]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/init/main.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * GK 2/5/95 - Changed to support mounting root fs via NFS | |
7 | * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96 | |
8 | * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96 | |
dd4d9fec | 9 | * Simplified starting of init: Michael A. Griffith <grif@acm.org> |
1da177e4 LT |
10 | */ |
11 | ||
ea676e84 AM |
12 | #define DEBUG /* Enable initcall_debug */ |
13 | ||
1da177e4 LT |
14 | #include <linux/types.h> |
15 | #include <linux/module.h> | |
16 | #include <linux/proc_fs.h> | |
1da177e4 LT |
17 | #include <linux/kernel.h> |
18 | #include <linux/syscalls.h> | |
9b5609fd | 19 | #include <linux/stackprotector.h> |
1da177e4 LT |
20 | #include <linux/string.h> |
21 | #include <linux/ctype.h> | |
22 | #include <linux/delay.h> | |
1da177e4 LT |
23 | #include <linux/ioport.h> |
24 | #include <linux/init.h> | |
1da177e4 | 25 | #include <linux/initrd.h> |
1da177e4 | 26 | #include <linux/bootmem.h> |
4a7a16dc | 27 | #include <linux/acpi.h> |
1da177e4 | 28 | #include <linux/tty.h> |
1da177e4 LT |
29 | #include <linux/percpu.h> |
30 | #include <linux/kmod.h> | |
db64fe02 | 31 | #include <linux/vmalloc.h> |
1da177e4 | 32 | #include <linux/kernel_stat.h> |
d7cd5611 | 33 | #include <linux/start_kernel.h> |
1da177e4 | 34 | #include <linux/security.h> |
3d442233 | 35 | #include <linux/smp.h> |
1da177e4 LT |
36 | #include <linux/profile.h> |
37 | #include <linux/rcupdate.h> | |
38 | #include <linux/moduleparam.h> | |
39 | #include <linux/kallsyms.h> | |
40 | #include <linux/writeback.h> | |
41 | #include <linux/cpu.h> | |
42 | #include <linux/cpuset.h> | |
ddbcc7e8 | 43 | #include <linux/cgroup.h> |
1da177e4 | 44 | #include <linux/efi.h> |
906568c9 | 45 | #include <linux/tick.h> |
6168a702 | 46 | #include <linux/interrupt.h> |
c757249a | 47 | #include <linux/taskstats_kern.h> |
ca74e92b | 48 | #include <linux/delayacct.h> |
1da177e4 LT |
49 | #include <linux/unistd.h> |
50 | #include <linux/rmap.h> | |
51 | #include <linux/mempolicy.h> | |
52 | #include <linux/key.h> | |
b6cd0b77 | 53 | #include <linux/buffer_head.h> |
eefa864b | 54 | #include <linux/page_ext.h> |
9a11b49a | 55 | #include <linux/debug_locks.h> |
3ac7fe5a | 56 | #include <linux/debugobjects.h> |
fbb9ce95 | 57 | #include <linux/lockdep.h> |
3c7b4e6b | 58 | #include <linux/kmemleak.h> |
84d73786 | 59 | #include <linux/pid_namespace.h> |
1f21782e | 60 | #include <linux/device.h> |
73c27992 | 61 | #include <linux/kthread.h> |
e6fe6649 | 62 | #include <linux/sched.h> |
a1c9eea9 | 63 | #include <linux/signal.h> |
199f0ca5 | 64 | #include <linux/idr.h> |
0b4b3827 | 65 | #include <linux/kgdb.h> |
68bf21aa | 66 | #include <linux/ftrace.h> |
22a9d645 | 67 | #include <linux/async.h> |
dfec072e | 68 | #include <linux/kmemcheck.h> |
6ae6996a | 69 | #include <linux/sfi.h> |
2b2af54a | 70 | #include <linux/shmem_fs.h> |
5a0e3ad6 | 71 | #include <linux/slab.h> |
24a24bb6 | 72 | #include <linux/perf_event.h> |
4a9d4b02 | 73 | #include <linux/file.h> |
a74fb73c | 74 | #include <linux/ptrace.h> |
bb813f4c TH |
75 | #include <linux/blkdev.h> |
76 | #include <linux/elevator.h> | |
38ff87f7 | 77 | #include <linux/sched_clock.h> |
65f382fd | 78 | #include <linux/context_tracking.h> |
47d06e53 | 79 | #include <linux/random.h> |
7b0b73d7 | 80 | #include <linux/list.h> |
c9cd2ce2 | 81 | #include <linux/integrity.h> |
1da177e4 LT |
82 | |
83 | #include <asm/io.h> | |
84 | #include <asm/bugs.h> | |
85 | #include <asm/setup.h> | |
a940199f | 86 | #include <asm/sections.h> |
37b73c82 | 87 | #include <asm/cacheflush.h> |
1da177e4 | 88 | |
1da177e4 LT |
89 | #ifdef CONFIG_X86_LOCAL_APIC |
90 | #include <asm/smp.h> | |
91 | #endif | |
92 | ||
aae5f662 | 93 | static int kernel_init(void *); |
1da177e4 LT |
94 | |
95 | extern void init_IRQ(void); | |
1da177e4 | 96 | extern void fork_init(unsigned long); |
1da177e4 | 97 | extern void radix_tree_init(void); |
37b73c82 AV |
98 | #ifndef CONFIG_DEBUG_RODATA |
99 | static inline void mark_rodata_ro(void) { } | |
100 | #endif | |
1da177e4 | 101 | |
2ce802f6 TH |
102 | /* |
103 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
104 | * where only the boot processor is running with IRQ disabled. This means | |
105 | * two things - IRQ must not be enabled before the flag is cleared and some | |
106 | * operations which are not allowed with IRQ disabled are allowed while the | |
107 | * flag is set. | |
108 | */ | |
109 | bool early_boot_irqs_disabled __read_mostly; | |
110 | ||
a6826048 | 111 | enum system_states system_state __read_mostly; |
1da177e4 LT |
112 | EXPORT_SYMBOL(system_state); |
113 | ||
114 | /* | |
115 | * Boot command-line arguments | |
116 | */ | |
117 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
118 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
119 | ||
120 | extern void time_init(void); | |
121 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 122 | void (*__initdata late_time_init)(void); |
1da177e4 | 123 | |
30d7e0d4 ABL |
124 | /* Untouched command line saved by arch-specific code. */ |
125 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
126 | /* Untouched saved command line (eg. for /proc) */ | |
127 | char *saved_command_line; | |
128 | /* Command line for parameter parsing */ | |
129 | static char *static_command_line; | |
08746a65 KM |
130 | /* Command line for per-initcall parameter parsing */ |
131 | static char *initcall_command_line; | |
1da177e4 LT |
132 | |
133 | static char *execute_command; | |
ffdfc409 | 134 | static char *ramdisk_execute_command; |
1da177e4 | 135 | |
c4b2c0c5 HFS |
136 | /* |
137 | * Used to generate warnings if static_key manipulation functions are used | |
138 | * before jump_label_init is called. | |
139 | */ | |
dd4d9fec | 140 | bool static_key_initialized __read_mostly; |
c4b2c0c5 HFS |
141 | EXPORT_SYMBOL_GPL(static_key_initialized); |
142 | ||
8b3b2955 JB |
143 | /* |
144 | * If set, this is an indication to the drivers that reset the underlying | |
145 | * device before going ahead with the initialization otherwise driver might | |
146 | * rely on the BIOS and skip the reset operation. | |
147 | * | |
148 | * This is useful if kernel is booting in an unreliable environment. | |
149 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
150 | * skipped and devices will be in unknown state. | |
151 | */ | |
152 | unsigned int reset_devices; | |
153 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 154 | |
7e96287d VG |
155 | static int __init set_reset_devices(char *str) |
156 | { | |
157 | reset_devices = 1; | |
158 | return 1; | |
159 | } | |
160 | ||
161 | __setup("reset_devices", set_reset_devices); | |
162 | ||
dd4d9fec FF |
163 | static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
164 | const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
1da177e4 LT |
165 | static const char *panic_later, *panic_param; |
166 | ||
914dcaa8 | 167 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
1da177e4 LT |
168 | |
169 | static int __init obsolete_checksetup(char *line) | |
170 | { | |
914dcaa8 | 171 | const struct obs_kernel_param *p; |
33df0d19 | 172 | int had_early_param = 0; |
1da177e4 LT |
173 | |
174 | p = __setup_start; | |
175 | do { | |
176 | int n = strlen(p->str); | |
b1e4d20c | 177 | if (parameqn(line, p->str, n)) { |
1da177e4 | 178 | if (p->early) { |
33df0d19 RR |
179 | /* Already done in parse_early_param? |
180 | * (Needs exact match on param part). | |
181 | * Keep iterating, as we can have early | |
182 | * params and __setups of same names 8( */ | |
1da177e4 | 183 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 184 | had_early_param = 1; |
1da177e4 | 185 | } else if (!p->setup_func) { |
ea676e84 AM |
186 | pr_warn("Parameter %s is obsolete, ignored\n", |
187 | p->str); | |
1da177e4 LT |
188 | return 1; |
189 | } else if (p->setup_func(line + n)) | |
190 | return 1; | |
191 | } | |
192 | p++; | |
193 | } while (p < __setup_end); | |
33df0d19 RR |
194 | |
195 | return had_early_param; | |
1da177e4 LT |
196 | } |
197 | ||
198 | /* | |
199 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
200 | * still work even if initially too large, it will just take slightly longer | |
201 | */ | |
202 | unsigned long loops_per_jiffy = (1<<12); | |
1da177e4 LT |
203 | EXPORT_SYMBOL(loops_per_jiffy); |
204 | ||
205 | static int __init debug_kernel(char *str) | |
206 | { | |
a8fe19eb | 207 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; |
f6f21c81 | 208 | return 0; |
1da177e4 LT |
209 | } |
210 | ||
211 | static int __init quiet_kernel(char *str) | |
212 | { | |
a8fe19eb | 213 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; |
f6f21c81 | 214 | return 0; |
1da177e4 LT |
215 | } |
216 | ||
f6f21c81 YL |
217 | early_param("debug", debug_kernel); |
218 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
219 | |
220 | static int __init loglevel(char *str) | |
221 | { | |
808bf29b AS |
222 | int newlevel; |
223 | ||
224 | /* | |
225 | * Only update loglevel value when a correct setting was passed, | |
226 | * to prevent blind crashes (when loglevel being set to 0) that | |
227 | * are quite hard to debug | |
228 | */ | |
229 | if (get_option(&str, &newlevel)) { | |
230 | console_loglevel = newlevel; | |
231 | return 0; | |
232 | } | |
233 | ||
234 | return -EINVAL; | |
1da177e4 LT |
235 | } |
236 | ||
f6f21c81 | 237 | early_param("loglevel", loglevel); |
1da177e4 | 238 | |
a99cd112 | 239 | /* Change NUL term back to "=", to make "param" the whole string. */ |
eb157427 | 240 | static int __init repair_env_string(char *param, char *val, const char *unused) |
1da177e4 | 241 | { |
1da177e4 LT |
242 | if (val) { |
243 | /* param=val or param="val"? */ | |
244 | if (val == param+strlen(param)+1) | |
245 | val[-1] = '='; | |
246 | else if (val == param+strlen(param)+2) { | |
247 | val[-2] = '='; | |
248 | memmove(val-1, val, strlen(val)+1); | |
249 | val--; | |
250 | } else | |
251 | BUG(); | |
252 | } | |
a99cd112 CM |
253 | return 0; |
254 | } | |
255 | ||
51e158c1 RR |
256 | /* Anything after -- gets handed straight to init. */ |
257 | static int __init set_init_arg(char *param, char *val, const char *unused) | |
258 | { | |
259 | unsigned int i; | |
260 | ||
261 | if (panic_later) | |
262 | return 0; | |
263 | ||
264 | repair_env_string(param, val, unused); | |
265 | ||
266 | for (i = 0; argv_init[i]; i++) { | |
267 | if (i == MAX_INIT_ARGS) { | |
268 | panic_later = "init"; | |
269 | panic_param = param; | |
270 | return 0; | |
271 | } | |
272 | } | |
273 | argv_init[i] = param; | |
274 | return 0; | |
275 | } | |
276 | ||
a99cd112 CM |
277 | /* |
278 | * Unknown boot options get handed to init, unless they look like | |
279 | * unused parameters (modprobe will find them in /proc/cmdline). | |
280 | */ | |
eb157427 | 281 | static int __init unknown_bootoption(char *param, char *val, const char *unused) |
a99cd112 | 282 | { |
eb157427 | 283 | repair_env_string(param, val, unused); |
1da177e4 LT |
284 | |
285 | /* Handle obsolete-style parameters */ | |
286 | if (obsolete_checksetup(param)) | |
287 | return 0; | |
288 | ||
f066a4f6 RR |
289 | /* Unused module parameter. */ |
290 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 291 | return 0; |
1da177e4 LT |
292 | |
293 | if (panic_later) | |
294 | return 0; | |
295 | ||
296 | if (val) { | |
297 | /* Environment option */ | |
298 | unsigned int i; | |
299 | for (i = 0; envp_init[i]; i++) { | |
300 | if (i == MAX_INIT_ENVS) { | |
499a4584 | 301 | panic_later = "env"; |
1da177e4 LT |
302 | panic_param = param; |
303 | } | |
304 | if (!strncmp(param, envp_init[i], val - param)) | |
305 | break; | |
306 | } | |
307 | envp_init[i] = param; | |
308 | } else { | |
309 | /* Command line option */ | |
310 | unsigned int i; | |
311 | for (i = 0; argv_init[i]; i++) { | |
312 | if (i == MAX_INIT_ARGS) { | |
499a4584 | 313 | panic_later = "init"; |
1da177e4 LT |
314 | panic_param = param; |
315 | } | |
316 | } | |
317 | argv_init[i] = param; | |
318 | } | |
319 | return 0; | |
320 | } | |
321 | ||
322 | static int __init init_setup(char *str) | |
323 | { | |
324 | unsigned int i; | |
325 | ||
326 | execute_command = str; | |
327 | /* | |
328 | * In case LILO is going to boot us with default command line, | |
329 | * it prepends "auto" before the whole cmdline which makes | |
330 | * the shell think it should execute a script with such name. | |
331 | * So we ignore all arguments entered _before_ init=... [MJ] | |
332 | */ | |
333 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
334 | argv_init[i] = NULL; | |
335 | return 1; | |
336 | } | |
337 | __setup("init=", init_setup); | |
338 | ||
ffdfc409 OJ |
339 | static int __init rdinit_setup(char *str) |
340 | { | |
341 | unsigned int i; | |
342 | ||
343 | ramdisk_execute_command = str; | |
344 | /* See "auto" comment in init_setup */ | |
345 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
346 | argv_init[i] = NULL; | |
347 | return 1; | |
348 | } | |
349 | __setup("rdinit=", rdinit_setup); | |
350 | ||
1da177e4 | 351 | #ifndef CONFIG_SMP |
34db18a0 | 352 | static const unsigned int setup_max_cpus = NR_CPUS; |
1da177e4 LT |
353 | #ifdef CONFIG_X86_LOCAL_APIC |
354 | static void __init smp_init(void) | |
355 | { | |
356 | APIC_init_uniprocessor(); | |
357 | } | |
358 | #else | |
359 | #define smp_init() do { } while (0) | |
360 | #endif | |
361 | ||
e0982e90 | 362 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 363 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
364 | #endif |
365 | ||
30d7e0d4 ABL |
366 | /* |
367 | * We need to store the untouched command line for future reference. | |
368 | * We also need to store the touched command line since the parameter | |
369 | * parsing is performed in place, and we should allow a component to | |
370 | * store reference of name/value for future reference. | |
371 | */ | |
372 | static void __init setup_command_line(char *command_line) | |
373 | { | |
098b081b SS |
374 | saved_command_line = |
375 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
376 | initcall_command_line = | |
377 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
378 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
dd4d9fec FF |
379 | strcpy(saved_command_line, boot_command_line); |
380 | strcpy(static_command_line, command_line); | |
30d7e0d4 ABL |
381 | } |
382 | ||
1da177e4 LT |
383 | /* |
384 | * We need to finalize in a non-__init function or else race conditions | |
385 | * between the root thread and the init thread may cause start_kernel to | |
386 | * be reaped by free_initmem before the root thread has proceeded to | |
387 | * cpu_idle. | |
388 | * | |
389 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
390 | */ | |
391 | ||
b433c3d4 PZ |
392 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
393 | ||
f99ebf0a | 394 | static noinline void __init_refok rest_init(void) |
1da177e4 | 395 | { |
73c27992 EB |
396 | int pid; |
397 | ||
7db905e6 | 398 | rcu_scheduler_starting(); |
b433c3d4 | 399 | /* |
97158569 | 400 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
401 | * the init task will end up wanting to create kthreads, which, if |
402 | * we schedule it before we create kthreadd, will OOPS. | |
403 | */ | |
34a1b723 | 404 | kernel_thread(kernel_init, NULL, CLONE_FS); |
1da177e4 | 405 | numa_default_policy(); |
73c27992 | 406 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 407 | rcu_read_lock(); |
5cd20455 | 408 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 409 | rcu_read_unlock(); |
b433c3d4 | 410 | complete(&kthreadd_done); |
f340c0d1 IM |
411 | |
412 | /* | |
413 | * The boot idle thread must execute schedule() | |
1df21055 | 414 | * at least once to get things moving: |
f340c0d1 | 415 | */ |
1df21055 | 416 | init_idle_bootup_task(current); |
bd2f5536 | 417 | schedule_preempt_disabled(); |
5bfb5d69 | 418 | /* Call into cpu_idle with preempt disabled */ |
a1a04ec3 | 419 | cpu_startup_entry(CPUHP_ONLINE); |
1df21055 | 420 | } |
1da177e4 LT |
421 | |
422 | /* Check for early params. */ | |
9fb48c74 | 423 | static int __init do_early_param(char *param, char *val, const char *unused) |
1da177e4 | 424 | { |
914dcaa8 | 425 | const struct obs_kernel_param *p; |
1da177e4 LT |
426 | |
427 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 428 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
429 | (strcmp(param, "console") == 0 && |
430 | strcmp(p->str, "earlycon") == 0) | |
431 | ) { | |
1da177e4 | 432 | if (p->setup_func(val) != 0) |
ea676e84 | 433 | pr_warn("Malformed early option '%s'\n", param); |
1da177e4 LT |
434 | } |
435 | } | |
436 | /* We accept everything at this stage. */ | |
437 | return 0; | |
438 | } | |
439 | ||
13977091 MD |
440 | void __init parse_early_options(char *cmdline) |
441 | { | |
026cee00 | 442 | parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); |
13977091 MD |
443 | } |
444 | ||
1da177e4 LT |
445 | /* Arch code calls this early on, or if not, just before other parsing. */ |
446 | void __init parse_early_param(void) | |
447 | { | |
dd4d9fec FF |
448 | static int done __initdata; |
449 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; | |
1da177e4 LT |
450 | |
451 | if (done) | |
452 | return; | |
453 | ||
454 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 455 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 456 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
457 | done = 1; |
458 | } | |
459 | ||
460 | /* | |
461 | * Activate the first processor. | |
462 | */ | |
463 | ||
44fd2299 SS |
464 | static void __init boot_cpu_init(void) |
465 | { | |
466 | int cpu = smp_processor_id(); | |
467 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 468 | set_cpu_online(cpu, true); |
933b0618 | 469 | set_cpu_active(cpu, true); |
915441b6 RR |
470 | set_cpu_present(cpu, true); |
471 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
472 | } |
473 | ||
839ad62e | 474 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
475 | { |
476 | } | |
477 | ||
eded09cc | 478 | # if THREAD_SIZE >= PAGE_SIZE |
8c9843e5 BH |
479 | void __init __weak thread_info_cache_init(void) |
480 | { | |
481 | } | |
eded09cc | 482 | #endif |
8c9843e5 | 483 | |
444f478f PE |
484 | /* |
485 | * Set up kernel memory allocators | |
486 | */ | |
487 | static void __init mm_init(void) | |
488 | { | |
eefa864b JK |
489 | /* |
490 | * page_ext requires contiguous pages, | |
491 | * bigger than MAX_ORDER unless SPARSEMEM. | |
492 | */ | |
493 | page_ext_init_flatmem(); | |
444f478f PE |
494 | mem_init(); |
495 | kmem_cache_init(); | |
099a19d9 | 496 | percpu_init_late(); |
b35f1819 | 497 | pgtable_init(); |
444f478f PE |
498 | vmalloc_init(); |
499 | } | |
500 | ||
722a9f92 | 501 | asmlinkage __visible void __init start_kernel(void) |
1da177e4 | 502 | { |
dd4d9fec FF |
503 | char *command_line; |
504 | char *after_dashes; | |
033ab7f8 | 505 | |
fbb9ce95 IM |
506 | /* |
507 | * Need to run as early as possible, to initialize the | |
508 | * lockdep hash: | |
509 | */ | |
510 | lockdep_init(); | |
d4311ff1 | 511 | set_task_stack_end_magic(&init_task); |
73839c5b | 512 | smp_setup_processor_id(); |
3ac7fe5a | 513 | debug_objects_early_init(); |
42059429 IM |
514 | |
515 | /* | |
516 | * Set up the the initial canary ASAP: | |
517 | */ | |
518 | boot_init_stack_canary(); | |
519 | ||
ddbcc7e8 | 520 | cgroup_init_early(); |
fbb9ce95 IM |
521 | |
522 | local_irq_disable(); | |
2ce802f6 | 523 | early_boot_irqs_disabled = true; |
fbb9ce95 | 524 | |
1da177e4 LT |
525 | /* |
526 | * Interrupts are still disabled. Do necessary setups, then | |
527 | * enable them | |
528 | */ | |
44fd2299 | 529 | boot_cpu_init(); |
1da177e4 | 530 | page_address_init(); |
ea676e84 | 531 | pr_notice("%s", linux_banner); |
1da177e4 | 532 | setup_arch(&command_line); |
6345d24d | 533 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 534 | setup_command_line(command_line); |
e0982e90 | 535 | setup_nr_cpu_ids(); |
d6647bdf | 536 | setup_per_cpu_areas(); |
44fd2299 | 537 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 538 | |
9adb62a5 | 539 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
540 | page_alloc_init(); |
541 | ||
ea676e84 | 542 | pr_notice("Kernel command line: %s\n", boot_command_line); |
83b519e8 | 543 | parse_early_param(); |
51e158c1 RR |
544 | after_dashes = parse_args("Booting kernel", |
545 | static_command_line, __start___param, | |
546 | __stop___param - __start___param, | |
547 | -1, -1, &unknown_bootoption); | |
3438cf54 | 548 | if (!IS_ERR_OR_NULL(after_dashes)) |
51e158c1 RR |
549 | parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, |
550 | set_init_arg); | |
97ce2c88 JF |
551 | |
552 | jump_label_init(); | |
553 | ||
83b519e8 PE |
554 | /* |
555 | * These use large bootmem allocations and must precede | |
556 | * kmem_cache_init() | |
557 | */ | |
162a7e75 | 558 | setup_log_buf(0); |
83b519e8 | 559 | pidhash_init(); |
83b519e8 PE |
560 | vfs_caches_init_early(); |
561 | sort_main_extable(); | |
562 | trap_init(); | |
444f478f | 563 | mm_init(); |
de03c72c | 564 | |
1da177e4 LT |
565 | /* |
566 | * Set up the scheduler prior starting any interrupts (such as the | |
567 | * timer interrupt). Full topology setup happens at smp_init() | |
568 | * time - but meanwhile we still have a functioning scheduler. | |
569 | */ | |
570 | sched_init(); | |
571 | /* | |
572 | * Disable preemption - early bootup scheduling is extremely | |
573 | * fragile until we cpu_idle() for the first time. | |
574 | */ | |
575 | preempt_disable(); | |
dd4d9fec FF |
576 | if (WARN(!irqs_disabled(), |
577 | "Interrupts were enabled *very* early, fixing it\n")) | |
c4a68306 | 578 | local_irq_disable(); |
9f58a205 | 579 | idr_init_cache(); |
1da177e4 | 580 | rcu_init(); |
5f893b26 SRRH |
581 | |
582 | /* trace_printk() and trace points may be used after this */ | |
583 | trace_init(); | |
584 | ||
65f382fd | 585 | context_tracking_init(); |
773e3eb7 | 586 | radix_tree_init(); |
0b8f1efa YL |
587 | /* init some links before init_ISA_irqs() */ |
588 | early_irq_init(); | |
1da177e4 | 589 | init_IRQ(); |
ad2b1353 | 590 | tick_init(); |
d6dd50e0 | 591 | rcu_init_nohz(); |
1da177e4 | 592 | init_timers(); |
c0a31329 | 593 | hrtimers_init(); |
1da177e4 | 594 | softirq_init(); |
ad596171 | 595 | timekeeping_init(); |
88fecaa2 | 596 | time_init(); |
38ff87f7 | 597 | sched_clock_postinit(); |
9e630205 | 598 | perf_event_init(); |
93e02814 | 599 | profile_init(); |
d8ad7d11 | 600 | call_function_init(); |
f91eb62f | 601 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); |
2ce802f6 | 602 | early_boot_irqs_disabled = false; |
93e02814 | 603 | local_irq_enable(); |
dcce284a | 604 | |
7e85ee0c | 605 | kmem_cache_init_late(); |
1da177e4 LT |
606 | |
607 | /* | |
608 | * HACK ALERT! This is early. We're enabling the console before | |
609 | * we've done PCI setups etc, and console_init() must be aware of | |
610 | * this. But we do want output early, in case something goes wrong. | |
611 | */ | |
612 | console_init(); | |
613 | if (panic_later) | |
499a4584 TH |
614 | panic("Too many boot %s vars at `%s'", panic_later, |
615 | panic_param); | |
fbb9ce95 IM |
616 | |
617 | lockdep_info(); | |
618 | ||
9a11b49a IM |
619 | /* |
620 | * Need to run this when irqs are enabled, because it wants | |
621 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
622 | * too: | |
623 | */ | |
624 | locking_selftest(); | |
625 | ||
1da177e4 LT |
626 | #ifdef CONFIG_BLK_DEV_INITRD |
627 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 628 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
ea676e84 | 629 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", |
bd673c7c GU |
630 | page_to_pfn(virt_to_page((void *)initrd_start)), |
631 | min_low_pfn); | |
1da177e4 LT |
632 | initrd_start = 0; |
633 | } | |
634 | #endif | |
eefa864b | 635 | page_ext_init(); |
3ac7fe5a | 636 | debug_objects_mem_init(); |
9b090f2d | 637 | kmemleak_init(); |
e7c8d5c9 | 638 | setup_per_cpu_pageset(); |
1da177e4 LT |
639 | numa_policy_init(); |
640 | if (late_time_init) | |
641 | late_time_init(); | |
fa84e9ee | 642 | sched_clock_init(); |
1da177e4 LT |
643 | calibrate_delay(); |
644 | pidmap_init(); | |
1da177e4 | 645 | anon_vma_init(); |
c4e1acbb | 646 | acpi_early_init(); |
11520e5e | 647 | #ifdef CONFIG_X86 |
83e68189 | 648 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
11520e5e | 649 | efi_enter_virtual_mode(); |
3891a04a | 650 | #endif |
197725de | 651 | #ifdef CONFIG_X86_ESPFIX64 |
3891a04a PA |
652 | /* Should be run before the first non-init thread is created */ |
653 | init_espfix_bsp(); | |
11520e5e | 654 | #endif |
8c9843e5 | 655 | thread_info_cache_init(); |
d84f4f99 | 656 | cred_init(); |
4481374c | 657 | fork_init(totalram_pages); |
1da177e4 LT |
658 | proc_caches_init(); |
659 | buffer_init(); | |
1da177e4 LT |
660 | key_init(); |
661 | security_init(); | |
0b4b3827 | 662 | dbg_late_init(); |
4481374c | 663 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
664 | signals_init(); |
665 | /* rootfs populating might need page-writeback */ | |
666 | page_writeback_init(); | |
1da177e4 | 667 | proc_root_init(); |
ddbcc7e8 | 668 | cgroup_init(); |
1da177e4 | 669 | cpuset_init(); |
c757249a | 670 | taskstats_init_early(); |
ca74e92b | 671 | delayacct_init(); |
1da177e4 LT |
672 | |
673 | check_bugs(); | |
674 | ||
6ae6996a | 675 | sfi_init_late(); |
1da177e4 | 676 | |
83e68189 | 677 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
2223af38 | 678 | efi_late_init(); |
78510792 | 679 | efi_free_boot_services(); |
2223af38 | 680 | } |
78510792 | 681 | |
68bf21aa SR |
682 | ftrace_init(); |
683 | ||
1da177e4 LT |
684 | /* Do the rest non-__init'ed, we're now alive */ |
685 | rest_init(); | |
686 | } | |
687 | ||
b99b87f7 PO |
688 | /* Call all constructor functions linked into the kernel. */ |
689 | static void __init do_ctors(void) | |
690 | { | |
691 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 692 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 693 | |
196a15b4 HS |
694 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
695 | (*fn)(); | |
b99b87f7 PO |
696 | #endif |
697 | } | |
698 | ||
2329abfa | 699 | bool initcall_debug; |
d0ea3d7d | 700 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 701 | |
7b0b73d7 PB |
702 | #ifdef CONFIG_KALLSYMS |
703 | struct blacklist_entry { | |
704 | struct list_head next; | |
705 | char *buf; | |
706 | }; | |
707 | ||
708 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); | |
709 | ||
710 | static int __init initcall_blacklist(char *str) | |
711 | { | |
712 | char *str_entry; | |
713 | struct blacklist_entry *entry; | |
714 | ||
715 | /* str argument is a comma-separated list of functions */ | |
716 | do { | |
717 | str_entry = strsep(&str, ","); | |
718 | if (str_entry) { | |
719 | pr_debug("blacklisting initcall %s\n", str_entry); | |
720 | entry = alloc_bootmem(sizeof(*entry)); | |
721 | entry->buf = alloc_bootmem(strlen(str_entry) + 1); | |
722 | strcpy(entry->buf, str_entry); | |
723 | list_add(&entry->next, &blacklisted_initcalls); | |
724 | } | |
725 | } while (str_entry); | |
726 | ||
727 | return 0; | |
728 | } | |
729 | ||
730 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
731 | { | |
732 | struct list_head *tmp; | |
733 | struct blacklist_entry *entry; | |
734 | char *fn_name; | |
735 | ||
736 | fn_name = kasprintf(GFP_KERNEL, "%pf", fn); | |
737 | if (!fn_name) | |
738 | return false; | |
739 | ||
740 | list_for_each(tmp, &blacklisted_initcalls) { | |
741 | entry = list_entry(tmp, struct blacklist_entry, next); | |
742 | if (!strcmp(fn_name, entry->buf)) { | |
743 | pr_debug("initcall %s blacklisted\n", fn_name); | |
744 | kfree(fn_name); | |
745 | return true; | |
746 | } | |
747 | } | |
748 | ||
749 | kfree(fn_name); | |
750 | return false; | |
751 | } | |
752 | #else | |
753 | static int __init initcall_blacklist(char *str) | |
754 | { | |
755 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); | |
756 | return 0; | |
757 | } | |
758 | ||
759 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
760 | { | |
761 | return false; | |
762 | } | |
763 | #endif | |
764 | __setup("initcall_blacklist=", initcall_blacklist); | |
765 | ||
e4461271 | 766 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 767 | { |
74239072 | 768 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
769 | unsigned long long duration; |
770 | int ret; | |
1da177e4 | 771 | |
d62cf815 | 772 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
22c5c03b | 773 | calltime = ktime_get(); |
30dbb20e | 774 | ret = fn(); |
22c5c03b KW |
775 | rettime = ktime_get(); |
776 | delta = ktime_sub(rettime, calltime); | |
777 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
d62cf815 | 778 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", |
ea676e84 | 779 | fn, ret, duration); |
1da177e4 | 780 | |
22c5c03b KW |
781 | return ret; |
782 | } | |
783 | ||
e4461271 | 784 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
785 | { |
786 | int count = preempt_count(); | |
787 | int ret; | |
ff1c8fac | 788 | char msgbuf[64]; |
22c5c03b | 789 | |
7b0b73d7 PB |
790 | if (initcall_blacklisted(fn)) |
791 | return -EPERM; | |
792 | ||
22c5c03b KW |
793 | if (initcall_debug) |
794 | ret = do_one_initcall_debug(fn); | |
795 | else | |
796 | ret = fn(); | |
8f0c45cd | 797 | |
e0df154f | 798 | msgbuf[0] = 0; |
e662e1cf | 799 | |
e0df154f | 800 | if (preempt_count() != count) { |
bf5d770b | 801 | sprintf(msgbuf, "preemption imbalance "); |
4a2b4b22 | 802 | preempt_count_set(count); |
1da177e4 | 803 | } |
e0df154f | 804 | if (irqs_disabled()) { |
a76bfd0d | 805 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
806 | local_irq_enable(); |
807 | } | |
f91eb62f | 808 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
59f9415f | 809 | |
30dbb20e | 810 | return ret; |
e0df154f LT |
811 | } |
812 | ||
813 | ||
026cee00 PM |
814 | extern initcall_t __initcall_start[]; |
815 | extern initcall_t __initcall0_start[]; | |
816 | extern initcall_t __initcall1_start[]; | |
817 | extern initcall_t __initcall2_start[]; | |
818 | extern initcall_t __initcall3_start[]; | |
819 | extern initcall_t __initcall4_start[]; | |
820 | extern initcall_t __initcall5_start[]; | |
821 | extern initcall_t __initcall6_start[]; | |
822 | extern initcall_t __initcall7_start[]; | |
823 | extern initcall_t __initcall_end[]; | |
824 | ||
825 | static initcall_t *initcall_levels[] __initdata = { | |
826 | __initcall0_start, | |
827 | __initcall1_start, | |
828 | __initcall2_start, | |
829 | __initcall3_start, | |
830 | __initcall4_start, | |
831 | __initcall5_start, | |
832 | __initcall6_start, | |
833 | __initcall7_start, | |
834 | __initcall_end, | |
835 | }; | |
836 | ||
96263d28 | 837 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 838 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
839 | "early", |
840 | "core", | |
841 | "postcore", | |
842 | "arch", | |
843 | "subsys", | |
844 | "fs", | |
845 | "device", | |
846 | "late", | |
026cee00 PM |
847 | }; |
848 | ||
026cee00 | 849 | static void __init do_initcall_level(int level) |
e0df154f | 850 | { |
196a15b4 | 851 | initcall_t *fn; |
e0df154f | 852 | |
08746a65 | 853 | strcpy(initcall_command_line, saved_command_line); |
026cee00 | 854 | parse_args(initcall_level_names[level], |
08746a65 | 855 | initcall_command_line, __start___param, |
026cee00 PM |
856 | __stop___param - __start___param, |
857 | level, level, | |
eb157427 | 858 | &repair_env_string); |
026cee00 PM |
859 | |
860 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 861 | do_one_initcall(*fn); |
1da177e4 LT |
862 | } |
863 | ||
026cee00 PM |
864 | static void __init do_initcalls(void) |
865 | { | |
866 | int level; | |
867 | ||
19efb72f | 868 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
869 | do_initcall_level(level); |
870 | } | |
871 | ||
1da177e4 LT |
872 | /* |
873 | * Ok, the machine is now initialized. None of the devices | |
874 | * have been touched yet, but the CPU subsystem is up and | |
875 | * running, and memory and process management works. | |
876 | * | |
877 | * Now we can finally start doing some real work.. | |
878 | */ | |
879 | static void __init do_basic_setup(void) | |
880 | { | |
759ee091 | 881 | cpuset_init_smp(); |
1da177e4 | 882 | usermodehelper_init(); |
41ffe5d5 | 883 | shmem_init(); |
1da177e4 | 884 | driver_init(); |
b04c3afb | 885 | init_irq_proc(); |
b99b87f7 | 886 | do_ctors(); |
d5767c53 | 887 | usermodehelper_enable(); |
b0f84374 | 888 | do_initcalls(); |
47d06e53 | 889 | random_int_secret_init(); |
1da177e4 LT |
890 | } |
891 | ||
7babe8db | 892 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 893 | { |
196a15b4 | 894 | initcall_t *fn; |
c2147a50 | 895 | |
026cee00 | 896 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 897 | do_one_initcall(*fn); |
c2147a50 EGM |
898 | } |
899 | ||
bb813f4c TH |
900 | /* |
901 | * This function requests modules which should be loaded by default and is | |
902 | * called twice right after initrd is mounted and right before init is | |
903 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
904 | * loaded before control is passed to userland. | |
905 | */ | |
906 | void __init load_default_modules(void) | |
907 | { | |
908 | load_default_elevator_module(); | |
909 | } | |
910 | ||
a74fb73c | 911 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
912 | { |
913 | argv_init[0] = init_filename; | |
c4ad8f98 | 914 | return do_execve(getname_kernel(init_filename), |
ae903caa AV |
915 | (const char __user *const __user *)argv_init, |
916 | (const char __user *const __user *)envp_init); | |
1da177e4 LT |
917 | } |
918 | ||
ba24762b MO |
919 | static int try_to_run_init_process(const char *init_filename) |
920 | { | |
921 | int ret; | |
922 | ||
923 | ret = run_init_process(init_filename); | |
924 | ||
925 | if (ret && ret != -ENOENT) { | |
926 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
927 | init_filename, ret); | |
928 | } | |
929 | ||
930 | return ret; | |
931 | } | |
932 | ||
f80b0c90 | 933 | static noinline void __init kernel_init_freeable(void); |
d6b21238 AV |
934 | |
935 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 936 | { |
ba24762b MO |
937 | int ret; |
938 | ||
d6b21238 | 939 | kernel_init_freeable(); |
22a9d645 AV |
940 | /* need to finish all async __init code before freeing the memory */ |
941 | async_synchronize_full(); | |
ee5bfa64 | 942 | free_initmem(); |
ee5bfa64 VG |
943 | mark_rodata_ro(); |
944 | system_state = SYSTEM_RUNNING; | |
945 | numa_default_policy(); | |
946 | ||
4a9d4b02 | 947 | flush_delayed_fput(); |
fae5fa44 | 948 | |
ee5bfa64 | 949 | if (ramdisk_execute_command) { |
ba24762b MO |
950 | ret = run_init_process(ramdisk_execute_command); |
951 | if (!ret) | |
a74fb73c | 952 | return 0; |
ba24762b MO |
953 | pr_err("Failed to execute %s (error %d)\n", |
954 | ramdisk_execute_command, ret); | |
ee5bfa64 VG |
955 | } |
956 | ||
957 | /* | |
958 | * We try each of these until one succeeds. | |
959 | * | |
960 | * The Bourne shell can be used instead of init if we are | |
961 | * trying to recover a really broken machine. | |
962 | */ | |
963 | if (execute_command) { | |
ba24762b MO |
964 | ret = run_init_process(execute_command); |
965 | if (!ret) | |
a74fb73c | 966 | return 0; |
6ef4536e AL |
967 | #ifndef CONFIG_INIT_FALLBACK |
968 | panic("Requested init %s failed (error %d).", | |
969 | execute_command, ret); | |
970 | #else | |
ba24762b | 971 | pr_err("Failed to execute %s (error %d). Attempting defaults...\n", |
6ef4536e AL |
972 | execute_command, ret); |
973 | #endif | |
ee5bfa64 | 974 | } |
ba24762b MO |
975 | if (!try_to_run_init_process("/sbin/init") || |
976 | !try_to_run_init_process("/etc/init") || | |
977 | !try_to_run_init_process("/bin/init") || | |
978 | !try_to_run_init_process("/bin/sh")) | |
a74fb73c | 979 | return 0; |
ee5bfa64 | 980 | |
ba24762b | 981 | panic("No working init found. Try passing init= option to kernel. " |
9a85b8d6 | 982 | "See Linux Documentation/init.txt for guidance."); |
ee5bfa64 VG |
983 | } |
984 | ||
f80b0c90 | 985 | static noinline void __init kernel_init_freeable(void) |
1da177e4 | 986 | { |
b433c3d4 PZ |
987 | /* |
988 | * Wait until kthreadd is all set-up. | |
989 | */ | |
990 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
991 | |
992 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
993 | gfp_allowed_mask = __GFP_BITS_MASK; | |
994 | ||
58568d2a MX |
995 | /* |
996 | * init can allocate pages on any node | |
997 | */ | |
3c466d46 | 998 | set_mems_allowed(node_states[N_MEMORY]); |
1da177e4 LT |
999 | /* |
1000 | * init can run on any cpu. | |
1001 | */ | |
1a2142af | 1002 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 1003 | |
9ec52099 CLG |
1004 | cad_pid = task_pid(current); |
1005 | ||
ca74a6f8 | 1006 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
1007 | |
1008 | do_pre_smp_initcalls(); | |
004417a6 | 1009 | lockup_detector_init(); |
1da177e4 | 1010 | |
1da177e4 LT |
1011 | smp_init(); |
1012 | sched_init_smp(); | |
1013 | ||
1da177e4 LT |
1014 | do_basic_setup(); |
1015 | ||
2bd3a997 EB |
1016 | /* Open the /dev/console on the rootfs, this should never fail */ |
1017 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
ea676e84 | 1018 | pr_err("Warning: unable to open an initial console.\n"); |
2bd3a997 EB |
1019 | |
1020 | (void) sys_dup(0); | |
1021 | (void) sys_dup(0); | |
1da177e4 LT |
1022 | /* |
1023 | * check if there is an early userspace init. If yes, let it do all | |
1024 | * the work | |
1025 | */ | |
ffdfc409 OJ |
1026 | |
1027 | if (!ramdisk_execute_command) | |
1028 | ramdisk_execute_command = "/init"; | |
1029 | ||
1030 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
1031 | ramdisk_execute_command = NULL; | |
1da177e4 | 1032 | prepare_namespace(); |
ffdfc409 | 1033 | } |
1da177e4 LT |
1034 | |
1035 | /* | |
1036 | * Ok, we have completed the initial bootup, and | |
1037 | * we're essentially up and running. Get rid of the | |
1038 | * initmem segments and start the user-mode stuff.. | |
c9cd2ce2 DK |
1039 | * |
1040 | * rootfs is available now, try loading the public keys | |
1041 | * and default modules | |
1da177e4 | 1042 | */ |
bb813f4c | 1043 | |
c9cd2ce2 | 1044 | integrity_load_keys(); |
bb813f4c | 1045 | load_default_modules(); |
1da177e4 | 1046 | } |