_from_string(optarg); |
2b5107e1 ZJS |
911 | if (r < 0) |
912 | return log_error_errno(r, "Failed to parse default standard error output setting \"%s\": %m", | |
913 | optarg); | |
c9e120e0 | 914 | arg_defaults.std_error = r; |
0a494f1f LP |
915 | break; |
916 | ||
2f198e2f | 917 | case ARG_UNIT: |
e6e242ad | 918 | r = free_and_strdup(&arg_default_unit, optarg); |
23bbb0de | 919 | if (r < 0) |
2b5107e1 | 920 | return log_error_errno(r, "Failed to set default unit \"%s\": %m", optarg); |
f170852a LP |
921 | |
922 | break; | |
923 | ||
edb9aaa8 | 924 | case ARG_SYSTEM: |
4870133b | 925 | arg_runtime_scope = RUNTIME_SCOPE_SYSTEM; |
edb9aaa8 | 926 | break; |
a5dab5ce | 927 | |
af2d49f7 | 928 | case ARG_USER: |
4870133b | 929 | arg_runtime_scope = RUNTIME_SCOPE_USER; |
9a9ca408 | 930 | user_arg_seen = true; |
a5dab5ce | 931 | break; |
a5dab5ce | 932 | |
e965d56d | 933 | case ARG_TEST: |
fa0f4d8a | 934 | arg_action = ACTION_TEST; |
b87c2aa6 ZJS |
935 | break; |
936 | ||
937 | case ARG_NO_PAGER: | |
0221d68a | 938 | arg_pager_flags |= PAGER_DISABLE; |
e965d56d LP |
939 | break; |
940 | ||
9ba0bc4e ZJS |
941 | case ARG_VERSION: |
942 | arg_action = ACTION_VERSION; | |
943 | break; | |
944 | ||
e537352b | 945 | case ARG_DUMP_CONFIGURATION_ITEMS: |
fa0f4d8a | 946 | arg_action = ACTION_DUMP_CONFIGURATION_ITEMS; |
e537352b LP |
947 | break; |
948 | ||
bbc1acab YW |
949 | case ARG_DUMP_BUS_PROPERTIES: |
950 | arg_action = ACTION_DUMP_BUS_PROPERTIES; | |
951 | break; | |
952 | ||
5c08257b ZJS |
953 | case ARG_BUS_INTROSPECT: |
954 | arg_bus_introspect = optarg; | |
955 | arg_action = ACTION_BUS_INTROSPECT; | |
956 | break; | |
957 | ||
9e58ff9c | 958 | case ARG_DUMP_CORE: |
599c7c54 ZJS |
959 | r = parse_boolean_argument("--dump-core", optarg, &arg_dump_core); |
960 | if (r < 0) | |
961 | return r; | |
b9e74c39 LP |
962 | break; |
963 | ||
964 | case ARG_CRASH_CHVT: | |
a07a7324 | 965 | r = parse_crash_chvt(optarg, &arg_crash_chvt); |
b9e74c39 | 966 | if (r < 0) |
2b5107e1 ZJS |
967 | return log_error_errno(r, "Failed to parse crash virtual terminal index: \"%s\": %m", |
968 | optarg); | |
9e58ff9c LP |
969 | break; |
970 | ||
971 | case ARG_CRASH_SHELL: | |
599c7c54 ZJS |
972 | r = parse_boolean_argument("--crash-shell", optarg, &arg_crash_shell); |
973 | if (r < 0) | |
974 | return r; | |
b9e74c39 LP |
975 | break; |
976 | ||
977 | case ARG_CRASH_REBOOT: | |
599c7c54 ZJS |
978 | r = parse_boolean_argument("--crash-reboot", optarg, &arg_crash_reboot); |
979 | if (r < 0) | |
980 | return r; | |
9e58ff9c LP |
981 | break; |
982 | ||
80876c20 | 983 | case ARG_CONFIRM_SPAWN: |
7d5ceb64 FB |
984 | arg_confirm_spawn = mfree(arg_confirm_spawn); |
985 | ||
986 | r = parse_confirm_spawn(optarg, &arg_confirm_spawn); | |
987 | if (r < 0) | |
2b5107e1 ZJS |
988 | return log_error_errno(r, "Failed to parse confirm spawn option: \"%s\": %m", |
989 | optarg); | |
80876c20 LP |
990 | break; |
991 | ||
2a12e32e | 992 | case ARG_SERVICE_WATCHDOGS: |
599c7c54 | 993 | r = parse_boolean_argument("--service-watchdogs=", optarg, &arg_service_watchdogs); |
2a12e32e | 994 | if (r < 0) |
599c7c54 | 995 | return r; |
2a12e32e JK |
996 | break; |
997 | ||
9e58ff9c | 998 | case ARG_SHOW_STATUS: |
d450b6f2 ZJS |
999 | if (optarg) { |
1000 | r = parse_show_status(optarg, &arg_show_status); | |
ac7ec288 | 1001 | if (r < 0) |
2b5107e1 ZJS |
1002 | return log_error_errno(r, "Failed to parse show status boolean: \"%s\": %m", |
1003 | optarg); | |
d450b6f2 ZJS |
1004 | } else |
1005 | arg_show_status = SHOW_STATUS_YES; | |
6e98720f | 1006 | break; |
a5d87bf0 | 1007 | |
a16e1123 LP |
1008 | case ARG_DESERIALIZE: { |
1009 | int fd; | |
1010 | FILE *f; | |
1011 | ||
e652663a | 1012 | fd = parse_fd(optarg); |
baaa35ad | 1013 | if (fd < 0) |
d2132d3d | 1014 | return log_error_errno(fd, "Failed to parse serialization fd \"%s\": %m", optarg); |
a16e1123 | 1015 | |
b9e74c39 | 1016 | (void) fd_cloexec(fd, true); |
01e10de3 LP |
1017 | |
1018 | f = fdopen(fd, "r"); | |
4a62c710 | 1019 | if (!f) |
2b5107e1 | 1020 | return log_error_errno(errno, "Failed to open serialization fd %d: %m", fd); |
a16e1123 | 1021 | |
74ca738f | 1022 | safe_fclose(arg_serialization); |
d3b1c508 | 1023 | arg_serialization = f; |
a16e1123 LP |
1024 | |
1025 | break; | |
1026 | } | |
1027 | ||
2660882b | 1028 | case ARG_SWITCHED_ROOT: |
bf4df7c3 | 1029 | arg_switched_root = true; |
d03bc1b8 HH |
1030 | break; |
1031 | ||
ee48dbd5 | 1032 | case ARG_MACHINE_ID: |
aea3f594 | 1033 | r = id128_from_string_nonzero(optarg, &arg_machine_id); |
54500613 | 1034 | if (r < 0) |
2b5107e1 | 1035 | return log_error_errno(r, "MachineID '%s' is not valid: %m", optarg); |
ee48dbd5 NC |
1036 | break; |
1037 | ||
f170852a | 1038 | case 'h': |
fa0f4d8a | 1039 | arg_action = ACTION_HELP; |
f170852a LP |
1040 | break; |
1041 | ||
1d2e23ab LP |
1042 | case 'D': |
1043 | log_set_max_level(LOG_DEBUG); | |
1044 | break; | |
1045 | ||
099663ff LP |
1046 | case 'b': |
1047 | case 's': | |
1048 | case 'z': | |
cd57038a ZJS |
1049 | /* Just to eat away the sysvinit kernel cmdline args that we'll parse in |
1050 | * parse_proc_cmdline_item() or ignore, without any getopt() error messages. | |
1051 | */ | |
099663ff | 1052 | case '?': |
df0ff127 | 1053 | if (getpid_cached() != 1) |
099663ff | 1054 | return -EINVAL; |
601185b4 ZJS |
1055 | else |
1056 | return 0; | |
099663ff | 1057 | |
601185b4 | 1058 | default: |
04499a70 | 1059 | assert_not_reached(); |
f170852a LP |
1060 | } |
1061 | ||
d7a0f1f4 | 1062 | if (optind < argc && getpid_cached() != 1) |
9a9ca408 ZJS |
1063 | /* Hmm, when we aren't run as init system let's complain about excess arguments */ |
1064 | return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Excess arguments."); | |
1065 | ||
4870133b | 1066 | if (arg_action == ACTION_RUN && arg_runtime_scope == RUNTIME_SCOPE_USER && !user_arg_seen) |
baaa35ad | 1067 | return log_error_errno(SYNTHETIC_ERRNO(EINVAL), |
9a9ca408 | 1068 | "Explicit --user argument required to run as user manager."); |
d821e6d6 | 1069 | |
f170852a LP |
1070 | return 0; |
1071 | } | |
1072 | ||
1073 | static int help(void) { | |
37ec0fdd LP |
1074 | _cleanup_free_ char *link = NULL; |
1075 | int r; | |
1076 | ||
1077 | r = terminal_urlify_man("systemd", "1", &link); | |
1078 | if (r < 0) | |
1079 | return log_oom(); | |
f170852a | 1080 | |
2e33c433 | 1081 | printf("%s [OPTIONS...]\n\n" |
7ae47326 ZJS |
1082 | "%sStarts and monitors system and user services.%s\n\n" |
1083 | "This program takes no positional arguments.\n\n" | |
1084 | "%sOptions%s:\n" | |
e537352b | 1085 | " -h --help Show this help\n" |
cb4069d9 | 1086 | " --version Show version\n" |
cd69e88b | 1087 | " --test Determine initial transaction, dump it and exit\n" |
03b3b55e ZJS |
1088 | " --system Combined with --test: operate in system mode\n" |
1089 | " --user Combined with --test: operate in user mode\n" | |
80876c20 | 1090 | " --dump-configuration-items Dump understood unit configuration items\n" |
bbc1acab | 1091 | " --dump-bus-properties Dump exposed bus properties\n" |
5c08257b | 1092 | " --bus-introspect=PATH Write XML introspection data\n" |
9e58ff9c | 1093 | " --unit=UNIT Set default unit\n" |
b9e74c39 LP |
1094 | " --dump-core[=BOOL] Dump core on crash\n" |
1095 | " --crash-vt=NR Change to specified VT on crash\n" | |
1096 | " --crash-reboot[=BOOL] Reboot on crash\n" | |
1097 | " --crash-shell[=BOOL] Run shell on crash\n" | |
1098 | " --confirm-spawn[=BOOL] Ask for confirmation when spawning processes\n" | |
03b3b55e ZJS |
1099 | " --show-status[=BOOL] Show status updates on the console during boot\n" |
1100 | " --log-target=TARGET Set log target (console, journal, kmsg,\n" | |
1101 | " journal-or-kmsg, null)\n" | |
1102 | " --log-level=LEVEL Set log level (debug, info, notice, warning,\n" | |
1103 | " err, crit, alert, emerg)\n" | |
b9e74c39 LP |
1104 | " --log-color[=BOOL] Highlight important log messages\n" |
1105 | " --log-location[=BOOL] Include code location in log messages\n" | |
c5673ed0 | 1106 | " --log-time[=BOOL] Prefix log messages with current time\n" |
0a494f1f | 1107 | " --default-standard-output= Set default standard output for services\n" |
37ec0fdd | 1108 | " --default-standard-error= Set default standard error output for services\n" |
03b3b55e | 1109 | " --no-pager Do not pipe output into a pager\n" |
bc556335 DDM |
1110 | "\nSee the %s for details.\n", |
1111 | program_invocation_short_name, | |
1112 | ansi_highlight(), | |
1113 | ansi_normal(), | |
1114 | ansi_underline(), | |
1115 | ansi_normal(), | |
1116 | link); | |
f170852a LP |
1117 | |
1118 | return 0; | |
1119 | } | |
1120 | ||
2cc856ac LP |
1121 | static int prepare_reexecute( |
1122 | Manager *m, | |
1123 | FILE **ret_f, | |
1124 | FDSet **ret_fds, | |
1125 | bool switching_root) { | |
1126 | ||
48b90859 LP |
1127 | _cleanup_fdset_free_ FDSet *fds = NULL; |
1128 | _cleanup_fclose_ FILE *f = NULL; | |
a16e1123 LP |
1129 | int r; |
1130 | ||
1131 | assert(m); | |
2cc856ac LP |
1132 | assert(ret_f); |
1133 | assert(ret_fds); | |
a16e1123 | 1134 | |
6b78f9b4 | 1135 | r = manager_open_serialization(m, &f); |
48b90859 LP |
1136 | if (r < 0) |
1137 | return log_error_errno(r, "Failed to create serialization file: %m"); | |
a16e1123 | 1138 | |
71445ae7 | 1139 | /* Make sure nothing is really destructed when we shut down */ |
313cefa1 | 1140 | m->n_reloading++; |
718db961 | 1141 | bus_manager_send_reloading(m, true); |
71445ae7 | 1142 | |
6b78f9b4 | 1143 | fds = fdset_new(); |
48b90859 LP |
1144 | if (!fds) |
1145 | return log_oom(); | |
a16e1123 | 1146 | |
b3680f49 | 1147 | r = manager_serialize(m, f, fds, switching_root); |
48b90859 | 1148 | if (r < 0) |
d68c645b | 1149 | return r; |
a16e1123 | 1150 | |
86cbbc6d | 1151 | if (fseeko(f, 0, SEEK_SET) < 0) |
48b90859 | 1152 | return log_error_errno(errno, "Failed to rewind serialization fd: %m"); |
a16e1123 | 1153 | |
6b78f9b4 | 1154 | r = fd_cloexec(fileno(f), false); |
48b90859 LP |
1155 | if (r < 0) |
1156 | return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization: %m"); | |
a16e1123 | 1157 | |
6b78f9b4 | 1158 | r = fdset_cloexec(fds, false); |
48b90859 LP |
1159 | if (r < 0) |
1160 | return log_error_errno(r, "Failed to disable O_CLOEXEC for serialization fds: %m"); | |
a16e1123 | 1161 | |
2cc856ac LP |
1162 | *ret_f = TAKE_PTR(f); |
1163 | *ret_fds = TAKE_PTR(fds); | |
a16e1123 | 1164 | |
48b90859 | 1165 | return 0; |
a16e1123 LP |
1166 | } |
1167 | ||
a8b627aa LP |
1168 | static void bump_file_max_and_nr_open(void) { |
1169 | ||
ff3a7019 ZJS |
1170 | /* Let's bump fs.file-max and fs.nr_open to their respective maximums. On current kernels large |
1171 | * numbers of file descriptors are no longer a performance problem and their memory is properly | |
1172 | * tracked by memcg, thus counting them and limiting them in another two layers of limits is | |
1173 | * unnecessary and just complicates things. This function hence turns off 2 of the 4 levels of limits | |
1174 | * on file descriptors, and makes RLIMIT_NOLIMIT (soft + hard) the only ones that really matter. */ | |
a8b627aa LP |
1175 | |
1176 | #if BUMP_PROC_SYS_FS_FILE_MAX || BUMP_PROC_SYS_FS_NR_OPEN | |
a8b627aa LP |
1177 | int r; |
1178 | #endif | |
1179 | ||
1180 | #if BUMP_PROC_SYS_FS_FILE_MAX | |
409607c1 ZJS |
1181 | /* The maximum the kernel allows for this since 5.2 is LONG_MAX, use that. (Previously things were |
1182 | * different, but the operation would fail silently.) */ | |
a5fac1df | 1183 | r = sysctl_write("fs/file-max", LONG_MAX_STR); |
a8b627aa | 1184 | if (r < 0) |
b47e0fac ZJS |
1185 | log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, |
1186 | r, "Failed to bump fs.file-max, ignoring: %m"); | |
a8b627aa LP |
1187 | #endif |
1188 | ||
a8b627aa LP |
1189 | #if BUMP_PROC_SYS_FS_NR_OPEN |
1190 | int v = INT_MAX; | |
1191 | ||
ff3a7019 ZJS |
1192 | /* Argh! The kernel enforces maximum and minimum values on the fs.nr_open, but we don't really know |
1193 | * what they are. The expression by which the maximum is determined is dependent on the architecture, | |
1194 | * and is something we don't really want to copy to userspace, as it is dependent on implementation | |
1195 | * details of the kernel. Since the kernel doesn't expose the maximum value to us, we can only try | |
1196 | * and hope. Hence, let's start with INT_MAX, and then keep halving the value until we find one that | |
1197 | * works. Ugly? Yes, absolutely, but kernel APIs are kernel APIs, so what do can we do... 🤯 */ | |
a8b627aa LP |
1198 | |
1199 | for (;;) { | |
1200 | int k; | |
1201 | ||
1202 | v &= ~(__SIZEOF_POINTER__ - 1); /* Round down to next multiple of the pointer size */ | |
1203 | if (v < 1024) { | |
1204 | log_warning("Can't bump fs.nr_open, value too small."); | |
1205 | break; | |
1206 | } | |
1207 | ||
1208 | k = read_nr_open(); | |
1209 | if (k < 0) { | |
1210 | log_error_errno(k, "Failed to read fs.nr_open: %m"); | |
1211 | break; | |
1212 | } | |
1213 | if (k >= v) { /* Already larger */ | |
1214 | log_debug("Skipping bump, value is already larger."); | |
1215 | break; | |
1216 | } | |
1217 | ||
b47e0fac | 1218 | r = sysctl_writef("fs/nr_open", "%i", v); |
a8b627aa LP |
1219 | if (r == -EINVAL) { |
1220 | log_debug("Couldn't write fs.nr_open as %i, halving it.", v); | |
1221 | v /= 2; | |
1222 | continue; | |
1223 | } | |
1224 | if (r < 0) { | |
1225 | log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, "Failed to bump fs.nr_open, ignoring: %m"); | |
1226 | break; | |
1227 | } | |
1228 | ||
1229 | log_debug("Successfully bumped fs.nr_open to %i", v); | |
1230 | break; | |
1231 | } | |
1232 | #endif | |
1233 | } | |
1234 | ||
eadb4f19 | 1235 | static int bump_rlimit_nofile(const struct rlimit *saved_rlimit) { |
cda7faa9 | 1236 | struct rlimit new_rlimit; |
9264cc39 | 1237 | int r, nr; |
4096d6f5 | 1238 | |
52d62075 LP |
1239 | /* Get the underlying absolute limit the kernel enforces */ |
1240 | nr = read_nr_open(); | |
1241 | ||
cda7faa9 LP |
1242 | /* Calculate the new limits to use for us. Never lower from what we inherited. */ |
1243 | new_rlimit = (struct rlimit) { | |
1244 | .rlim_cur = MAX((rlim_t) nr, saved_rlimit->rlim_cur), | |
1245 | .rlim_max = MAX((rlim_t) nr, saved_rlimit->rlim_max), | |
1246 | }; | |
1247 | ||
1248 | /* Shortcut if nothing changes. */ | |
1249 | if (saved_rlimit->rlim_max >= new_rlimit.rlim_max && | |
1250 | saved_rlimit->rlim_cur >= new_rlimit.rlim_cur) { | |
1251 | log_debug("RLIMIT_NOFILE is already as high or higher than we need it, not bumping."); | |
1252 | return 0; | |
1253 | } | |
1254 | ||
52d62075 LP |
1255 | /* Bump up the resource limit for ourselves substantially, all the way to the maximum the kernel allows, for |
1256 | * both hard and soft. */ | |
cda7faa9 | 1257 | r = setrlimit_closest(RLIMIT_NOFILE, &new_rlimit); |
23bbb0de | 1258 | if (r < 0) |
3ce40911 | 1259 | return log_warning_errno(r, "Setting RLIMIT_NOFILE failed, ignoring: %m"); |
4096d6f5 LP |
1260 | |
1261 | return 0; | |
1262 | } | |
1263 | ||
eadb4f19 | 1264 | static int bump_rlimit_memlock(const struct rlimit *saved_rlimit) { |
cda7faa9 | 1265 | struct rlimit new_rlimit; |
04d1ee0f | 1266 | uint64_t mm; |
fb3ae275 LP |
1267 | int r; |
1268 | ||
ff3a7019 ZJS |
1269 | /* BPF_MAP_TYPE_LPM_TRIE bpf maps are charged against RLIMIT_MEMLOCK, even if we have CAP_IPC_LOCK |
1270 | * which should normally disable such checks. We need them to implement IPAddressAllow= and | |
1271 | * IPAddressDeny=, hence let's bump the value high enough for our user. */ | |
fb3ae275 | 1272 | |
cda7faa9 LP |
1273 | /* Using MAX() on resource limits only is safe if RLIM_INFINITY is > 0. POSIX declares that rlim_t |
1274 | * must be unsigned, hence this is a given, but let's make this clear here. */ | |
1275 | assert_cc(RLIM_INFINITY > 0); | |
1276 | ||
ff3a7019 ZJS |
1277 | mm = physical_memory_scale(1, 8); /* Let's scale how much we allow to be locked by the amount of |
1278 | * physical RAM. We allow an eighth to be locked by us, just to | |
1279 | * pick a value. */ | |
04d1ee0f | 1280 | |
cda7faa9 | 1281 | new_rlimit = (struct rlimit) { |
04d1ee0f LP |
1282 | .rlim_cur = MAX3(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_cur, mm), |
1283 | .rlim_max = MAX3(HIGH_RLIMIT_MEMLOCK, saved_rlimit->rlim_max, mm), | |
cda7faa9 LP |
1284 | }; |
1285 | ||
1286 | if (saved_rlimit->rlim_max >= new_rlimit.rlim_cur && | |
1287 | saved_rlimit->rlim_cur >= new_rlimit.rlim_max) { | |
1288 | log_debug("RLIMIT_MEMLOCK is already as high or higher than we need it, not bumping."); | |
1289 | return 0; | |
1290 | } | |
1291 | ||
1292 | r = setrlimit_closest(RLIMIT_MEMLOCK, &new_rlimit); | |
fb3ae275 LP |
1293 | if (r < 0) |
1294 | return log_warning_errno(r, "Setting RLIMIT_MEMLOCK failed, ignoring: %m"); | |
1295 | ||
1296 | return 0; | |
1297 | } | |
1298 | ||
80758717 | 1299 | static void test_usr(void) { |
80758717 | 1300 | |
796ac4c1 | 1301 | /* Check that /usr is either on the same file system as / or mounted already. */ |
80758717 | 1302 | |
db55bbf2 | 1303 | if (dir_is_empty("/usr", /* ignore_hidden_or_backup= */ false) <= 0) |
871c44a7 LP |
1304 | return; |
1305 | ||
8b173b5e | 1306 | log_warning("/usr appears to be on its own filesystem and is not already mounted. This is not a supported setup. " |
871c44a7 | 1307 | "Some things will probably break (sometimes even silently) in mysterious ways. " |
a25d9395 | 1308 | "Consult https://www.freedesktop.org/wiki/Software/systemd/separate-usr-is-broken for more information."); |
871c44a7 LP |
1309 | } |
1310 | ||
d3b1c508 | 1311 | static int enforce_syscall_archs(Set *archs) { |
349cc4a5 | 1312 | #if HAVE_SECCOMP |
d3b1c508 LP |
1313 | int r; |
1314 | ||
83f12b27 FS |
1315 | if (!is_seccomp_available()) |
1316 | return 0; | |
1317 | ||
469830d1 | 1318 | r = seccomp_restrict_archs(arg_syscall_archs); |
d3b1c508 | 1319 | if (r < 0) |
469830d1 | 1320 | return log_error_errno(r, "Failed to enforce system call architecture restrication: %m"); |
d3b1c508 | 1321 | #endif |
469830d1 | 1322 | return 0; |
d3b1c508 LP |
1323 | } |
1324 | ||
4bd03515 ZJS |
1325 | static int os_release_status(void) { |
1326 | _cleanup_free_ char *pretty_name = NULL, *name = NULL, *version = NULL, | |
1327 | *ansi_color = NULL, *support_end = NULL; | |
b6e2f329 LP |
1328 | int r; |
1329 | ||
d58ad743 LP |
1330 | r = parse_os_release(NULL, |
1331 | "PRETTY_NAME", &pretty_name, | |
4bd03515 ZJS |
1332 | "NAME", &name, |
1333 | "VERSION", &version, | |
1334 | "ANSI_COLOR", &ansi_color, | |
1335 | "SUPPORT_END", &support_end); | |
d58ad743 | 1336 | if (r < 0) |
4bd03515 ZJS |
1337 | return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, |
1338 | "Failed to read os-release file, ignoring: %m"); | |
b6e2f329 | 1339 | |
02b7005e | 1340 | const char *label = os_release_pretty_name(pretty_name, name); |
4bd03515 ZJS |
1341 | |
1342 | if (show_status_on(arg_show_status)) { | |
1343 | if (log_get_show_color()) | |
1344 | status_printf(NULL, 0, | |
1345 | "\nWelcome to \x1B[%sm%s\x1B[0m!\n", | |
1346 | empty_to_null(ansi_color) ?: "1", | |
1347 | label); | |
1348 | else | |
1349 | status_printf(NULL, 0, | |
1350 | "\nWelcome to %s!\n", | |
1351 | label); | |
1352 | } | |
1353 | ||
6bfe9b3b | 1354 | if (support_end && os_release_support_ended(support_end, /* quiet */ false, NULL) > 0) |
4bd03515 ZJS |
1355 | /* pretty_name may include the version already, so we'll print the version only if we |
1356 | * have it and we're not using pretty_name. */ | |
1357 | status_printf(ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL, 0, | |
1358 | "This OS version (%s%s%s) is past its end-of-support date (%s)", | |
1359 | label, | |
1360 | (pretty_name || !version) ? "" : " version ", | |
1361 | (pretty_name || !version) ? "" : version, | |
1362 | support_end); | |
1363 | ||
1364 | return 0; | |
b6e2f329 LP |
1365 | } |
1366 | ||
3f37a825 LB |
1367 | static int setup_os_release(RuntimeScope scope) { |
1368 | _cleanup_free_ char *os_release_dst = NULL; | |
1369 | const char *os_release_src = "/etc/os-release"; | |
1370 | int r; | |
1371 | ||
1372 | if (access("/etc/os-release", F_OK) < 0) { | |
1373 | if (errno != ENOENT) | |
1374 | log_debug_errno(errno, "Failed to check if /etc/os-release exists, ignoring: %m"); | |
1375 | ||
1376 | os_release_src = "/usr/lib/os-release"; | |
1377 | } | |
1378 | ||
1379 | if (scope == RUNTIME_SCOPE_SYSTEM) { | |
663e2756 | 1380 | os_release_dst = strdup("/run/systemd/propagate/.os-release-stage/os-release"); |
3f37a825 LB |
1381 | if (!os_release_dst) |
1382 | return log_oom_debug(); | |
1383 | } else { | |
663e2756 | 1384 | if (asprintf(&os_release_dst, "/run/user/" UID_FMT "/systemd/propagate/.os-release-stage/os-release", geteuid()) < 0) |
3f37a825 LB |
1385 | return log_oom_debug(); |
1386 | } | |
1387 | ||
1388 | r = mkdir_parents_label(os_release_dst, 0755); | |
bf85c239 | 1389 | if (r < 0) |
3f37a825 LB |
1390 | return log_debug_errno(r, "Failed to create parent directory of %s, ignoring: %m", os_release_dst); |
1391 | ||
663e2756 | 1392 | r = copy_file_atomic(os_release_src, os_release_dst, 0644, COPY_MAC_CREATE|COPY_REPLACE); |
3f37a825 LB |
1393 | if (r < 0) |
1394 | return log_debug_errno(r, "Failed to create %s, ignoring: %m", os_release_dst); | |
1395 | ||
1396 | return 0; | |
1397 | } | |
1398 | ||
fdd25311 LP |
1399 | static int write_container_id(void) { |
1400 | const char *c; | |
7756528e | 1401 | int r = 0; /* avoid false maybe-uninitialized warning */ |
fdd25311 LP |
1402 | |
1403 | c = getenv("container"); | |
1404 | if (isempty(c)) | |
1405 | return 0; | |
1406 | ||
2053593f | 1407 | WITH_UMASK(0022) |
8612da97 | 1408 | r = write_string_file("/run/systemd/container", c, WRITE_STRING_FILE_CREATE); |
19854865 | 1409 | if (r < 0) |
f1f849b0 | 1410 | return log_warning_errno(r, "Failed to write /run/systemd/container, ignoring: %m"); |
19854865 LP |
1411 | |
1412 | return 1; | |
1413 | } | |
1414 | ||
1415 | static int bump_unix_max_dgram_qlen(void) { | |
1416 | _cleanup_free_ char *qlen = NULL; | |
1417 | unsigned long v; | |
1418 | int r; | |
1419 | ||
ff3a7019 ZJS |
1420 | /* Let's bump the net.unix.max_dgram_qlen sysctl. The kernel default of 16 is simply too low. We set |
1421 | * the value really really early during boot, so that it is actually applied to all our sockets, | |
1422 | * including the $NOTIFY_SOCKET one. */ | |
19854865 LP |
1423 | |
1424 | r = read_one_line_file("/proc/sys/net/unix/max_dgram_qlen", &qlen); | |
1425 | if (r < 0) | |
ff3a7019 ZJS |
1426 | return log_full_errno(r == -ENOENT ? LOG_DEBUG : LOG_WARNING, r, |
1427 | "Failed to read AF_UNIX datagram queue length, ignoring: %m"); | |
19854865 LP |
1428 | |
1429 | r = safe_atolu(qlen, &v); | |
1430 | if (r < 0) | |
3130fca5 | 1431 | return log_warning_errno(r, "Failed to parse AF_UNIX datagram queue length '%s', ignoring: %m", qlen); |
19854865 LP |
1432 | |
1433 | if (v >= DEFAULT_UNIX_MAX_DGRAM_QLEN) | |
1434 | return 0; | |
1435 | ||
27c8ca43 | 1436 | r = sysctl_write("net/unix/max_dgram_qlen", STRINGIFY(DEFAULT_UNIX_MAX_DGRAM_QLEN)); |
19854865 LP |
1437 | if (r < 0) |
1438 | return log_full_errno(IN_SET(r, -EROFS, -EPERM, -EACCES) ? LOG_DEBUG : LOG_WARNING, r, | |
1439 | "Failed to bump AF_UNIX datagram queue length, ignoring: %m"); | |
1440 | ||
1441 | return 1; | |
fdd25311 LP |
1442 | } |
1443 | ||
32391275 FB |
1444 | static int fixup_environment(void) { |
1445 | _cleanup_free_ char *term = NULL; | |
4dc63c4b | 1446 | const char *t; |
32391275 FB |
1447 | int r; |
1448 | ||
43db615b LP |
1449 | /* Only fix up the environment when we are started as PID 1 */ |
1450 | if (getpid_cached() != 1) | |
1451 | return 0; | |
1452 | ||
1453 | /* We expect the environment to be set correctly if run inside a container. */ | |
84af7821 LP |
1454 | if (detect_container() > 0) |
1455 | return 0; | |
1456 | ||
ff3a7019 ZJS |
1457 | /* When started as PID1, the kernel uses /dev/console for our stdios and uses TERM=linux whatever the |
1458 | * backend device used by the console. We try to make a better guess here since some consoles might | |
1459 | * not have support for color mode for example. | |
32391275 | 1460 | * |
43db615b | 1461 | * However if TERM was configured through the kernel command line then leave it alone. */ |
1d84ad94 | 1462 | r = proc_cmdline_get_key("TERM", 0, &term); |
32391275 FB |
1463 | if (r < 0) |
1464 | return r; | |
32391275 | 1465 | |
153d5dfd DDM |
1466 | if (r == 0) { |
1467 | r = proc_cmdline_get_key("systemd.tty.term.console", 0, &term); | |
1468 | if (r < 0) | |
1469 | return r; | |
1470 | } | |
1471 | ||
4dc63c4b LP |
1472 | t = term ?: default_term_for_tty("/dev/console"); |
1473 | ||
1474 | if (setenv("TERM", t, 1) < 0) | |
32391275 FB |
1475 | return -errno; |
1476 | ||
9d48671c | 1477 | /* The kernels sets HOME=/ for init. Let's undo this. */ |
44ee03d1 ZJS |
1478 | if (path_equal_ptr(getenv("HOME"), "/")) |
1479 | assert_se(unsetenv("HOME") == 0); | |
9d48671c | 1480 | |
32391275 FB |
1481 | return 0; |
1482 | } | |
1483 | ||
6808a0bc LP |
1484 | static void redirect_telinit(int argc, char *argv[]) { |
1485 | ||
1486 | /* This is compatibility support for SysV, where calling init as a user is identical to telinit. */ | |
1487 | ||
1488 | #if HAVE_SYSV_COMPAT | |
1489 | if (getpid_cached() == 1) | |
1490 | return; | |
1491 | ||
2306d177 | 1492 | if (!invoked_as(argv, "init")) |
6808a0bc LP |
1493 | return; |
1494 | ||
1495 | execv(SYSTEMCTL_BINARY_PATH, argv); | |
1496 | log_error_errno(errno, "Failed to exec " SYSTEMCTL_BINARY_PATH ": %m"); | |
a45d7127 | 1497 | exit(EXIT_FAILURE); |
6808a0bc LP |
1498 | #endif |
1499 | } | |
1500 | ||
b22d392d | 1501 | static int become_shutdown(int objective, int retval) { |
b22d392d VC |
1502 | static const char* const table[_MANAGER_OBJECTIVE_MAX] = { |
1503 | [MANAGER_EXIT] = "exit", | |
1504 | [MANAGER_REBOOT] = "reboot", | |
1505 | [MANAGER_POWEROFF] = "poweroff", | |
1506 | [MANAGER_HALT] = "halt", | |
1507 | [MANAGER_KEXEC] = "kexec", | |
1508 | }; | |
4a36297c | 1509 | |
d2ebd50d | 1510 | char log_level[STRLEN("--log-level=") + DECIMAL_STR_MAX(int)], |
4688b089 ZJS |
1511 | timeout[STRLEN("--timeout=") + DECIMAL_STR_MAX(usec_t) + STRLEN("us")], |
1512 | exit_code[STRLEN("--exit-code=") + DECIMAL_STR_MAX(uint8_t)]; | |
4a36297c LP |
1513 | |
1514 | _cleanup_strv_free_ char **env_block = NULL; | |
f16890f8 | 1515 | usec_t watchdog_timer = 0; |
4a36297c LP |
1516 | int r; |
1517 | ||
b22d392d VC |
1518 | assert(objective >= 0 && objective < _MANAGER_OBJECTIVE_MAX); |
1519 | assert(table[objective]); | |
4a36297c | 1520 | |
d2ebd50d | 1521 | xsprintf(log_level, "--log-level=%d", log_get_max_level()); |
c9e120e0 | 1522 | xsprintf(timeout, "--timeout=%" PRI_USEC "us", arg_defaults.timeout_stop_usec); |
4a36297c | 1523 | |
4688b089 ZJS |
1524 | const char* command_line[10] = { |
1525 | SYSTEMD_SHUTDOWN_BINARY_PATH, | |
1526 | table[objective], | |
1527 | log_level, | |
1528 | timeout, | |
1529 | /* Note that the last position is a terminator and must contain NULL. */ | |
1530 | }; | |
1531 | size_t pos = 4; | |
1532 | ||
1533 | assert(command_line[pos-1]); | |
1534 | assert(!command_line[pos]); | |
1535 | ||
4a36297c LP |
1536 | switch (log_get_target()) { |
1537 | ||
1538 | case LOG_TARGET_KMSG: | |
1539 | case LOG_TARGET_JOURNAL_OR_KMSG: | |
1540 | case LOG_TARGET_SYSLOG_OR_KMSG: | |
d2ebd50d | 1541 | command_line[pos++] = "--log-target=kmsg"; |
4a36297c LP |
1542 | break; |
1543 | ||
1544 | case LOG_TARGET_NULL: | |
6b7f150b | 1545 | command_line[pos++] = "--log-target=null"; |
4a36297c LP |
1546 | break; |
1547 | ||
1548 | case LOG_TARGET_CONSOLE: | |
1549 | default: | |
d2ebd50d | 1550 | command_line[pos++] = "--log-target=console"; |
4a36297c LP |
1551 | break; |
1552 | }; | |
1553 | ||
1554 | if (log_get_show_color()) | |
1555 | command_line[pos++] = "--log-color"; | |
1556 | ||
1557 | if (log_get_show_location()) | |
1558 | command_line[pos++] = "--log-location"; | |
1559 | ||
c5673ed0 DS |
1560 | if (log_get_show_time()) |
1561 | command_line[pos++] = "--log-time"; | |
1562 | ||
3f92250f DDM |
1563 | xsprintf(exit_code, "--exit-code=%d", retval); |
1564 | command_line[pos++] = exit_code; | |
4a36297c | 1565 | |
14ecfc1c | 1566 | assert(pos < ELEMENTSOF(command_line)); |
4688b089 ZJS |
1567 | |
1568 | /* The watchdog: */ | |
4a36297c | 1569 | |
b22d392d | 1570 | if (objective == MANAGER_REBOOT) |
65224c1d | 1571 | watchdog_timer = arg_reboot_watchdog; |
b22d392d | 1572 | else if (objective == MANAGER_KEXEC) |
acafd7d8 LB |
1573 | watchdog_timer = arg_kexec_watchdog; |
1574 | ||
f16890f8 | 1575 | /* If we reboot or kexec let's set the shutdown watchdog and tell the |
5717062e CK |
1576 | * shutdown binary to repeatedly ping it. |
1577 | * Disable the pretimeout watchdog, as we do not support it from the shutdown binary. */ | |
1578 | (void) watchdog_setup_pretimeout(0); | |
aff3a9e1 | 1579 | (void) watchdog_setup_pretimeout_governor(NULL); |
f16890f8 FB |
1580 | r = watchdog_setup(watchdog_timer); |
1581 | watchdog_close(r < 0); | |
4a36297c | 1582 | |
4688b089 ZJS |
1583 | /* The environment block: */ |
1584 | ||
1585 | env_block = strv_copy(environ); | |
1586 | ||
f16890f8 FB |
1587 | /* Tell the binary how often to ping, ignore failure */ |
1588 | (void) strv_extendf(&env_block, "WATCHDOG_USEC="USEC_FMT, watchdog_timer); | |
8a2c1fbf | 1589 | |
f16890f8 FB |
1590 | if (arg_watchdog_device) |
1591 | (void) strv_extendf(&env_block, "WATCHDOG_DEVICE=%s", arg_watchdog_device); | |
4a36297c | 1592 | |
3178d23d FB |
1593 | /* Avoid the creation of new processes forked by the kernel; at this |
1594 | * point, we will not listen to the signals anyway */ | |
4a36297c LP |
1595 | if (detect_container() <= 0) |
1596 | (void) cg_uninstall_release_agent(SYSTEMD_CGROUP_CONTROLLER); | |
1597 | ||
1598 | execve(SYSTEMD_SHUTDOWN_BINARY_PATH, (char **) command_line, env_block); | |
1599 | return -errno; | |
1600 | } | |
1601 | ||
e839bafd LP |
1602 | static void initialize_clock(void) { |
1603 | int r; | |
1604 | ||
3753325b LP |
1605 | /* This is called very early on, before we parse the kernel command line or otherwise figure out why |
1606 | * we are running, but only once. */ | |
1607 | ||
e839bafd LP |
1608 | if (clock_is_localtime(NULL) > 0) { |
1609 | int min; | |
1610 | ||
ff3a7019 | 1611 | /* The very first call of settimeofday() also does a time warp in the kernel. |
e839bafd | 1612 | * |
ff3a7019 ZJS |
1613 | * In the rtc-in-local time mode, we set the kernel's timezone, and rely on external tools to |
1614 | * take care of maintaining the RTC and do all adjustments. This matches the behavior of | |
1615 | * Windows, which leaves the RTC alone if the registry tells that the RTC runs in UTC. | |
e839bafd LP |
1616 | */ |
1617 | r = clock_set_timezone(&min); | |
1618 | if (r < 0) | |
1619 | log_error_errno(r, "Failed to apply local time delta, ignoring: %m"); | |
1620 | else | |
1621 | log_info("RTC configured in localtime, applying delta of %i minutes to system time.", min); | |
1622 | ||
d46b79bb | 1623 | } else if (!in_initrd()) |
e839bafd LP |
1624 | /* |
1625 | * Do a dummy very first call to seal the kernel's time warp magic. | |
1626 | * | |
ff3a7019 ZJS |
1627 | * Do not call this from inside the initrd. The initrd might not carry /etc/adjtime with |
1628 | * LOCAL, but the real system could be set up that way. In such case, we need to delay the | |
1629 | * time-warp or the sealing until we reach the real system. | |
e839bafd | 1630 | * |
ff3a7019 ZJS |
1631 | * Do no set the kernel's timezone. The concept of local time cannot be supported reliably, |
1632 | * the time will jump or be incorrect at every daylight saving time change. All kernel local | |
1633 | * time concepts will be treated as UTC that way. | |
e839bafd LP |
1634 | */ |
1635 | (void) clock_reset_timewarp(); | |
e839bafd | 1636 | |
b10abe4b EI |
1637 | ClockChangeDirection change_dir; |
1638 | r = clock_apply_epoch(&change_dir); | |
1639 | if (r > 0 && change_dir == CLOCK_CHANGE_FORWARD) | |
e839bafd | 1640 | log_info("System time before build time, advancing clock."); |
b10abe4b EI |
1641 | else if (r > 0 && change_dir == CLOCK_CHANGE_BACKWARD) |
1642 | log_info("System time is further ahead than %s after build time, resetting clock to build time.", | |
1643 | FORMAT_TIMESPAN(CLOCK_VALID_RANGE_USEC_MAX, USEC_PER_DAY)); | |
1644 | else if (r < 0 && change_dir == CLOCK_CHANGE_FORWARD) | |
1645 | log_error_errno(r, "Current system time is before build time, but cannot correct: %m"); | |
1646 | else if (r < 0 && change_dir == CLOCK_CHANGE_BACKWARD) | |
1647 | log_error_errno(r, "Current system time is further ahead %s after build time, but cannot correct: %m", | |
1648 | FORMAT_TIMESPAN(CLOCK_VALID_RANGE_USEC_MAX, USEC_PER_DAY)); | |
e839bafd LP |
1649 | } |
1650 | ||
3753325b | 1651 | static void apply_clock_update(void) { |
3753325b LP |
1652 | /* This is called later than initialize_clock(), i.e. after we parsed configuration files/kernel |
1653 | * command line and such. */ | |
1654 | ||
1655 | if (arg_clock_usec == 0) | |
1656 | return; | |
1657 | ||
45250e66 LP |
1658 | if (getpid_cached() != 1) |
1659 | return; | |
1660 | ||
52bb308c | 1661 | if (clock_settime(CLOCK_REALTIME, TIMESPEC_STORE(arg_clock_usec)) < 0) |
3753325b | 1662 | log_error_errno(errno, "Failed to set system clock to time specified on kernel command line: %m"); |
04f5c018 | 1663 | else |
3753325b | 1664 | log_info("Set system clock to %s, as specified on the kernel command line.", |
04f5c018 | 1665 | FORMAT_TIMESTAMP(arg_clock_usec)); |
3753325b LP |
1666 | } |
1667 | ||
d247f232 | 1668 | static void cmdline_take_random_seed(void) { |
d247f232 LP |
1669 | size_t suggested; |
1670 | int r; | |
1671 | ||
1672 | if (arg_random_seed_size == 0) | |
1673 | return; | |
1674 | ||
1675 | if (getpid_cached() != 1) | |
1676 | return; | |
1677 | ||
1678 | assert(arg_random_seed); | |
1679 | suggested = random_pool_size(); | |
1680 | ||
1681 | if (arg_random_seed_size < suggested) | |
1682 | log_warning("Random seed specified on kernel command line has size %zu, but %zu bytes required to fill entropy pool.", | |
1683 | arg_random_seed_size, suggested); | |
1684 | ||
61bd7d1e | 1685 | r = random_write_entropy(-1, arg_random_seed, arg_random_seed_size, true); |
d247f232 LP |
1686 | if (r < 0) { |
1687 | log_warning_errno(r, "Failed to credit entropy specified on kernel command line, ignoring: %m"); | |
1688 | return; | |
1689 | } | |
1690 | ||
1691 | log_notice("Successfully credited entropy passed on kernel command line.\n" | |
ff3a7019 ZJS |
1692 | "Note that the seed provided this way is accessible to unprivileged programs. " |
1693 | "This functionality should not be used outside of testing environments."); | |
d247f232 LP |
1694 | } |
1695 | ||
1e41242e | 1696 | static void initialize_coredump(bool skip_setup) { |
1e41242e LP |
1697 | if (getpid_cached() != 1) |
1698 | return; | |
1699 | ||
ff3a7019 ZJS |
1700 | /* Don't limit the core dump size, so that coredump handlers such as systemd-coredump (which honour |
1701 | * the limit) will process core dumps for system services by default. */ | |
1e41242e LP |
1702 | if (setrlimit(RLIMIT_CORE, &RLIMIT_MAKE_CONST(RLIM_INFINITY)) < 0) |
1703 | log_warning_errno(errno, "Failed to set RLIMIT_CORE: %m"); | |
1704 | ||
ff3a7019 ZJS |
1705 | /* But at the same time, turn off the core_pattern logic by default, so that no coredumps are stored |
1706 | * until the systemd-coredump tool is enabled via sysctl. However it can be changed via the kernel | |
32e27670 | 1707 | * command line later so core dumps can still be generated during early startup and in initrd. */ |
1e41242e | 1708 | if (!skip_setup) |
e557b1a6 | 1709 | disable_coredumps(); |
1e41242e LP |
1710 | } |
1711 | ||
c6885f5f FB |
1712 | static void initialize_core_pattern(bool skip_setup) { |
1713 | int r; | |
1714 | ||
1715 | if (skip_setup || !arg_early_core_pattern) | |
1716 | return; | |
1717 | ||
1718 | if (getpid_cached() != 1) | |
1719 | return; | |
1720 | ||
57512c89 | 1721 | r = write_string_file("/proc/sys/kernel/core_pattern", arg_early_core_pattern, WRITE_STRING_FILE_DISABLE_BUFFER); |
c6885f5f | 1722 | if (r < 0) |
ff3a7019 ZJS |
1723 | log_warning_errno(r, "Failed to write '%s' to /proc/sys/kernel/core_pattern, ignoring: %m", |
1724 | arg_early_core_pattern); | |
c6885f5f FB |
1725 | } |
1726 | ||
ffc1ec73 LP |
1727 | static void apply_protect_system(bool skip_setup) { |
1728 | int r; | |
1729 | ||
1730 | if (skip_setup || getpid_cached() != 1 || arg_protect_system == 0) | |
1731 | return; | |
1732 | ||
1733 | if (arg_protect_system < 0 && !in_initrd()) { | |
1734 | log_debug("ProtectSystem=auto selected, but not running in an initrd, skipping."); | |
1735 | return; | |
1736 | } | |
1737 | ||
1738 | r = make_mount_point("/usr"); | |
1739 | if (r < 0) { | |
1740 | log_warning_errno(r, "Failed to make /usr/ a mount point, ignoring: %m"); | |
1741 | return; | |
1742 | } | |
1743 | ||
1744 | if (mount_nofollow_verbose( | |
1745 | LOG_WARNING, | |
1746 | /* what= */ NULL, | |
1747 | "/usr", | |
1748 | /* fstype= */ NULL, | |
1749 | MS_BIND|MS_REMOUNT|MS_RDONLY, | |
1750 | /* options= */ NULL) < 0) | |
1751 | return; | |
1752 | ||
1753 | log_info("Successfully made /usr/ read-only."); | |
1754 | } | |
1755 | ||
61fbbac1 ZJS |
1756 | static void update_cpu_affinity(bool skip_setup) { |
1757 | _cleanup_free_ char *mask = NULL; | |
1758 | ||
1759 | if (skip_setup || !arg_cpu_affinity.set) | |
1760 | return; | |
1761 | ||
1762 | assert(arg_cpu_affinity.allocated > 0); | |
1763 | ||
667030bf ZJS |
1764 | mask = cpu_set_to_range_string(&arg_cpu_affinity); |
1765 | log_debug("Setting CPU affinity to {%s}.", strnull(mask)); | |
61fbbac1 ZJS |
1766 | |
1767 | if (sched_setaffinity(0, arg_cpu_affinity.allocated, arg_cpu_affinity.set) < 0) | |
6b1fa539 | 1768 | log_warning_errno(errno, "Failed to set CPU affinity, ignoring: %m"); |
61fbbac1 ZJS |
1769 | } |
1770 | ||
b070c7c0 MS |
1771 | static void update_numa_policy(bool skip_setup) { |
1772 | int r; | |
1773 | _cleanup_free_ char *nodes = NULL; | |
1774 | const char * policy = NULL; | |
1775 | ||
1776 | if (skip_setup || !mpol_is_valid(numa_policy_get_type(&arg_numa_policy))) | |
1777 | return; | |
1778 | ||
1779 | if (DEBUG_LOGGING) { | |
1780 | policy = mpol_to_string(numa_policy_get_type(&arg_numa_policy)); | |
1781 | nodes = cpu_set_to_range_string(&arg_numa_policy.nodes); | |
667030bf | 1782 | log_debug("Setting NUMA policy to %s, with nodes {%s}.", strnull(policy), strnull(nodes)); |
b070c7c0 MS |
1783 | } |
1784 | ||
1785 | r = apply_numa_policy(&arg_numa_policy); | |
1786 | if (r == -EOPNOTSUPP) | |
1787 | log_debug_errno(r, "NUMA support not available, ignoring."); | |
1788 | else if (r < 0) | |
6b1fa539 | 1789 | log_warning_errno(r, "Failed to set NUMA memory policy, ignoring: %m"); |
b070c7c0 MS |
1790 | } |
1791 | ||
19fd72df LP |
1792 | static void filter_args( |
1793 | const char* dst[], | |
1794 | size_t *dst_index, | |
1795 | char **src, | |
1796 | int argc) { | |
1797 | ||
846f1da4 | 1798 | assert(dst); |
19fd72df | 1799 | assert(dst_index); |
846f1da4 ZJS |
1800 | |
1801 | /* Copy some filtered arguments into the dst array from src. */ | |
1802 | for (int i = 1; i < argc; i++) { | |
1803 | if (STR_IN_SET(src[i], | |
1804 | "--switched-root", | |
1805 | "--system", | |
1806 | "--user")) | |
1807 | continue; | |
1808 | ||
1809 | if (startswith(src[i], "--deserialize=")) | |
1810 | continue; | |
1811 | if (streq(src[i], "--deserialize")) { | |
1812 | i++; /* Skip the argument too */ | |
1813 | continue; | |
1814 | } | |
1815 | ||
1816 | /* Skip target unit designators. We already acted upon this information and have queued | |
1817 | * appropriate jobs. We don't want to redo all this after reexecution. */ | |
1818 | if (startswith(src[i], "--unit=")) | |
1819 | continue; | |
1820 | if (streq(src[i], "--unit")) { | |
1821 | i++; /* Skip the argument too */ | |
1822 | continue; | |
1823 | } | |
1824 | ||
846f1da4 | 1825 | /* Seems we have a good old option. Let's pass it over to the new instance. */ |
19fd72df | 1826 | dst[(*dst_index)++] = src[i]; |
846f1da4 ZJS |
1827 | } |
1828 | } | |
1829 | ||
05be3e8b LB |
1830 | static void finish_remaining_processes(ManagerObjective objective) { |
1831 | assert(objective >= 0 && objective < _MANAGER_OBJECTIVE_MAX); | |
1832 | ||
1833 | /* Kill all remaining processes from the initrd, but don't wait for them, so that we can handle the | |
1834 | * SIGCHLD for them after deserializing. */ | |
1835 | if (IN_SET(objective, MANAGER_SWITCH_ROOT, MANAGER_SOFT_REBOOT)) | |
c9e120e0 | 1836 | broadcast_signal(SIGTERM, /* wait_for_exit= */ false, /* send_sighup= */ true, arg_defaults.timeout_stop_usec); |
05be3e8b | 1837 | |
559214cb LB |
1838 | /* On soft reboot really make sure nothing is left. Note that this will skip cgroups |
1839 | * of units that were configured with SurviveFinalKillSignal=yes. */ | |
05be3e8b | 1840 | if (objective == MANAGER_SOFT_REBOOT) |
c9e120e0 | 1841 | broadcast_signal(SIGKILL, /* wait_for_exit= */ false, /* send_sighup= */ false, arg_defaults.timeout_stop_usec); |
05be3e8b LB |
1842 | } |
1843 | ||
1e3eee8c ZJS |
1844 | static int do_reexecute( |
1845 | ManagerObjective objective, | |
3c7878f9 | 1846 | int argc, |
846f1da4 | 1847 | char* argv[], |
3c7878f9 LP |
1848 | const struct rlimit *saved_rlimit_nofile, |
1849 | const struct rlimit *saved_rlimit_memlock, | |
1850 | FDSet *fds, | |
1851 | const char *switch_root_dir, | |
1852 | const char *switch_root_init, | |
1853 | const char **ret_error_message) { | |
1854 | ||
19fd72df | 1855 | size_t i, args_size; |
3c7878f9 LP |
1856 | const char **args; |
1857 | int r; | |
1858 | ||
13ffc607 | 1859 | assert(IN_SET(objective, MANAGER_REEXECUTE, MANAGER_SWITCH_ROOT, MANAGER_SOFT_REBOOT)); |
19fd72df | 1860 | assert(argc >= 0); |
3c7878f9 LP |
1861 | assert(saved_rlimit_nofile); |
1862 | assert(saved_rlimit_memlock); | |
1863 | assert(ret_error_message); | |
1864 | ||
aaa27e2e DDM |
1865 | if (switch_root_init) { |
1866 | r = chase(switch_root_init, switch_root_dir, CHASE_PREFIX_ROOT, NULL, NULL); | |
1867 | if (r < 0) | |
1868 | log_warning_errno(r, "Failed to chase configured init %s/%s: %m", | |
1869 | strempty(switch_root_dir), switch_root_init); | |
1870 | } else { | |
1871 | r = chase(SYSTEMD_BINARY_PATH, switch_root_dir, CHASE_PREFIX_ROOT, NULL, NULL); | |
1872 | if (r < 0) | |
1873 | log_debug_errno(r, "Failed to chase our own binary %s/%s: %m", | |
1874 | strempty(switch_root_dir), SYSTEMD_BINARY_PATH); | |
1875 | } | |
1876 | ||
1877 | if (r < 0) { | |
1878 | r = chase("/sbin/init", switch_root_dir, CHASE_PREFIX_ROOT, NULL, NULL); | |
1879 | if (r < 0) | |
1880 | return log_error_errno(r, "Failed to chase %s/sbin/init", strempty(switch_root_dir)); | |
1881 | } | |
1882 | ||
ff3a7019 ZJS |
1883 | /* Close and disarm the watchdog, so that the new instance can reinitialize it, but doesn't get |
1884 | * rebooted while we do that */ | |
3c7878f9 LP |
1885 | watchdog_close(true); |
1886 | ||
ddfa8b0b LP |
1887 | /* Reset RLIMIT_NOFILE + RLIMIT_MEMLOCK back to the kernel defaults, so that the new systemd can pass |
1888 | * the kernel default to its child processes */ | |
1889 | if (saved_rlimit_nofile->rlim_cur != 0) | |
3c7878f9 | 1890 | (void) setrlimit(RLIMIT_NOFILE, saved_rlimit_nofile); |
ddfa8b0b | 1891 | if (saved_rlimit_memlock->rlim_cur != RLIM_INFINITY) |
3c7878f9 LP |
1892 | (void) setrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock); |
1893 | ||
05be3e8b | 1894 | finish_remaining_processes(objective); |
13ffc607 LP |
1895 | |
1896 | if (!switch_root_dir && objective == MANAGER_SOFT_REBOOT) { | |
1897 | /* If no switch root dir is specified, then check if /run/nextroot/ qualifies and use that */ | |
1898 | r = path_is_os_tree("/run/nextroot"); | |
1899 | if (r < 0 && r != -ENOENT) | |
1900 | log_debug_errno(r, "Failed to determine if /run/nextroot/ is a valid OS tree, ignoring: %m"); | |
1901 | else if (r > 0) | |
1902 | switch_root_dir = "/run/nextroot"; | |
1903 | } | |
3c7878f9 | 1904 | |
13ffc607 | 1905 | if (switch_root_dir) { |
13ffc607 LP |
1906 | r = switch_root(/* new_root= */ switch_root_dir, |
1907 | /* old_root_after= */ NULL, | |
b12d41a8 | 1908 | /* flags= */ (objective == MANAGER_SWITCH_ROOT ? SWITCH_ROOT_DESTROY_OLD_ROOT : 0) | |
c2d62118 | 1909 | (objective == MANAGER_SOFT_REBOOT ? 0 : SWITCH_ROOT_RECURSIVE_RUN)); |
3c7878f9 LP |
1910 | if (r < 0) |
1911 | log_error_errno(r, "Failed to switch root, trying to continue: %m"); | |
1912 | } | |
1913 | ||
d2ebd50d | 1914 | args_size = argc + 5; |
3c7878f9 LP |
1915 | args = newa(const char*, args_size); |
1916 | ||
1917 | if (!switch_root_init) { | |
d2ebd50d | 1918 | char sfd[STRLEN("--deserialize=") + DECIMAL_STR_MAX(int)]; |
3c7878f9 | 1919 | |
ff3a7019 ZJS |
1920 | /* First try to spawn ourselves with the right path, and with full serialization. We do this |
1921 | * only if the user didn't specify an explicit init to spawn. */ | |
3c7878f9 LP |
1922 | |
1923 | assert(arg_serialization); | |
1924 | assert(fds); | |
1925 | ||
d2ebd50d | 1926 | xsprintf(sfd, "--deserialize=%i", fileno(arg_serialization)); |
3c7878f9 | 1927 | |
846f1da4 | 1928 | i = 1; /* Leave args[0] empty for now. */ |
846f1da4 | 1929 | |
06afda6b FS |
1930 | /* Put our stuff first to make sure it always gets parsed in case |
1931 | * we get weird stuff from the kernel cmdline (like --) */ | |
13ffc607 | 1932 | if (IN_SET(objective, MANAGER_SWITCH_ROOT, MANAGER_SOFT_REBOOT)) |
3c7878f9 | 1933 | args[i++] = "--switched-root"; |
40d73340 | 1934 | args[i++] = runtime_scope_cmdline_option_to_string(arg_runtime_scope); |
3c7878f9 | 1935 | args[i++] = sfd; |
06afda6b FS |
1936 | |
1937 | filter_args(args, &i, argv, argc); | |
1938 | ||
3c7878f9 LP |
1939 | args[i++] = NULL; |
1940 | ||
1941 | assert(i <= args_size); | |
1942 | ||
1943 | /* | |
50b35193 ZJS |
1944 | * We want valgrind to print its memory usage summary before reexecution. Valgrind won't do |
1945 | * this is on its own on exec(), but it will do it on exit(). Hence, to ensure we get a | |
ff3a7019 ZJS |
1946 | * summary here, fork() off a child, let it exit() cleanly, so that it prints the summary, |
1947 | * and wait() for it in the parent, before proceeding into the exec(). | |
3c7878f9 LP |
1948 | */ |
1949 | valgrind_summary_hack(); | |
1950 | ||
846f1da4 | 1951 | args[0] = SYSTEMD_BINARY_PATH; |
3c7878f9 | 1952 | (void) execv(args[0], (char* const*) args); |
1e3eee8c ZJS |
1953 | |
1954 | if (objective == MANAGER_REEXECUTE) { | |
1955 | *ret_error_message = "Failed to execute our own binary"; | |
1956 | return log_error_errno(errno, "Failed to execute our own binary %s: %m", args[0]); | |
1957 | } | |
1958 | ||
846f1da4 | 1959 | log_debug_errno(errno, "Failed to execute our own binary %s, trying fallback: %m", args[0]); |
3c7878f9 LP |
1960 | } |
1961 | ||
ff3a7019 ZJS |
1962 | /* Try the fallback, if there is any, without any serialization. We pass the original argv[] and |
1963 | * envp[]. (Well, modulo the ordering changes due to getopt() in argv[], and some cleanups in envp[], | |
1964 | * but let's hope that doesn't matter.) */ | |
3c7878f9 LP |
1965 | |
1966 | arg_serialization = safe_fclose(arg_serialization); | |
1967 | fds = fdset_free(fds); | |
1968 | ||
1969 | /* Reopen the console */ | |
1970 | (void) make_console_stdio(); | |
1971 | ||
846f1da4 ZJS |
1972 | i = 1; /* Leave args[0] empty for now. */ |
1973 | for (int j = 1; j <= argc; j++) | |
3c7878f9 | 1974 | args[i++] = argv[j]; |
3c7878f9 LP |
1975 | assert(i <= args_size); |
1976 | ||
55c041b4 | 1977 | /* Re-enable any blocked signals, especially important if we switch from initrd to init=... */ |
3c7878f9 LP |
1978 | (void) reset_all_signal_handlers(); |
1979 | (void) reset_signal_mask(); | |
595225af | 1980 | (void) rlimit_nofile_safe(); |
3c7878f9 LP |
1981 | |
1982 | if (switch_root_init) { | |
1983 | args[0] = switch_root_init; | |
a5cede8c | 1984 | (void) execve(args[0], (char* const*) args, saved_env); |
846f1da4 | 1985 | log_warning_errno(errno, "Failed to execute configured init %s, trying fallback: %m", args[0]); |
3c7878f9 LP |
1986 | } |
1987 | ||
1988 | args[0] = "/sbin/init"; | |
1989 | (void) execv(args[0], (char* const*) args); | |
1990 | r = -errno; | |
1991 | ||
1992 | manager_status_printf(NULL, STATUS_TYPE_EMERGENCY, | |
1993 | ANSI_HIGHLIGHT_RED " !! " ANSI_NORMAL, | |
1994 | "Failed to execute /sbin/init"); | |
1995 | ||
1e3eee8c | 1996 | *ret_error_message = "Failed to execute fallback shell"; |
3c7878f9 LP |
1997 | if (r == -ENOENT) { |
1998 | log_warning("No /sbin/init, trying fallback"); | |
1999 | ||
2000 | args[0] = "/bin/sh"; | |
2001 | args[1] = NULL; | |
a5cede8c | 2002 | (void) execve(args[0], (char* const*) args, saved_env); |
1e3eee8c | 2003 | return log_error_errno(errno, "Failed to execute /bin/sh, giving up: %m"); |
3c7878f9 | 2004 | } else |
6b1fa539 | 2005 | return log_error_errno(r, "Failed to execute /sbin/init, giving up: %m"); |
3c7878f9 LP |
2006 | } |
2007 | ||
7eb35049 LP |
2008 | static int invoke_main_loop( |
2009 | Manager *m, | |
a9fd4cd1 FB |
2010 | const struct rlimit *saved_rlimit_nofile, |
2011 | const struct rlimit *saved_rlimit_memlock, | |
7eb35049 | 2012 | int *ret_retval, /* Return parameters relevant for shutting down */ |
7eb35049 LP |
2013 | FDSet **ret_fds, /* Return parameters for reexecuting */ |
2014 | char **ret_switch_root_dir, /* … */ | |
2015 | char **ret_switch_root_init, /* … */ | |
2016 | const char **ret_error_message) { | |
2017 | ||
2018 | int r; | |
2019 | ||
2020 | assert(m); | |
a9fd4cd1 FB |
2021 | assert(saved_rlimit_nofile); |
2022 | assert(saved_rlimit_memlock); | |
7eb35049 | 2023 | assert(ret_retval); |
7eb35049 LP |
2024 | assert(ret_fds); |
2025 | assert(ret_switch_root_dir); | |
2026 | assert(ret_switch_root_init); | |
2027 | assert(ret_error_message); | |
2028 | ||
2029 | for (;;) { | |
5409c6fc ZJS |
2030 | int objective = manager_loop(m); |
2031 | if (objective < 0) { | |
7eb35049 | 2032 | *ret_error_message = "Failed to run main loop"; |
ad5db940 OJ |
2033 | return log_struct_errno(LOG_EMERG, objective, |
2034 | LOG_MESSAGE("Failed to run main loop: %m"), | |
2035 | "MESSAGE_ID=" SD_MESSAGE_CORE_MAINLOOP_FAILED_STR); | |
7eb35049 LP |
2036 | } |
2037 | ||
5409c6fc | 2038 | switch (objective) { |
7eb35049 | 2039 | |
a6ecbf83 | 2040 | case MANAGER_RELOAD: { |
bda7d78b | 2041 | LogTarget saved_log_target; |
a6ecbf83 FB |
2042 | int saved_log_level; |
2043 | ||
dd0ab174 LP |
2044 | manager_send_reloading(m); |
2045 | ||
af2fb2f2 | 2046 | log_info("Reloading..."); |
7eb35049 | 2047 | |
ff3a7019 ZJS |
2048 | /* First, save any overridden log level/target, then parse the configuration file, |
2049 | * which might change the log level to new settings. */ | |
bda7d78b | 2050 | |
a6ecbf83 | 2051 | saved_log_level = m->log_level_overridden ? log_get_max_level() : -1; |
bda7d78b | 2052 | saved_log_target = m->log_target_overridden ? log_get_target() : _LOG_TARGET_INVALID; |
a6ecbf83 | 2053 | |
a9fd4cd1 | 2054 | (void) parse_configuration(saved_rlimit_nofile, saved_rlimit_memlock); |
7eb35049 LP |
2055 | |
2056 | set_manager_defaults(m); | |
986935cf | 2057 | set_manager_settings(m); |
7eb35049 | 2058 | |
61fbbac1 | 2059 | update_cpu_affinity(false); |
b070c7c0 | 2060 | update_numa_policy(false); |
61fbbac1 | 2061 | |
a6ecbf83 FB |
2062 | if (saved_log_level >= 0) |
2063 | manager_override_log_level(m, saved_log_level); | |
bda7d78b FB |
2064 | if (saved_log_target >= 0) |
2065 | manager_override_log_target(m, saved_log_target); | |
a6ecbf83 | 2066 | |
5409c6fc | 2067 | if (manager_reload(m) < 0) |
ff3a7019 ZJS |
2068 | /* Reloading failed before the point of no return. |
2069 | * Let's continue running as if nothing happened. */ | |
7a35fa24 | 2070 | m->objective = MANAGER_OK; |
af2fb2f2 LB |
2071 | else |
2072 | log_info("Reloading finished in " USEC_FMT " ms.", | |
2073 | usec_sub_unsigned(now(CLOCK_MONOTONIC), m->timestamps[MANAGER_TIMESTAMP_UNITS_LOAD].monotonic) / USEC_PER_MSEC); | |
7eb35049 | 2074 | |
5409c6fc | 2075 | continue; |
a6ecbf83 | 2076 | } |
7eb35049 LP |
2077 | |
2078 | case MANAGER_REEXECUTE: | |
dd0ab174 LP |
2079 | |
2080 | manager_send_reloading(m); /* From the perspective of the manager calling us this is | |
2081 | * pretty much the same as a reload */ | |
2082 | ||
7eb35049 LP |
2083 | r = prepare_reexecute(m, &arg_serialization, ret_fds, false); |
2084 | if (r < 0) { | |
2085 | *ret_error_message = "Failed to prepare for reexecution"; | |
2086 | return r; | |
2087 | } | |
2088 | ||
2089 | log_notice("Reexecuting."); | |
2090 | ||
7eb35049 | 2091 | *ret_retval = EXIT_SUCCESS; |
7eb35049 LP |
2092 | *ret_switch_root_dir = *ret_switch_root_init = NULL; |
2093 | ||
5409c6fc | 2094 | return objective; |
7eb35049 LP |
2095 | |
2096 | case MANAGER_SWITCH_ROOT: | |
dd0ab174 LP |
2097 | |
2098 | manager_send_reloading(m); /* From the perspective of the manager calling us this is | |
2099 | * pretty much the same as a reload */ | |
2100 | ||
d35fe8c0 FB |
2101 | manager_set_switching_root(m, true); |
2102 | ||
7eb35049 LP |
2103 | if (!m->switch_root_init) { |
2104 | r = prepare_reexecute(m, &arg_serialization, ret_fds, true); | |
2105 | if (r < 0) { | |
2106 | *ret_error_message = "Failed to prepare for reexecution"; | |
2107 | return r; | |
2108 | } | |
2109 | } else | |
2110 | *ret_fds = NULL; | |
2111 | ||
2112 | log_notice("Switching root."); | |
2113 | ||
7eb35049 | 2114 | *ret_retval = EXIT_SUCCESS; |
7eb35049 LP |
2115 | |
2116 | /* Steal the switch root parameters */ | |
49052946 YW |
2117 | *ret_switch_root_dir = TAKE_PTR(m->switch_root); |
2118 | *ret_switch_root_init = TAKE_PTR(m->switch_root_init); | |
7eb35049 | 2119 | |
5409c6fc | 2120 | return objective; |
7eb35049 | 2121 | |
13ffc607 LP |
2122 | case MANAGER_SOFT_REBOOT: |
2123 | manager_send_reloading(m); | |
2124 | manager_set_switching_root(m, true); | |
2125 | ||
2126 | r = prepare_reexecute(m, &arg_serialization, ret_fds, /* switching_root= */ true); | |
2127 | if (r < 0) { | |
2128 | *ret_error_message = "Failed to prepare for reexecution"; | |
2129 | return r; | |
2130 | } | |
2131 | ||
2132 | log_notice("Soft-rebooting."); | |
2133 | ||
2134 | *ret_retval = EXIT_SUCCESS; | |
2135 | *ret_switch_root_dir = TAKE_PTR(m->switch_root); | |
2136 | *ret_switch_root_init = NULL; | |
2137 | ||
2138 | return objective; | |
2139 | ||
7eb35049 | 2140 | case MANAGER_EXIT: |
7eb35049 LP |
2141 | if (MANAGER_IS_USER(m)) { |
2142 | log_debug("Exit."); | |
2143 | ||
7eb35049 | 2144 | *ret_retval = m->return_value; |
7eb35049 LP |
2145 | *ret_fds = NULL; |
2146 | *ret_switch_root_dir = *ret_switch_root_init = NULL; | |
2147 | ||
5409c6fc | 2148 | return objective; |
7eb35049 LP |
2149 | } |
2150 | ||
2151 | _fallthrough_; | |
2152 | case MANAGER_REBOOT: | |
2153 | case MANAGER_POWEROFF: | |
2154 | case MANAGER_HALT: | |
2155 | case MANAGER_KEXEC: { | |
7eb35049 LP |
2156 | log_notice("Shutting down."); |
2157 | ||
7eb35049 | 2158 | *ret_retval = m->return_value; |
7eb35049 LP |
2159 | *ret_fds = NULL; |
2160 | *ret_switch_root_dir = *ret_switch_root_init = NULL; | |
2161 | ||
5409c6fc | 2162 | return objective; |
7eb35049 LP |
2163 | } |
2164 | ||
2165 | default: | |
04499a70 | 2166 | assert_not_reached(); |
7eb35049 LP |
2167 | } |
2168 | } | |
2169 | } | |
2170 | ||
31aef7ff | 2171 | static void log_execution_mode(bool *ret_first_boot) { |
7cd43e34 | 2172 | bool first_boot = false; |
4870133b | 2173 | int r; |
7cd43e34 | 2174 | |
31aef7ff LP |
2175 | assert(ret_first_boot); |
2176 | ||
4870133b LP |
2177 | switch (arg_runtime_scope) { |
2178 | ||
2179 | case RUNTIME_SCOPE_SYSTEM: { | |
40efaaed | 2180 | struct utsname uts; |
31aef7ff LP |
2181 | int v; |
2182 | ||
e7b18106 | 2183 | log_info("systemd " GIT_VERSION " running in %ssystem mode (%s)", |
91b79ba8 ZJS |
2184 | arg_action == ACTION_TEST ? "test " : "", |
2185 | systemd_features); | |
31aef7ff LP |
2186 | |
2187 | v = detect_virtualization(); | |
2188 | if (v > 0) | |
2189 | log_info("Detected virtualization %s.", virtualization_to_string(v)); | |
2190 | ||
024469dd DB |
2191 | v = detect_confidential_virtualization(); |
2192 | if (v > 0) | |
2193 | log_info("Detected confidential virtualization %s.", confidential_virtualization_to_string(v)); | |
2194 | ||
31aef7ff LP |
2195 | log_info("Detected architecture %s.", architecture_to_string(uname_architecture())); |
2196 | ||
7cd43e34 | 2197 | if (in_initrd()) |
55c041b4 | 2198 | log_info("Running in initrd."); |
7cd43e34 | 2199 | else { |
583cef3b HS |
2200 | _cleanup_free_ char *id_text = NULL; |
2201 | ||
7cd43e34 | 2202 | /* Let's check whether we are in first boot. First, check if an override was |
7c52d523 | 2203 | * specified on the kernel command line. If yes, we honour that. */ |
7cd43e34 | 2204 | |
3787934b | 2205 | r = proc_cmdline_get_bool("systemd.condition-first-boot", /* flags = */ 0, &first_boot); |
7cd43e34 | 2206 | if (r < 0) |
7c52d523 | 2207 | log_debug_errno(r, "Failed to parse systemd.condition-first-boot= kernel command line argument, ignoring: %m"); |
7cd43e34 ZJS |
2208 | |
2209 | if (r > 0) | |
2210 | log_full(first_boot ? LOG_INFO : LOG_DEBUG, | |
7c52d523 | 2211 | "Kernel command line argument says we are %s first boot.", |
7cd43e34 ZJS |
2212 | first_boot ? "in" : "not in"); |
2213 | else { | |
2214 | /* Second, perform autodetection. We use /etc/machine-id as flag file for | |
2215 | * this: If it is missing or contains the value "uninitialized", this is the | |
2216 | * first boot. In other cases, it is not. This allows container managers and | |
2217 | * installers to provision a couple of files in /etc but still permit the | |
2218 | * first-boot initialization to occur. If the container manager wants to | |
2219 | * provision the machine ID it should pass $container_uuid to PID 1. */ | |
2220 | ||
2221 | r = read_one_line_file("/etc/machine-id", &id_text); | |
2222 | if (r < 0 || streq(id_text, "uninitialized")) { | |
2223 | if (r < 0 && r != -ENOENT) | |
60e4b429 | 2224 | log_warning_errno(r, "Unexpected error while reading /etc/machine-id, assuming first boot: %m"); |
7cd43e34 ZJS |
2225 | |
2226 | first_boot = true; | |
2227 | log_info("Detected first boot."); | |
2228 | } else | |
2229 | log_debug("Detected initialized system, this is not the first boot."); | |
583cef3b | 2230 | } |
31aef7ff | 2231 | } |
40efaaed | 2232 | |
5180394b | 2233 | assert_se(uname(&uts) >= 0); |
40efaaed LP |
2234 | |
2235 | if (strverscmp_improved(uts.release, KERNEL_BASELINE_VERSION) < 0) | |
2236 | log_warning("Warning! Reported kernel version %s is older than systemd's required baseline kernel version %s. " | |
2237 | "Your mileage may vary.", uts.release, KERNEL_BASELINE_VERSION); | |
2238 | else | |
2239 | log_debug("Kernel version %s, our baseline is %s", uts.release, KERNEL_BASELINE_VERSION); | |
4870133b LP |
2240 | |
2241 | break; | |
2242 | } | |
2243 | ||
2244 | case RUNTIME_SCOPE_USER: | |
b9e90f3a | 2245 | if (DEBUG_LOGGING) { |
c2b2df60 | 2246 | _cleanup_free_ char *t = NULL; |
31aef7ff | 2247 | |
b9e90f3a | 2248 | t = uid_to_name(getuid()); |
91b79ba8 ZJS |
2249 | log_debug("systemd " GIT_VERSION " running in %suser mode for user " UID_FMT "/%s. (%s)", |
2250 | arg_action == ACTION_TEST ? " test" : "", | |
2251 | getuid(), strna(t), systemd_features); | |
b9e90f3a | 2252 | } |
4870133b LP |
2253 | |
2254 | break; | |
2255 | ||
2256 | default: | |
2257 | assert_not_reached(); | |
31aef7ff | 2258 | } |
7cd43e34 ZJS |
2259 | |
2260 | *ret_first_boot = first_boot; | |
31aef7ff LP |
2261 | } |
2262 | ||
5afbaa36 LP |
2263 | static int initialize_runtime( |
2264 | bool skip_setup, | |
3023f2fe | 2265 | bool first_boot, |
5afbaa36 LP |
2266 | struct rlimit *saved_rlimit_nofile, |
2267 | struct rlimit *saved_rlimit_memlock, | |
2268 | const char **ret_error_message) { | |
5afbaa36 LP |
2269 | int r; |
2270 | ||
2271 | assert(ret_error_message); | |
2272 | ||
2273 | /* Sets up various runtime parameters. Many of these initializations are conditionalized: | |
2274 | * | |
2275 | * - Some only apply to --system instances | |
2276 | * - Some only apply to --user instances | |
2277 | * - Some only apply when we first start up, but not when we reexecute | |
2278 | */ | |
2279 | ||
2d776038 LP |
2280 | if (arg_action != ACTION_RUN) |
2281 | return 0; | |
2282 | ||
61fbbac1 | 2283 | update_cpu_affinity(skip_setup); |
b070c7c0 | 2284 | update_numa_policy(skip_setup); |
61fbbac1 | 2285 | |
4870133b LP |
2286 | switch (arg_runtime_scope) { |
2287 | ||
2288 | case RUNTIME_SCOPE_SYSTEM: | |
5238e957 | 2289 | /* Make sure we leave a core dump without panicking the kernel. */ |
3c3c6cb9 | 2290 | install_crash_handler(); |
5afbaa36 | 2291 | |
3c3c6cb9 | 2292 | if (!skip_setup) { |
143fadf3 | 2293 | r = mount_cgroup_controllers(); |
3c3c6cb9 LP |
2294 | if (r < 0) { |
2295 | *ret_error_message = "Failed to mount cgroup hierarchies"; | |
2296 | return r; | |
2297 | } | |
2298 | ||
deb0d489 LP |
2299 | /* Pull credentials from various sources into a common credential directory (we do |
2300 | * this here, before setting up the machine ID, so that we can use credential info | |
2301 | * for setting up the machine ID) */ | |
2302 | (void) import_credentials(); | |
2303 | ||
4bd03515 | 2304 | (void) os_release_status(); |
b6fad306 | 2305 | (void) hostname_setup(true); |
3023f2fe | 2306 | /* Force transient machine-id on first boot. */ |
deb0d489 | 2307 | machine_id_setup(/* root= */ NULL, /* force_transient= */ first_boot, arg_machine_id, /* ret_machine_id */ NULL); |
df883de9 | 2308 | (void) loopback_setup(); |
3c3c6cb9 | 2309 | bump_unix_max_dgram_qlen(); |
a8b627aa | 2310 | bump_file_max_and_nr_open(); |
3c3c6cb9 LP |
2311 | test_usr(); |
2312 | write_container_id(); | |
3f37a825 LB |
2313 | |
2314 | /* Copy os-release to the propagate directory, so that we update it for services running | |
2315 | * under RootDirectory=/RootImage= when we do a soft reboot. */ | |
2316 | r = setup_os_release(RUNTIME_SCOPE_SYSTEM); | |
2317 | if (r < 0) | |
2318 | log_warning_errno(r, "Failed to copy os-release for propagation, ignoring: %m"); | |
3c3c6cb9 | 2319 | } |
8a2c1fbf | 2320 | |
82d7a151 YW |
2321 | r = watchdog_set_device(arg_watchdog_device); |
2322 | if (r < 0) | |
2323 | log_warning_errno(r, "Failed to set watchdog device to %s, ignoring: %m", arg_watchdog_device); | |
4870133b LP |
2324 | |
2325 | break; | |
2326 | ||
2327 | case RUNTIME_SCOPE_USER: { | |
32429805 LP |
2328 | _cleanup_free_ char *p = NULL; |
2329 | ||
2330 | /* Create the runtime directory and place the inaccessible device nodes there, if we run in | |
2331 | * user mode. In system mode mount_setup() already did that. */ | |
2332 | ||
2333 | r = xdg_user_runtime_dir(&p, "/systemd"); | |
2334 | if (r < 0) { | |
2335 | *ret_error_message = "$XDG_RUNTIME_DIR is not set"; | |
ad5db940 OJ |
2336 | return log_struct_errno(LOG_EMERG, r, |
2337 | LOG_MESSAGE("Failed to determine $XDG_RUNTIME_DIR path: %m"), | |
2338 | "MESSAGE_ID=" SD_MESSAGE_CORE_NO_XDGDIR_PATH_STR); | |
32429805 LP |
2339 | } |
2340 | ||
e813a74a | 2341 | (void) mkdir_p_label(p, 0755); |
32429805 | 2342 | (void) make_inaccessible_nodes(p, UID_INVALID, GID_INVALID); |
3f37a825 LB |
2343 | r = setup_os_release(RUNTIME_SCOPE_USER); |
2344 | if (r < 0) | |
2345 | log_warning_errno(r, "Failed to copy os-release for propagation, ignoring: %m"); | |
4870133b LP |
2346 | break; |
2347 | } | |
2348 | ||
2349 | default: | |
2350 | assert_not_reached(); | |
3c3c6cb9 | 2351 | } |
5afbaa36 LP |
2352 | |
2353 | if (arg_timer_slack_nsec != NSEC_INFINITY) | |
2354 | if (prctl(PR_SET_TIMERSLACK, arg_timer_slack_nsec) < 0) | |
3a671cd1 | 2355 | log_warning_errno(errno, "Failed to adjust timer slack, ignoring: %m"); |
5afbaa36 | 2356 | |
4870133b | 2357 | if (arg_runtime_scope == RUNTIME_SCOPE_SYSTEM) { |
5afbaa36 | 2358 | |
4870133b LP |
2359 | if (!cap_test_all(arg_capability_bounding_set)) { |
2360 | r = capability_bounding_set_drop_usermode(arg_capability_bounding_set); | |
2361 | if (r < 0) { | |
2362 | *ret_error_message = "Failed to drop capability bounding set of usermode helpers"; | |
ad5db940 OJ |
2363 | return log_struct_errno(LOG_EMERG, r, |
2364 | LOG_MESSAGE("Failed to drop capability bounding set of usermode helpers: %m"), | |
2365 | "MESSAGE_ID=" SD_MESSAGE_CORE_CAPABILITY_BOUNDING_USER_STR); | |
4870133b LP |
2366 | } |
2367 | ||
2368 | r = capability_bounding_set_drop(arg_capability_bounding_set, true); | |
2369 | if (r < 0) { | |
2370 | *ret_error_message = "Failed to drop capability bounding set"; | |
ad5db940 OJ |
2371 | return log_struct_errno(LOG_EMERG, r, |
2372 | LOG_MESSAGE("Failed to drop capability bounding set: %m"), | |
2373 | "MESSAGE_ID=" SD_MESSAGE_CORE_CAPABILITY_BOUNDING_STR); | |
4870133b | 2374 | } |
5afbaa36 | 2375 | } |
5afbaa36 | 2376 | |
4870133b LP |
2377 | if (arg_no_new_privs) { |
2378 | if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) < 0) { | |
2379 | *ret_error_message = "Failed to disable new privileges"; | |
ad5db940 OJ |
2380 | return log_struct_errno(LOG_EMERG, errno, |
2381 | LOG_MESSAGE("Failed to disable new privileges: %m"), | |
2382 | "MESSAGE_ID=" SD_MESSAGE_CORE_DISABLE_PRIVILEGES_STR); | |
4870133b | 2383 | } |
39362f6f JB |
2384 | } |
2385 | } | |
2386 | ||
5afbaa36 LP |
2387 | if (arg_syscall_archs) { |
2388 | r = enforce_syscall_archs(arg_syscall_archs); | |
2389 | if (r < 0) { | |
2390 | *ret_error_message = "Failed to set syscall architectures"; | |
2391 | return r; | |
2392 | } | |
2393 | } | |
2394 | ||
8c3fe1b5 LP |
2395 | r = make_reaper_process(true); |
2396 | if (r < 0) | |
2397 | log_warning_errno(r, "Failed to make us a subreaper, ignoring: %m"); | |
5afbaa36 | 2398 | |
a17c1712 LP |
2399 | /* Bump up RLIMIT_NOFILE for systemd itself */ |
2400 | (void) bump_rlimit_nofile(saved_rlimit_nofile); | |
2401 | (void) bump_rlimit_memlock(saved_rlimit_memlock); | |
5afbaa36 LP |
2402 | |
2403 | return 0; | |
2404 | } | |
2405 | ||
6acca5fc LP |
2406 | static int do_queue_default_job( |
2407 | Manager *m, | |
2408 | const char **ret_error_message) { | |
2409 | ||
2410 | _cleanup_(sd_bus_error_free) sd_bus_error error = SD_BUS_ERROR_NULL; | |
f1d075dc ZJS |
2411 | const char *unit; |
2412 | Job *job; | |
2413 | Unit *target; | |
6acca5fc LP |
2414 | int r; |
2415 | ||
8755dbad | 2416 | if (arg_default_unit) |
f1d075dc | 2417 | unit = arg_default_unit; |
8755dbad | 2418 | else if (in_initrd()) |
f1d075dc | 2419 | unit = SPECIAL_INITRD_TARGET; |
8755dbad | 2420 | else |
f1d075dc | 2421 | unit = SPECIAL_DEFAULT_TARGET; |
8755dbad | 2422 | |
f1d075dc | 2423 | log_debug("Activating default unit: %s", unit); |
8755dbad | 2424 | |
f1d075dc | 2425 | r = manager_load_startable_unit_or_warn(m, unit, NULL, &target); |
8755dbad ZJS |
2426 | if (r < 0 && in_initrd() && !arg_default_unit) { |
2427 | /* Fall back to default.target, which we used to always use by default. Only do this if no | |
2428 | * explicit configuration was given. */ | |
2429 | ||
2430 | log_info("Falling back to " SPECIAL_DEFAULT_TARGET "."); | |
6acca5fc | 2431 | |
8755dbad ZJS |
2432 | r = manager_load_startable_unit_or_warn(m, SPECIAL_DEFAULT_TARGET, NULL, &target); |
2433 | } | |
4109ede7 | 2434 | if (r < 0) { |
8755dbad | 2435 | log_info("Falling back to " SPECIAL_RESCUE_TARGET "."); |
6acca5fc | 2436 | |
4109ede7 | 2437 | r = manager_load_startable_unit_or_warn(m, SPECIAL_RESCUE_TARGET, NULL, &target); |
6acca5fc | 2438 | if (r < 0) { |
8755dbad ZJS |
2439 | *ret_error_message = r == -ERFKILL ? SPECIAL_RESCUE_TARGET " masked" |
2440 | : "Failed to load " SPECIAL_RESCUE_TARGET; | |
4109ede7 | 2441 | return r; |
6acca5fc LP |
2442 | } |
2443 | } | |
2444 | ||
2445 | assert(target->load_state == UNIT_LOADED); | |
2446 | ||
f1d075dc | 2447 | r = manager_add_job(m, JOB_START, target, JOB_ISOLATE, NULL, &error, &job); |
6acca5fc LP |
2448 | if (r == -EPERM) { |
2449 | log_debug_errno(r, "Default target could not be isolated, starting instead: %s", bus_error_message(&error, r)); | |
2450 | ||
2451 | sd_bus_error_free(&error); | |
2452 | ||
f1d075dc | 2453 | r = manager_add_job(m, JOB_START, target, JOB_REPLACE, NULL, &error, &job); |
6acca5fc LP |
2454 | if (r < 0) { |
2455 | *ret_error_message = "Failed to start default target"; | |
ad5db940 OJ |
2456 | return log_struct_errno(LOG_EMERG, r, |
2457 | LOG_MESSAGE("Failed to start default target: %s", bus_error_message(&error, r)), | |
2458 | "MESSAGE_ID=" SD_MESSAGE_CORE_START_TARGET_FAILED_STR); | |
6acca5fc LP |
2459 | } |
2460 | ||
2461 | } else if (r < 0) { | |
2462 | *ret_error_message = "Failed to isolate default target"; | |
ad5db940 OJ |
2463 | return log_struct_errno(LOG_EMERG, r, |
2464 | LOG_MESSAGE("Failed to isolate default target: %s", bus_error_message(&error, r)), | |
2465 | "MESSAGE_ID=" SD_MESSAGE_CORE_ISOLATE_TARGET_FAILED_STR); | |
c86c31d9 ZJS |
2466 | } else |
2467 | log_info("Queued %s job for default target %s.", | |
2468 | job_type_to_string(job->type), | |
04d232d8 | 2469 | unit_status_string(job->unit, NULL)); |
6acca5fc | 2470 | |
f1d075dc | 2471 | m->default_unit_job_id = job->id; |
6acca5fc LP |
2472 | |
2473 | return 0; | |
2474 | } | |
2475 | ||
a9fd4cd1 FB |
2476 | static void save_rlimits(struct rlimit *saved_rlimit_nofile, |
2477 | struct rlimit *saved_rlimit_memlock) { | |
2478 | ||
2479 | assert(saved_rlimit_nofile); | |
2480 | assert(saved_rlimit_memlock); | |
2481 | ||
2482 | if (getrlimit(RLIMIT_NOFILE, saved_rlimit_nofile) < 0) | |
2483 | log_warning_errno(errno, "Reading RLIMIT_NOFILE failed, ignoring: %m"); | |
2484 | ||
2485 | if (getrlimit(RLIMIT_MEMLOCK, saved_rlimit_memlock) < 0) | |
2486 | log_warning_errno(errno, "Reading RLIMIT_MEMLOCK failed, ignoring: %m"); | |
2487 | } | |
2488 | ||
2489 | static void fallback_rlimit_nofile(const struct rlimit *saved_rlimit_nofile) { | |
2490 | struct rlimit *rl; | |
2491 | ||
c9e120e0 | 2492 | if (arg_defaults.rlimit[RLIMIT_NOFILE]) |
a9fd4cd1 FB |
2493 | return; |
2494 | ||
2495 | /* Make sure forked processes get limits based on the original kernel setting */ | |
2496 | ||
2497 | rl = newdup(struct rlimit, saved_rlimit_nofile, 1); | |
2498 | if (!rl) { | |
2499 | log_oom(); | |
2500 | return; | |
2501 | } | |
2502 | ||
2503 | /* Bump the hard limit for system services to a substantially higher value. The default | |
2504 | * hard limit current kernels set is pretty low (4K), mostly for historical | |
2505 | * reasons. According to kernel developers, the fd handling in recent kernels has been | |
2506 | * optimized substantially enough, so that we can bump the limit now, without paying too | |
2507 | * high a price in memory or performance. Note however that we only bump the hard limit, | |
2508 | * not the soft limit. That's because select() works the way it works, and chokes on fds | |
2509 | * >= 1024. If we'd bump the soft limit globally, it might accidentally happen to | |
2510 | * unexpecting programs that they get fds higher than what they can process using | |
2511 | * select(). By only bumping the hard limit but leaving the low limit as it is we avoid | |
2512 | * this pitfall: programs that are written by folks aware of the select() problem in mind | |
2513 | * (and thus use poll()/epoll instead of select(), the way everybody should) can | |
2514 | * explicitly opt into high fds by bumping their soft limit beyond 1024, to the hard limit | |
2515 | * we pass. */ | |
4870133b | 2516 | if (arg_runtime_scope == RUNTIME_SCOPE_SYSTEM) { |
a9fd4cd1 FB |
2517 | int nr; |
2518 | ||
2519 | /* Get the underlying absolute limit the kernel enforces */ | |
2520 | nr = read_nr_open(); | |
2521 | ||
2522 | rl->rlim_max = MIN((rlim_t) nr, MAX(rl->rlim_max, (rlim_t) HIGH_RLIMIT_NOFILE)); | |
2523 | } | |
2524 | ||
2525 | /* If for some reason we were invoked with a soft limit above 1024 (which should never | |
2526 | * happen!, but who knows what we get passed in from pam_limit when invoked as --user | |
2527 | * instance), then lower what we pass on to not confuse our children */ | |
2528 | rl->rlim_cur = MIN(rl->rlim_cur, (rlim_t) FD_SETSIZE); | |
2529 | ||
c9e120e0 | 2530 | arg_defaults.rlimit[RLIMIT_NOFILE] = rl; |
a9fd4cd1 FB |
2531 | } |
2532 | ||
2533 | static void fallback_rlimit_memlock(const struct rlimit *saved_rlimit_memlock) { | |
2534 | struct rlimit *rl; | |
2535 | ||
2536 | /* Pass the original value down to invoked processes */ | |
2537 | ||
c9e120e0 | 2538 | if (arg_defaults.rlimit[RLIMIT_MEMLOCK]) |
a9fd4cd1 FB |
2539 | return; |
2540 | ||
2541 | rl = newdup(struct rlimit, saved_rlimit_memlock, 1); | |
2542 | if (!rl) { | |
2543 | log_oom(); | |
2544 | return; | |
2545 | } | |
2546 | ||
4870133b | 2547 | if (arg_runtime_scope == RUNTIME_SCOPE_SYSTEM) { |
852b6250 LP |
2548 | /* Raise the default limit to 8M also on old kernels and in containers (8M is the kernel |
2549 | * default for this since kernel 5.16) */ | |
2550 | rl->rlim_max = MAX(rl->rlim_max, (rlim_t) DEFAULT_RLIMIT_MEMLOCK); | |
2551 | rl->rlim_cur = MAX(rl->rlim_cur, (rlim_t) DEFAULT_RLIMIT_MEMLOCK); | |
2552 | } | |
2553 | ||
c9e120e0 | 2554 | arg_defaults.rlimit[RLIMIT_MEMLOCK] = rl; |
a9fd4cd1 FB |
2555 | } |
2556 | ||
d55ed7de | 2557 | static void setenv_manager_environment(void) { |
d55ed7de ZJS |
2558 | int r; |
2559 | ||
2560 | STRV_FOREACH(p, arg_manager_environment) { | |
2561 | log_debug("Setting '%s' in our own environment.", *p); | |
2562 | ||
2563 | r = putenv_dup(*p, true); | |
2564 | if (r < 0) | |
2565 | log_warning_errno(errno, "Failed to setenv \"%s\", ignoring: %m", *p); | |
2566 | } | |
2567 | } | |
2568 | ||
fb39af4c ZJS |
2569 | static void reset_arguments(void) { |
2570 | /* Frees/resets arg_* variables, with a few exceptions commented below. */ | |
970777b5 LP |
2571 | |
2572 | arg_default_unit = mfree(arg_default_unit); | |
fb39af4c | 2573 | |
4870133b | 2574 | /* arg_runtime_scope — ignore */ |
fb39af4c ZJS |
2575 | |
2576 | arg_dump_core = true; | |
2577 | arg_crash_chvt = -1; | |
2578 | arg_crash_shell = false; | |
2579 | arg_crash_reboot = false; | |
970777b5 | 2580 | arg_confirm_spawn = mfree(arg_confirm_spawn); |
fb39af4c | 2581 | arg_show_status = _SHOW_STATUS_INVALID; |
36cf4507 | 2582 | arg_status_unit_format = STATUS_UNIT_FORMAT_DEFAULT; |
fb39af4c ZJS |
2583 | arg_switched_root = false; |
2584 | arg_pager_flags = 0; | |
2585 | arg_service_watchdogs = true; | |
c9e120e0 LP |
2586 | |
2587 | unit_defaults_done(&arg_defaults); | |
ea09a416 | 2588 | unit_defaults_init(&arg_defaults, arg_runtime_scope); |
c9e120e0 | 2589 | |
fb39af4c | 2590 | arg_runtime_watchdog = 0; |
65224c1d | 2591 | arg_reboot_watchdog = 10 * USEC_PER_MINUTE; |
acafd7d8 | 2592 | arg_kexec_watchdog = 0; |
5717062e | 2593 | arg_pretimeout_watchdog = 0; |
919ea64f ŁS |
2594 | arg_early_core_pattern = mfree(arg_early_core_pattern); |
2595 | arg_watchdog_device = mfree(arg_watchdog_device); | |
aff3a9e1 | 2596 | arg_watchdog_pretimeout_governor = mfree(arg_watchdog_pretimeout_governor); |
fb39af4c | 2597 | |
970777b5 | 2598 | arg_default_environment = strv_free(arg_default_environment); |
d55ed7de | 2599 | arg_manager_environment = strv_free(arg_manager_environment); |
fb39af4c | 2600 | |
3fd5190b | 2601 | arg_capability_bounding_set = CAP_MASK_UNSET; |
fb39af4c | 2602 | arg_no_new_privs = false; |
ffc1ec73 | 2603 | arg_protect_system = -1; |
fb39af4c | 2604 | arg_timer_slack_nsec = NSEC_INFINITY; |
fb39af4c | 2605 | |
970777b5 | 2606 | arg_syscall_archs = set_free(arg_syscall_archs); |
61fbbac1 | 2607 | |
fb39af4c ZJS |
2608 | /* arg_serialization — ignore */ |
2609 | ||
fb39af4c ZJS |
2610 | arg_machine_id = (sd_id128_t) {}; |
2611 | arg_cad_burst_action = EMERGENCY_ACTION_REBOOT_FORCE; | |
fb39af4c | 2612 | |
61fbbac1 | 2613 | cpu_set_reset(&arg_cpu_affinity); |
b070c7c0 | 2614 | numa_policy_reset(&arg_numa_policy); |
d247f232 LP |
2615 | |
2616 | arg_random_seed = mfree(arg_random_seed); | |
2617 | arg_random_seed_size = 0; | |
33d943d1 | 2618 | arg_clock_usec = 0; |
d4a402e4 | 2619 | |
856bfaeb LB |
2620 | arg_reload_limit_interval_sec = 0; |
2621 | arg_reload_limit_burst = 0; | |
d4a402e4 LP |
2622 | } |
2623 | ||
2624 | static void determine_default_oom_score_adjust(void) { | |
2625 | int r, a, b; | |
2626 | ||
2627 | /* Run our services at slightly higher OOM score than ourselves. But let's be conservative here, and | |
2628 | * do this only if we don't run as root (i.e. only if we are run in user mode, for an unprivileged | |
2629 | * user). */ | |
2630 | ||
c9e120e0 | 2631 | if (arg_defaults.oom_score_adjust_set) |
d4a402e4 LP |
2632 | return; |
2633 | ||
2634 | if (getuid() == 0) | |
2635 | return; | |
2636 | ||
2637 | r = get_oom_score_adjust(&a); | |
2638 | if (r < 0) | |
2639 | return (void) log_warning_errno(r, "Failed to determine current OOM score adjustment value, ignoring: %m"); | |
2640 | ||
2641 | assert_cc(100 <= OOM_SCORE_ADJ_MAX); | |
2642 | b = a >= OOM_SCORE_ADJ_MAX - 100 ? OOM_SCORE_ADJ_MAX : a + 100; | |
2643 | ||
2644 | if (a == b) | |
2645 | return; | |
2646 | ||
c9e120e0 LP |
2647 | arg_defaults.oom_score_adjust = b; |
2648 | arg_defaults.oom_score_adjust_set = true; | |
970777b5 LP |
2649 | } |
2650 | ||
a9fd4cd1 FB |
2651 | static int parse_configuration(const struct rlimit *saved_rlimit_nofile, |
2652 | const struct rlimit *saved_rlimit_memlock) { | |
97d1fb94 LP |
2653 | int r; |
2654 | ||
a9fd4cd1 FB |
2655 | assert(saved_rlimit_nofile); |
2656 | assert(saved_rlimit_memlock); | |
2657 | ||
fb39af4c ZJS |
2658 | /* Assign configuration defaults */ |
2659 | reset_arguments(); | |
2660 | ||
97d1fb94 | 2661 | r = parse_config_file(); |
470a5e6d ZJS |
2662 | if (r < 0) |
2663 | log_warning_errno(r, "Failed to parse config file, ignoring: %m"); | |
97d1fb94 | 2664 | |
4870133b | 2665 | if (arg_runtime_scope == RUNTIME_SCOPE_SYSTEM) { |
97d1fb94 LP |
2666 | r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0); |
2667 | if (r < 0) | |
2668 | log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m"); | |
2669 | } | |
2670 | ||
a9fd4cd1 FB |
2671 | /* Initialize some default rlimits for services if they haven't been configured */ |
2672 | fallback_rlimit_nofile(saved_rlimit_nofile); | |
2673 | fallback_rlimit_memlock(saved_rlimit_memlock); | |
2674 | ||
97d1fb94 LP |
2675 | /* Note that this also parses bits from the kernel command line, including "debug". */ |
2676 | log_parse_environment(); | |
2677 | ||
db33214b | 2678 | /* Initialize the show status setting if it hasn't been set explicitly yet */ |
7a293242 | 2679 | if (arg_show_status == _SHOW_STATUS_INVALID) |
db33214b LP |
2680 | arg_show_status = SHOW_STATUS_YES; |
2681 | ||
d4a402e4 LP |
2682 | /* Slightly raise the OOM score for our services if we are running for unprivileged users. */ |
2683 | determine_default_oom_score_adjust(); | |
2684 | ||
d55ed7de ZJS |
2685 | /* Push variables into the manager environment block */ |
2686 | setenv_manager_environment(); | |
2687 | ||
a4303b40 DDM |
2688 | /* Parse log environment variables again to take into account any new environment variables. */ |
2689 | log_parse_environment(); | |
2690 | ||
97d1fb94 LP |
2691 | return 0; |
2692 | } | |
2693 | ||
b0d7c989 LP |
2694 | static int safety_checks(void) { |
2695 | ||
febf46a4 | 2696 | if (getpid_cached() == 1 && |
baaa35ad ZJS |
2697 | arg_action != ACTION_RUN) |
2698 | return log_error_errno(SYNTHETIC_ERRNO(EPERM), | |
2699 | "Unsupported execution mode while PID 1."); | |
febf46a4 LP |
2700 | |
2701 | if (getpid_cached() == 1 && | |
4870133b | 2702 | arg_runtime_scope == RUNTIME_SCOPE_USER) |
baaa35ad ZJS |
2703 | return log_error_errno(SYNTHETIC_ERRNO(EPERM), |
2704 | "Can't run --user mode as PID 1."); | |
febf46a4 LP |
2705 | |
2706 | if (arg_action == ACTION_RUN && | |
4870133b | 2707 | arg_runtime_scope == RUNTIME_SCOPE_SYSTEM && |
baaa35ad ZJS |
2708 | getpid_cached() != 1) |
2709 | return log_error_errno(SYNTHETIC_ERRNO(EPERM), | |
2710 | "Can't run system mode unless PID 1."); | |
febf46a4 | 2711 | |
b0d7c989 | 2712 | if (arg_action == ACTION_TEST && |
baaa35ad ZJS |
2713 | geteuid() == 0) |
2714 | return log_error_errno(SYNTHETIC_ERRNO(EPERM), | |
2715 | "Don't run test mode as root."); | |
b0d7c989 | 2716 | |
4870133b LP |
2717 | switch (arg_runtime_scope) { |
2718 | ||
2719 | case RUNTIME_SCOPE_USER: | |
2720 | ||
2721 | if (arg_action == ACTION_RUN && | |
2722 | sd_booted() <= 0) | |
2723 | return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP), | |
2724 | "Trying to run as user instance, but the system has not been booted with systemd."); | |
2725 | ||
2726 | if (arg_action == ACTION_RUN && | |
2727 | !getenv("XDG_RUNTIME_DIR")) | |
2728 | return log_error_errno(SYNTHETIC_ERRNO(EUNATCH), | |
2729 | "Trying to run as user instance, but $XDG_RUNTIME_DIR is not set."); | |
2730 | ||
2731 | break; | |
b0d7c989 | 2732 | |
4870133b LP |
2733 | case RUNTIME_SCOPE_SYSTEM: |
2734 | if (arg_action == ACTION_RUN && | |
2735 | running_in_chroot() > 0) | |
2736 | return log_error_errno(SYNTHETIC_ERRNO(EOPNOTSUPP), | |
2737 | "Cannot be run in a chroot() environment."); | |
2738 | break; | |
b0d7c989 | 2739 | |
4870133b LP |
2740 | default: |
2741 | assert_not_reached(); | |
2742 | } | |
b0d7c989 LP |
2743 | |
2744 | return 0; | |
2745 | } | |
2746 | ||
74da609f LP |
2747 | static int initialize_security( |
2748 | bool *loaded_policy, | |
2749 | dual_timestamp *security_start_timestamp, | |
2750 | dual_timestamp *security_finish_timestamp, | |
2751 | const char **ret_error_message) { | |
2752 | ||
2753 | int r; | |
2754 | ||
2755 | assert(loaded_policy); | |
2756 | assert(security_start_timestamp); | |
2757 | assert(security_finish_timestamp); | |
2758 | assert(ret_error_message); | |
2759 | ||
fa5a0251 | 2760 | dual_timestamp_now(security_start_timestamp); |
74da609f | 2761 | |
97149f40 | 2762 | r = mac_selinux_setup(loaded_policy); |
74da609f LP |
2763 | if (r < 0) { |
2764 | *ret_error_message = "Failed to load SELinux policy"; | |
2765 | return r; | |
2766 | } | |
2767 | ||
2768 | r = mac_smack_setup(loaded_policy); | |
2769 | if (r < 0) { | |
2770 | *ret_error_message = "Failed to load SMACK policy"; | |
2771 | return r; | |
2772 | } | |
2773 | ||
2ffadd3c Y |
2774 | r = mac_apparmor_setup(); |
2775 | if (r < 0) { | |
2776 | *ret_error_message = "Failed to load AppArmor policy"; | |
2777 | return r; | |
2778 | } | |
2779 | ||
74da609f LP |
2780 | r = ima_setup(); |
2781 | if (r < 0) { | |
2782 | *ret_error_message = "Failed to load IMA policy"; | |
2783 | return r; | |
2784 | } | |
2785 | ||
fa5a0251 | 2786 | dual_timestamp_now(security_finish_timestamp); |
74da609f LP |
2787 | return 0; |
2788 | } | |
2789 | ||
efeb853f LP |
2790 | static int collect_fds(FDSet **ret_fds, const char **ret_error_message) { |
2791 | int r; | |
2792 | ||
2793 | assert(ret_fds); | |
2794 | assert(ret_error_message); | |
2795 | ||
a3dff21a LP |
2796 | /* Pick up all fds passed to us. We apply a filter here: we only take the fds that have O_CLOEXEC |
2797 | * off. All fds passed via execve() to us must have O_CLOEXEC off, and our own code and dependencies | |
2798 | * should be clean enough to set O_CLOEXEC universally. Thus checking the bit should be a safe | |
2799 | * mechanism to distinguish passed in fds from our own. | |
2800 | * | |
2801 | * Why bother? Some subsystems we initialize early, specifically selinux might keep fds open in our | |
2802 | * process behind our back. We should not take possession of that (and then accidentally close | |
2803 | * it). SELinux thankfully sets O_CLOEXEC on its fds, so this test should work. */ | |
2804 | r = fdset_new_fill(/* filter_cloexec= */ 0, ret_fds); | |
efeb853f LP |
2805 | if (r < 0) { |
2806 | *ret_error_message = "Failed to allocate fd set"; | |
ad5db940 OJ |
2807 | return log_struct_errno(LOG_EMERG, r, |
2808 | LOG_MESSAGE("Failed to allocate fd set: %m"), | |
2809 | "MESSAGE_ID=" SD_MESSAGE_CORE_FD_SET_FAILED_STR); | |
efeb853f LP |
2810 | } |
2811 | ||
a3dff21a LP |
2812 | /* The serialization fd should have O_CLOEXEC turned on already, let's verify that we didn't pick it up here */ |
2813 | assert_se(!arg_serialization || !fdset_contains(*ret_fds, fileno(arg_serialization))); | |
efeb853f LP |
2814 | |
2815 | return 0; | |
2816 | } | |
2817 | ||
2e51b31c LP |
2818 | static void setup_console_terminal(bool skip_setup) { |
2819 | ||
4870133b | 2820 | if (arg_runtime_scope != RUNTIME_SCOPE_SYSTEM) |
2e51b31c LP |
2821 | return; |
2822 | ||
2823 | /* Become a session leader if we aren't one yet. */ | |
2824 | (void) setsid(); | |
2825 | ||
ff3a7019 ZJS |
2826 | /* If we are init, we connect stdin/stdout/stderr to /dev/null and make sure we don't have a |
2827 | * controlling tty. */ | |
2e51b31c LP |
2828 | (void) release_terminal(); |
2829 | ||
2830 | /* Reset the console, but only if this is really init and we are freshly booted */ | |
2831 | if (getpid_cached() == 1 && !skip_setup) | |
2832 | (void) console_setup(); | |
2833 | } | |
2834 | ||
aa40ff07 LP |
2835 | static bool early_skip_setup_check(int argc, char *argv[]) { |
2836 | bool found_deserialize = false; | |
aa40ff07 | 2837 | |
ff3a7019 ZJS |
2838 | /* Determine if this is a reexecution or normal bootup. We do the full command line parsing much |
2839 | * later, so let's just have a quick peek here. Note that if we have switched root, do all the | |
2840 | * special setup things anyway, even if in that case we also do deserialization. */ | |
aa40ff07 | 2841 | |
431733b8 | 2842 | for (int i = 1; i < argc; i++) |
aa40ff07 LP |
2843 | if (streq(argv[i], "--switched-root")) |
2844 | return false; /* If we switched root, don't skip the setup. */ | |
09567df7 | 2845 | else if (startswith(argv[i], "--deserialize=") || streq(argv[i], "--deserialize")) |
aa40ff07 | 2846 | found_deserialize = true; |
aa40ff07 LP |
2847 | |
2848 | return found_deserialize; /* When we are deserializing, then we are reexecuting, hence avoid the extensive setup */ | |
2849 | } | |
2850 | ||
0e06a031 LP |
2851 | static int save_env(void) { |
2852 | char **l; | |
2853 | ||
2854 | l = strv_copy(environ); | |
2855 | if (!l) | |
2856 | return -ENOMEM; | |
2857 | ||
2858 | strv_free_and_replace(saved_env, l); | |
2859 | return 0; | |
2860 | } | |
2861 | ||
60918275 | 2862 | int main(int argc, char *argv[]) { |
5409c6fc ZJS |
2863 | dual_timestamp |
2864 | initrd_timestamp = DUAL_TIMESTAMP_NULL, | |
2865 | userspace_timestamp = DUAL_TIMESTAMP_NULL, | |
2866 | kernel_timestamp = DUAL_TIMESTAMP_NULL, | |
2867 | security_start_timestamp = DUAL_TIMESTAMP_NULL, | |
2868 | security_finish_timestamp = DUAL_TIMESTAMP_NULL; | |
ddfa8b0b LP |
2869 | struct rlimit saved_rlimit_nofile = RLIMIT_MAKE_CONST(0), |
2870 | saved_rlimit_memlock = RLIMIT_MAKE_CONST(RLIM_INFINITY); /* The original rlimits we passed | |
2871 | * in. Note we use different values | |
2872 | * for the two that indicate whether | |
2873 | * these fields are initialized! */ | |
5409c6fc | 2874 | bool skip_setup, loaded_policy = false, queue_default_job = false, first_boot = false; |
625e8690 | 2875 | char *switch_root_dir = NULL, *switch_root_init = NULL; |
9d76d730 | 2876 | usec_t before_startup, after_startup; |
625e8690 | 2877 | static char systemd[] = "systemd"; |
b22d392d | 2878 | const char *error_message = NULL; |
625e8690 LP |
2879 | int r, retval = EXIT_FAILURE; |
2880 | Manager *m = NULL; | |
a16e1123 | 2881 | FDSet *fds = NULL; |
27b14a22 | 2882 | |
61b9769b | 2883 | assert_se(argc > 0 && !isempty(argv[0])); |
cf3095ac | 2884 | |
d72a8f10 | 2885 | /* SysV compatibility: redirect init → telinit */ |
6808a0bc | 2886 | redirect_telinit(argc, argv); |
2cb1a60d | 2887 | |
d72a8f10 | 2888 | /* Take timestamps early on */ |
c3a170f3 | 2889 | dual_timestamp_from_monotonic(&kernel_timestamp, 0); |
fa5a0251 | 2890 | dual_timestamp_now(&userspace_timestamp); |
c3a170f3 | 2891 | |
d72a8f10 | 2892 | /* Figure out whether we need to do initialize the system, or if we already did that because we are |
ff3a7019 | 2893 | * reexecuting. */ |
aa40ff07 | 2894 | skip_setup = early_skip_setup_check(argc, argv); |
d03bc1b8 | 2895 | |
ff3a7019 ZJS |
2896 | /* If we get started via the /sbin/init symlink then we are called 'init'. After a subsequent |
2897 | * reexecution we are then called 'systemd'. That is confusing, hence let's call us systemd | |
2898 | * right-away. */ | |
f3b6a3ed | 2899 | program_invocation_short_name = systemd; |
eee8b7ab | 2900 | (void) prctl(PR_SET_NAME, systemd); |
5d6b1584 | 2901 | |
d72a8f10 | 2902 | /* Save the original command line */ |
36fea155 | 2903 | save_argc_argv(argc, argv); |
f3b6a3ed | 2904 | |
0e06a031 LP |
2905 | /* Save the original environment as we might need to restore it if we're requested to execute another |
2906 | * system manager later. */ | |
2907 | r = save_env(); | |
2908 | if (r < 0) { | |
2909 | error_message = "Failed to copy environment block"; | |
2910 | goto finish; | |
2911 | } | |
a5cede8c | 2912 | |
6fdb8de4 | 2913 | /* Make sure that if the user says "syslog" we actually log to the journal. */ |
c1dc6153 | 2914 | log_set_upgrade_syslog_to_journal(true); |
bbe63281 | 2915 | |
df0ff127 | 2916 | if (getpid_cached() == 1) { |
b5752d23 | 2917 | /* When we run as PID 1 force system mode */ |
4870133b | 2918 | arg_runtime_scope = RUNTIME_SCOPE_SYSTEM; |
b5752d23 | 2919 | |
48a601fe | 2920 | /* Disable the umask logic */ |
90dc8c2e MG |
2921 | umask(0); |
2922 | ||
ff3a7019 ZJS |
2923 | /* Make sure that at least initially we do not ever log to journald/syslogd, because it might |
2924 | * not be activated yet (even though the log socket for it exists). */ | |
d075092f LP |
2925 | log_set_prohibit_ipc(true); |
2926 | ||
ff3a7019 ZJS |
2927 | /* Always reopen /dev/console when running as PID 1 or one of its pre-execve() children. This |
2928 | * is important so that we never end up logging to any foreign stderr, for example if we have | |
2929 | * to log in a child process right before execve()'ing the actual binary, at a point in time | |
2930 | * where socket activation stderr/stdout area already set up. */ | |
48a601fe | 2931 | log_set_always_reopen_console(true); |
48a601fe | 2932 | |
92890452 | 2933 | if (detect_container() <= 0) { |
4f8d551f | 2934 | |
92890452 | 2935 | /* Running outside of a container as PID 1 */ |
1e344c1d | 2936 | log_set_target_and_open(LOG_TARGET_KMSG); |
a866073d | 2937 | |
92890452 LP |
2938 | if (in_initrd()) |
2939 | initrd_timestamp = userspace_timestamp; | |
c3ba6250 | 2940 | |
92890452 LP |
2941 | if (!skip_setup) { |
2942 | r = mount_setup_early(); | |
2943 | if (r < 0) { | |
2944 | error_message = "Failed to mount early API filesystems"; | |
2945 | goto finish; | |
2946 | } | |
d2f57745 DDM |
2947 | } |
2948 | ||
2949 | /* We might have just mounted /proc, so let's try to parse the kernel | |
2950 | * command line log arguments immediately. */ | |
2951 | log_parse_environment(); | |
92890452 | 2952 | |
d2f57745 DDM |
2953 | /* Let's open the log backend a second time, in case the first time didn't |
2954 | * work. Quite possibly we have mounted /dev just now, so /dev/kmsg became | |
2955 | * available, and it previously wasn't. */ | |
2956 | log_open(); | |
0a2eef1e | 2957 | |
d2f57745 | 2958 | if (!skip_setup) { |
6123dfaa ZJS |
2959 | disable_printk_ratelimit(); |
2960 | ||
92890452 LP |
2961 | r = initialize_security( |
2962 | &loaded_policy, | |
2963 | &security_start_timestamp, | |
2964 | &security_finish_timestamp, | |
2965 | &error_message); | |
2966 | if (r < 0) | |
2967 | goto finish; | |
d723cd65 | 2968 | } |
eee8b7ab | 2969 | |
a452c807 DDM |
2970 | if (mac_init() < 0) { |
2971 | error_message = "Failed to initialize MAC support"; | |
96694e99 | 2972 | goto finish; |
92890452 | 2973 | } |
0b3325e7 | 2974 | |
92890452 LP |
2975 | if (!skip_setup) |
2976 | initialize_clock(); | |
2977 | ||
ff3a7019 ZJS |
2978 | /* Set the default for later on, but don't actually open the logs like this for |
2979 | * now. Note that if we are transitioning from the initrd there might still be | |
2980 | * journal fd open, and we shouldn't attempt opening that before we parsed | |
2981 | * /proc/cmdline which might redirect output elsewhere. */ | |
92890452 LP |
2982 | log_set_target(LOG_TARGET_JOURNAL_OR_KMSG); |
2983 | ||
2984 | } else { | |
2985 | /* Running inside a container, as PID 1 */ | |
1e344c1d | 2986 | log_set_target_and_open(LOG_TARGET_CONSOLE); |
92890452 LP |
2987 | |
2988 | /* For later on, see above... */ | |
2989 | log_set_target(LOG_TARGET_JOURNAL); | |
2990 | ||
45250e66 | 2991 | /* clear the kernel timestamp, because we are in a container */ |
92890452 | 2992 | kernel_timestamp = DUAL_TIMESTAMP_NULL; |
cb6531be | 2993 | } |
7948c4df | 2994 | |
92890452 | 2995 | initialize_coredump(skip_setup); |
a866073d | 2996 | |
92890452 LP |
2997 | r = fixup_environment(); |
2998 | if (r < 0) { | |
ad5db940 OJ |
2999 | log_struct_errno(LOG_EMERG, r, |
3000 | LOG_MESSAGE("Failed to fix up PID 1 environment: %m"), | |
3001 | "MESSAGE_ID=" SD_MESSAGE_CORE_PID1_ENVIRONMENT_STR); | |
92890452 LP |
3002 | error_message = "Failed to fix up PID1 environment"; |
3003 | goto finish; | |
3004 | } | |
a866073d | 3005 | |
ff3a7019 ZJS |
3006 | /* Try to figure out if we can use colors with the console. No need to do that for user |
3007 | * instances since they never log into the console. */ | |
3a18b604 | 3008 | log_show_color(colors_enabled()); |
92890452 | 3009 | |
c76cf844 AK |
3010 | r = make_null_stdio(); |
3011 | if (r < 0) | |
92890452 | 3012 | log_warning_errno(r, "Failed to redirect standard streams to /dev/null, ignoring: %m"); |
f84f9974 | 3013 | |
a132bef0 | 3014 | /* Load the kernel modules early. */ |
2e75e2a8 | 3015 | if (!skip_setup) |
e921a00d | 3016 | (void) kmod_setup(); |
2e75e2a8 | 3017 | |
3196e423 | 3018 | /* Mount /proc, /sys and friends, so that /proc/cmdline and /proc/$PID/fd is available. */ |
f74349d8 | 3019 | r = mount_setup(loaded_policy, skip_setup); |
cb6531be ZJS |
3020 | if (r < 0) { |
3021 | error_message = "Failed to mount API filesystems"; | |
8efe3c01 | 3022 | goto finish; |
cb6531be | 3023 | } |
c18ecf03 | 3024 | |
0be72218 JD |
3025 | /* The efivarfs is now mounted, let's lock down the system token. */ |
3026 | lock_down_efi_variables(); | |
209b2592 FB |
3027 | |
3028 | /* Cache command-line options passed from EFI variables */ | |
3029 | if (!skip_setup) | |
3030 | (void) cache_efi_options_variable(); | |
3196e423 LP |
3031 | } else { |
3032 | /* Running as user instance */ | |
4870133b | 3033 | arg_runtime_scope = RUNTIME_SCOPE_USER; |
2a646b1d | 3034 | log_set_always_reopen_console(true); |
1e344c1d | 3035 | log_set_target_and_open(LOG_TARGET_AUTO); |
3196e423 LP |
3036 | |
3037 | /* clear the kernel timestamp, because we are not PID 1 */ | |
3038 | kernel_timestamp = DUAL_TIMESTAMP_NULL; | |
3039 | ||
963b6b90 ŁS |
3040 | /* Clear ambient capabilities, so services do not inherit them implicitly. Dropping them does |
3041 | * not affect the permitted and effective sets which are important for the manager itself to | |
3042 | * operate. */ | |
3043 | capability_ambient_set_apply(0, /* also_inherit= */ false); | |
3044 | ||
a452c807 DDM |
3045 | if (mac_init() < 0) { |
3046 | error_message = "Failed to initialize MAC support"; | |
3196e423 LP |
3047 | goto finish; |
3048 | } | |
0c85a4f3 | 3049 | } |
4ade7963 | 3050 | |
a9fd4cd1 FB |
3051 | /* Save the original RLIMIT_NOFILE/RLIMIT_MEMLOCK so that we can reset it later when |
3052 | * transitioning from the initrd to the main systemd or suchlike. */ | |
3053 | save_rlimits(&saved_rlimit_nofile, &saved_rlimit_memlock); | |
3054 | ||
4ade7963 | 3055 | /* Reset all signal handlers. */ |
ce30c8dc | 3056 | (void) reset_all_signal_handlers(); |
9c274488 | 3057 | (void) ignore_signals(SIGNALS_IGNORE); |
078e4539 | 3058 | |
ffe5c01e FB |
3059 | (void) parse_configuration(&saved_rlimit_nofile, &saved_rlimit_memlock); |
3060 | ||
3061 | r = parse_argv(argc, argv); | |
3062 | if (r < 0) { | |
7c52d523 | 3063 | error_message = "Failed to parse command line arguments"; |
f170852a | 3064 | goto finish; |
ffe5c01e | 3065 | } |
10c961b9 | 3066 | |
b0d7c989 LP |
3067 | r = safety_checks(); |
3068 | if (r < 0) | |
fe783b03 | 3069 | goto finish; |
fe783b03 | 3070 | |
5c08257b | 3071 | if (IN_SET(arg_action, ACTION_TEST, ACTION_HELP, ACTION_DUMP_CONFIGURATION_ITEMS, ACTION_DUMP_BUS_PROPERTIES, ACTION_BUS_INTROSPECT)) |
384c2c32 | 3072 | pager_open(arg_pager_flags); |
b0d7c989 LP |
3073 | |
3074 | if (arg_action != ACTION_RUN) | |
74e7579c | 3075 | skip_setup = true; |
b87c2aa6 | 3076 | |
fa0f4d8a | 3077 | if (arg_action == ACTION_HELP) { |
37ec0fdd | 3078 | retval = help() < 0 ? EXIT_FAILURE : EXIT_SUCCESS; |
f170852a | 3079 | goto finish; |
9ba0bc4e ZJS |
3080 | } else if (arg_action == ACTION_VERSION) { |
3081 | retval = version(); | |
3082 | goto finish; | |
fa0f4d8a | 3083 | } else if (arg_action == ACTION_DUMP_CONFIGURATION_ITEMS) { |
e537352b | 3084 | unit_dump_config_items(stdout); |
22f4096c | 3085 | retval = EXIT_SUCCESS; |
e537352b | 3086 | goto finish; |
bbc1acab YW |
3087 | } else if (arg_action == ACTION_DUMP_BUS_PROPERTIES) { |
3088 | dump_bus_properties(stdout); | |
3089 | retval = EXIT_SUCCESS; | |
3090 | goto finish; | |
5c08257b ZJS |
3091 | } else if (arg_action == ACTION_BUS_INTROSPECT) { |
3092 | r = bus_manager_introspect_implementations(stdout, arg_bus_introspect); | |
3093 | retval = r >= 0 ? EXIT_SUCCESS : EXIT_FAILURE; | |
3094 | goto finish; | |
f170852a LP |
3095 | } |
3096 | ||
4c701096 | 3097 | assert_se(IN_SET(arg_action, ACTION_RUN, ACTION_TEST)); |
f170852a | 3098 | |
5a2e0c62 LP |
3099 | /* Move out of the way, so that we won't block unmounts */ |
3100 | assert_se(chdir("/") == 0); | |
3101 | ||
dea374e8 | 3102 | if (arg_action == ACTION_RUN) { |
d247f232 LP |
3103 | if (!skip_setup) { |
3104 | /* Apply the systemd.clock_usec= kernel command line switch */ | |
45250e66 | 3105 | apply_clock_update(); |
a70c72a0 | 3106 | |
d247f232 LP |
3107 | /* Apply random seed from kernel command line */ |
3108 | cmdline_take_random_seed(); | |
3109 | } | |
3110 | ||
ffc1ec73 | 3111 | /* A core pattern might have been specified via the cmdline. */ |
c6885f5f FB |
3112 | initialize_core_pattern(skip_setup); |
3113 | ||
ffc1ec73 LP |
3114 | /* Make /usr/ read-only */ |
3115 | apply_protect_system(skip_setup); | |
3116 | ||
efeb853f | 3117 | /* Close logging fds, in order not to confuse collecting passed fds and terminal logic below */ |
a70c72a0 LP |
3118 | log_close(); |
3119 | ||
3120 | /* Remember open file descriptors for later deserialization */ | |
efeb853f LP |
3121 | r = collect_fds(&fds, &error_message); |
3122 | if (r < 0) | |
dea374e8 | 3123 | goto finish; |
a16e1123 | 3124 | |
2e51b31c LP |
3125 | /* Give up any control of the console, but make sure its initialized. */ |
3126 | setup_console_terminal(skip_setup); | |
56d96fc0 | 3127 | |
a70c72a0 LP |
3128 | /* Open the logging devices, if possible and necessary */ |
3129 | log_open(); | |
56d96fc0 | 3130 | } |
4ade7963 | 3131 | |
31aef7ff | 3132 | log_execution_mode(&first_boot); |
a5dab5ce | 3133 | |
2d776038 | 3134 | r = initialize_runtime(skip_setup, |
3023f2fe | 3135 | first_boot, |
2d776038 LP |
3136 | &saved_rlimit_nofile, |
3137 | &saved_rlimit_memlock, | |
3138 | &error_message); | |
3139 | if (r < 0) | |
3140 | goto finish; | |
4096d6f5 | 3141 | |
4870133b | 3142 | r = manager_new(arg_runtime_scope, |
e0a3da1f ZJS |
3143 | arg_action == ACTION_TEST ? MANAGER_TEST_FULL : 0, |
3144 | &m); | |
e96d6be7 | 3145 | if (r < 0) { |
ad5db940 OJ |
3146 | log_struct_errno(LOG_EMERG, r, |
3147 | LOG_MESSAGE("Failed to allocate manager object: %m"), | |
3148 | "MESSAGE_ID=" SD_MESSAGE_CORE_MANAGER_ALLOCATE_STR); | |
cb6531be | 3149 | error_message = "Failed to allocate manager object"; |
60918275 LP |
3150 | goto finish; |
3151 | } | |
3152 | ||
9f9f0342 LP |
3153 | m->timestamps[MANAGER_TIMESTAMP_KERNEL] = kernel_timestamp; |
3154 | m->timestamps[MANAGER_TIMESTAMP_INITRD] = initrd_timestamp; | |
3155 | m->timestamps[MANAGER_TIMESTAMP_USERSPACE] = userspace_timestamp; | |
d4ee7bd8 YW |
3156 | m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_START)] = security_start_timestamp; |
3157 | m->timestamps[manager_timestamp_initrd_mangle(MANAGER_TIMESTAMP_SECURITY_FINISH)] = security_finish_timestamp; | |
9e58ff9c | 3158 | |
85cb4151 | 3159 | set_manager_defaults(m); |
7b46fc6a | 3160 | set_manager_settings(m); |
fd130612 | 3161 | manager_set_first_boot(m, first_boot); |
d35fe8c0 | 3162 | manager_set_switching_root(m, arg_switched_root); |
27d340c7 | 3163 | |
bf4df7c3 | 3164 | /* Remember whether we should queue the default job */ |
d3b1c508 | 3165 | queue_default_job = !arg_serialization || arg_switched_root; |
bf4df7c3 | 3166 | |
9d76d730 LP |
3167 | before_startup = now(CLOCK_MONOTONIC); |
3168 | ||
2a7cf953 | 3169 | r = manager_startup(m, arg_serialization, fds, /* root= */ NULL); |
58f88d92 | 3170 | if (r < 0) { |
cefb3eda | 3171 | error_message = "Failed to start up manager"; |
58f88d92 ZJS |
3172 | goto finish; |
3173 | } | |
a16e1123 | 3174 | |
6acca5fc | 3175 | /* This will close all file descriptors that were opened, but not claimed by any unit. */ |
2feceb5e | 3176 | fds = fdset_free(fds); |
74ca738f | 3177 | arg_serialization = safe_fclose(arg_serialization); |
bf4df7c3 LP |
3178 | |
3179 | if (queue_default_job) { | |
6acca5fc | 3180 | r = do_queue_default_job(m, &error_message); |
718db961 | 3181 | if (r < 0) |
37d88da7 | 3182 | goto finish; |
6acca5fc | 3183 | } |
ab17a050 | 3184 | |
6acca5fc | 3185 | after_startup = now(CLOCK_MONOTONIC); |
60918275 | 3186 | |
6acca5fc LP |
3187 | log_full(arg_action == ACTION_TEST ? LOG_INFO : LOG_DEBUG, |
3188 | "Loaded units and determined initial transaction in %s.", | |
5291f26d | 3189 | FORMAT_TIMESPAN(after_startup - before_startup, 100 * USEC_PER_MSEC)); |
07672f49 | 3190 | |
6acca5fc | 3191 | if (arg_action == ACTION_TEST) { |
2a341bb9 | 3192 | manager_test_summary(m); |
6acca5fc LP |
3193 | retval = EXIT_SUCCESS; |
3194 | goto finish; | |
e965d56d | 3195 | } |
d46de8a1 | 3196 | |
5409c6fc ZJS |
3197 | r = invoke_main_loop(m, |
3198 | &saved_rlimit_nofile, | |
3199 | &saved_rlimit_memlock, | |
3200 | &retval, | |
5409c6fc ZJS |
3201 | &fds, |
3202 | &switch_root_dir, | |
3203 | &switch_root_init, | |
3204 | &error_message); | |
3205 | assert(r < 0 || IN_SET(r, MANAGER_EXIT, /* MANAGER_OK is not expected here. */ | |
3206 | MANAGER_RELOAD, | |
3207 | MANAGER_REEXECUTE, | |
3208 | MANAGER_REBOOT, | |
13ffc607 | 3209 | MANAGER_SOFT_REBOOT, |
5409c6fc ZJS |
3210 | MANAGER_POWEROFF, |
3211 | MANAGER_HALT, | |
3212 | MANAGER_KEXEC, | |
3213 | MANAGER_SWITCH_ROOT)); | |
f170852a | 3214 | |
60918275 | 3215 | finish: |
b87c2aa6 ZJS |
3216 | pager_close(); |
3217 | ||
92890452 | 3218 | if (m) { |
986935cf FB |
3219 | arg_reboot_watchdog = manager_get_watchdog(m, WATCHDOG_REBOOT); |
3220 | arg_kexec_watchdog = manager_get_watchdog(m, WATCHDOG_KEXEC); | |
92890452 LP |
3221 | m = manager_free(m); |
3222 | } | |
60918275 | 3223 | |
cc56fafe | 3224 | mac_selinux_finish(); |
b2bb3dbe | 3225 | |
13ffc607 | 3226 | if (IN_SET(r, MANAGER_REEXECUTE, MANAGER_SWITCH_ROOT, MANAGER_SOFT_REBOOT)) |
1e3eee8c ZJS |
3227 | r = do_reexecute(r, |
3228 | argc, argv, | |
3229 | &saved_rlimit_nofile, | |
3230 | &saved_rlimit_memlock, | |
3231 | fds, | |
3232 | switch_root_dir, | |
3233 | switch_root_init, | |
3234 | &error_message); /* This only returns if reexecution failed */ | |
a16e1123 | 3235 | |
74ca738f | 3236 | arg_serialization = safe_fclose(arg_serialization); |
2feceb5e | 3237 | fds = fdset_free(fds); |
a16e1123 | 3238 | |
0e06a031 LP |
3239 | saved_env = strv_free(saved_env); |
3240 | ||
349cc4a5 | 3241 | #if HAVE_VALGRIND_VALGRIND_H |
54b434b1 LP |
3242 | /* If we are PID 1 and running under valgrind, then let's exit |
3243 | * here explicitly. valgrind will only generate nice output on | |
3244 | * exit(), not on exec(), hence let's do the former not the | |
3245 | * latter here. */ | |
8a2c1fbf EJ |
3246 | if (getpid_cached() == 1 && RUNNING_ON_VALGRIND) { |
3247 | /* Cleanup watchdog_device strings for valgrind. We need them | |
3248 | * in become_shutdown() so normally we cannot free them yet. */ | |
3249 | watchdog_free_device(); | |
7d9eea2b | 3250 | reset_arguments(); |
27fe58b7 | 3251 | return retval; |
8a2c1fbf | 3252 | } |
54b434b1 LP |
3253 | #endif |
3254 | ||
7e11a95e | 3255 | #if HAS_FEATURE_ADDRESS_SANITIZER |
ae5ce7e3 FS |
3256 | /* At this stage we most likely don't have stdio/stderr open, so the following |
3257 | * LSan check would not print any actionable information and would just crash | |
3258 | * PID 1. To make this a bit more helpful, let's try to open /dev/console, | |
3259 | * and if we succeed redirect LSan's report there. */ | |
7b19bd60 FS |
3260 | if (getpid_cached() == 1) { |
3261 | _cleanup_close_ int tty_fd = -EBADF; | |
ae5ce7e3 | 3262 | |
7b19bd60 FS |
3263 | tty_fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); |
3264 | if (tty_fd >= 0) | |
3265 | __sanitizer_set_report_fd((void*) (intptr_t) tty_fd); | |
ae5ce7e3 | 3266 | |
7b19bd60 FS |
3267 | __lsan_do_leak_check(); |
3268 | } | |
7e11a95e EV |
3269 | #endif |
3270 | ||
88eec29d DDM |
3271 | if (r < 0) |
3272 | (void) sd_notifyf(0, "ERRNO=%i", -r); | |
3273 | ||
5409c6fc ZJS |
3274 | /* Try to invoke the shutdown binary unless we already failed. |
3275 | * If we failed above, we want to freeze after finishing cleanup. */ | |
4870133b LP |
3276 | if (arg_runtime_scope == RUNTIME_SCOPE_SYSTEM && |
3277 | IN_SET(r, MANAGER_EXIT, MANAGER_REBOOT, MANAGER_POWEROFF, MANAGER_HALT, MANAGER_KEXEC)) { | |
b22d392d | 3278 | r = become_shutdown(r, retval); |
4a36297c | 3279 | log_error_errno(r, "Failed to execute shutdown binary, %s: %m", getpid_cached() == 1 ? "freezing" : "quitting"); |
9b9881d7 | 3280 | error_message = "Failed to execute shutdown binary"; |
b9080b03 FF |
3281 | } |
3282 | ||
3a89cb84 DDM |
3283 | /* This is primarily useful when running systemd in a VM, as it provides the user running the VM with |
3284 | * a mechanism to pick up systemd's exit status in the VM. */ | |
3285 | (void) sd_notifyf(0, "EXIT_STATUS=%i", retval); | |
3286 | ||
8a2c1fbf EJ |
3287 | watchdog_free_device(); |
3288 | arg_watchdog_device = mfree(arg_watchdog_device); | |
3289 | ||
df0ff127 | 3290 | if (getpid_cached() == 1) { |
cb6531be ZJS |
3291 | if (error_message) |
3292 | manager_status_printf(NULL, STATUS_TYPE_EMERGENCY, | |
1fc464f6 | 3293 | ANSI_HIGHLIGHT_RED "!!!!!!" ANSI_NORMAL, |
bb259772 LP |
3294 | "%s.", error_message); |
3295 | freeze_or_exit_or_reboot(); | |
cb6531be | 3296 | } |
c3b3c274 | 3297 | |
7d9eea2b | 3298 | reset_arguments(); |
60918275 LP |
3299 | return retval; |
3300 | } |