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> |
e149ed2b | 82 | #include <linux/proc_ns.h> |
1da177e4 LT |
83 | |
84 | #include <asm/io.h> | |
85 | #include <asm/bugs.h> | |
86 | #include <asm/setup.h> | |
a940199f | 87 | #include <asm/sections.h> |
37b73c82 | 88 | #include <asm/cacheflush.h> |
1da177e4 | 89 | |
aae5f662 | 90 | static int kernel_init(void *); |
1da177e4 LT |
91 | |
92 | extern void init_IRQ(void); | |
1da177e4 | 93 | extern void fork_init(unsigned long); |
1da177e4 | 94 | extern void radix_tree_init(void); |
37b73c82 AV |
95 | #ifndef CONFIG_DEBUG_RODATA |
96 | static inline void mark_rodata_ro(void) { } | |
97 | #endif | |
1da177e4 | 98 | |
2ce802f6 TH |
99 | /* |
100 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
101 | * where only the boot processor is running with IRQ disabled. This means | |
102 | * two things - IRQ must not be enabled before the flag is cleared and some | |
103 | * operations which are not allowed with IRQ disabled are allowed while the | |
104 | * flag is set. | |
105 | */ | |
106 | bool early_boot_irqs_disabled __read_mostly; | |
107 | ||
a6826048 | 108 | enum system_states system_state __read_mostly; |
1da177e4 LT |
109 | EXPORT_SYMBOL(system_state); |
110 | ||
111 | /* | |
112 | * Boot command-line arguments | |
113 | */ | |
114 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
115 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
116 | ||
117 | extern void time_init(void); | |
118 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 119 | void (*__initdata late_time_init)(void); |
1da177e4 | 120 | |
30d7e0d4 ABL |
121 | /* Untouched command line saved by arch-specific code. */ |
122 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
123 | /* Untouched saved command line (eg. for /proc) */ | |
124 | char *saved_command_line; | |
125 | /* Command line for parameter parsing */ | |
126 | static char *static_command_line; | |
08746a65 KM |
127 | /* Command line for per-initcall parameter parsing */ |
128 | static char *initcall_command_line; | |
1da177e4 LT |
129 | |
130 | static char *execute_command; | |
ffdfc409 | 131 | static char *ramdisk_execute_command; |
1da177e4 | 132 | |
c4b2c0c5 HFS |
133 | /* |
134 | * Used to generate warnings if static_key manipulation functions are used | |
135 | * before jump_label_init is called. | |
136 | */ | |
dd4d9fec | 137 | bool static_key_initialized __read_mostly; |
c4b2c0c5 HFS |
138 | EXPORT_SYMBOL_GPL(static_key_initialized); |
139 | ||
8b3b2955 JB |
140 | /* |
141 | * If set, this is an indication to the drivers that reset the underlying | |
142 | * device before going ahead with the initialization otherwise driver might | |
143 | * rely on the BIOS and skip the reset operation. | |
144 | * | |
145 | * This is useful if kernel is booting in an unreliable environment. | |
fc454fdc | 146 | * For ex. kdump situation where previous kernel has crashed, BIOS has been |
8b3b2955 JB |
147 | * skipped and devices will be in unknown state. |
148 | */ | |
149 | unsigned int reset_devices; | |
150 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 151 | |
7e96287d VG |
152 | static int __init set_reset_devices(char *str) |
153 | { | |
154 | reset_devices = 1; | |
155 | return 1; | |
156 | } | |
157 | ||
158 | __setup("reset_devices", set_reset_devices); | |
159 | ||
dd4d9fec FF |
160 | static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
161 | const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
1da177e4 LT |
162 | static const char *panic_later, *panic_param; |
163 | ||
914dcaa8 | 164 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
1da177e4 LT |
165 | |
166 | static int __init obsolete_checksetup(char *line) | |
167 | { | |
914dcaa8 | 168 | const struct obs_kernel_param *p; |
33df0d19 | 169 | int had_early_param = 0; |
1da177e4 LT |
170 | |
171 | p = __setup_start; | |
172 | do { | |
173 | int n = strlen(p->str); | |
b1e4d20c | 174 | if (parameqn(line, p->str, n)) { |
1da177e4 | 175 | if (p->early) { |
33df0d19 RR |
176 | /* Already done in parse_early_param? |
177 | * (Needs exact match on param part). | |
178 | * Keep iterating, as we can have early | |
179 | * params and __setups of same names 8( */ | |
1da177e4 | 180 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 181 | had_early_param = 1; |
1da177e4 | 182 | } else if (!p->setup_func) { |
ea676e84 AM |
183 | pr_warn("Parameter %s is obsolete, ignored\n", |
184 | p->str); | |
1da177e4 LT |
185 | return 1; |
186 | } else if (p->setup_func(line + n)) | |
187 | return 1; | |
188 | } | |
189 | p++; | |
190 | } while (p < __setup_end); | |
33df0d19 RR |
191 | |
192 | return had_early_param; | |
1da177e4 LT |
193 | } |
194 | ||
195 | /* | |
196 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
197 | * still work even if initially too large, it will just take slightly longer | |
198 | */ | |
199 | unsigned long loops_per_jiffy = (1<<12); | |
1da177e4 LT |
200 | EXPORT_SYMBOL(loops_per_jiffy); |
201 | ||
202 | static int __init debug_kernel(char *str) | |
203 | { | |
a8fe19eb | 204 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; |
f6f21c81 | 205 | return 0; |
1da177e4 LT |
206 | } |
207 | ||
208 | static int __init quiet_kernel(char *str) | |
209 | { | |
a8fe19eb | 210 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; |
f6f21c81 | 211 | return 0; |
1da177e4 LT |
212 | } |
213 | ||
f6f21c81 YL |
214 | early_param("debug", debug_kernel); |
215 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
216 | |
217 | static int __init loglevel(char *str) | |
218 | { | |
808bf29b AS |
219 | int newlevel; |
220 | ||
221 | /* | |
222 | * Only update loglevel value when a correct setting was passed, | |
223 | * to prevent blind crashes (when loglevel being set to 0) that | |
224 | * are quite hard to debug | |
225 | */ | |
226 | if (get_option(&str, &newlevel)) { | |
227 | console_loglevel = newlevel; | |
228 | return 0; | |
229 | } | |
230 | ||
231 | return -EINVAL; | |
1da177e4 LT |
232 | } |
233 | ||
f6f21c81 | 234 | early_param("loglevel", loglevel); |
1da177e4 | 235 | |
a99cd112 | 236 | /* Change NUL term back to "=", to make "param" the whole string. */ |
eb157427 | 237 | static int __init repair_env_string(char *param, char *val, const char *unused) |
1da177e4 | 238 | { |
1da177e4 LT |
239 | if (val) { |
240 | /* param=val or param="val"? */ | |
241 | if (val == param+strlen(param)+1) | |
242 | val[-1] = '='; | |
243 | else if (val == param+strlen(param)+2) { | |
244 | val[-2] = '='; | |
245 | memmove(val-1, val, strlen(val)+1); | |
246 | val--; | |
247 | } else | |
248 | BUG(); | |
249 | } | |
a99cd112 CM |
250 | return 0; |
251 | } | |
252 | ||
51e158c1 RR |
253 | /* Anything after -- gets handed straight to init. */ |
254 | static int __init set_init_arg(char *param, char *val, const char *unused) | |
255 | { | |
256 | unsigned int i; | |
257 | ||
258 | if (panic_later) | |
259 | return 0; | |
260 | ||
261 | repair_env_string(param, val, unused); | |
262 | ||
263 | for (i = 0; argv_init[i]; i++) { | |
264 | if (i == MAX_INIT_ARGS) { | |
265 | panic_later = "init"; | |
266 | panic_param = param; | |
267 | return 0; | |
268 | } | |
269 | } | |
270 | argv_init[i] = param; | |
271 | return 0; | |
272 | } | |
273 | ||
a99cd112 CM |
274 | /* |
275 | * Unknown boot options get handed to init, unless they look like | |
276 | * unused parameters (modprobe will find them in /proc/cmdline). | |
277 | */ | |
eb157427 | 278 | static int __init unknown_bootoption(char *param, char *val, const char *unused) |
a99cd112 | 279 | { |
eb157427 | 280 | repair_env_string(param, val, unused); |
1da177e4 LT |
281 | |
282 | /* Handle obsolete-style parameters */ | |
283 | if (obsolete_checksetup(param)) | |
284 | return 0; | |
285 | ||
f066a4f6 RR |
286 | /* Unused module parameter. */ |
287 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 288 | return 0; |
1da177e4 LT |
289 | |
290 | if (panic_later) | |
291 | return 0; | |
292 | ||
293 | if (val) { | |
294 | /* Environment option */ | |
295 | unsigned int i; | |
296 | for (i = 0; envp_init[i]; i++) { | |
297 | if (i == MAX_INIT_ENVS) { | |
499a4584 | 298 | panic_later = "env"; |
1da177e4 LT |
299 | panic_param = param; |
300 | } | |
301 | if (!strncmp(param, envp_init[i], val - param)) | |
302 | break; | |
303 | } | |
304 | envp_init[i] = param; | |
305 | } else { | |
306 | /* Command line option */ | |
307 | unsigned int i; | |
308 | for (i = 0; argv_init[i]; i++) { | |
309 | if (i == MAX_INIT_ARGS) { | |
499a4584 | 310 | panic_later = "init"; |
1da177e4 LT |
311 | panic_param = param; |
312 | } | |
313 | } | |
314 | argv_init[i] = param; | |
315 | } | |
316 | return 0; | |
317 | } | |
318 | ||
319 | static int __init init_setup(char *str) | |
320 | { | |
321 | unsigned int i; | |
322 | ||
323 | execute_command = str; | |
324 | /* | |
325 | * In case LILO is going to boot us with default command line, | |
326 | * it prepends "auto" before the whole cmdline which makes | |
327 | * the shell think it should execute a script with such name. | |
328 | * So we ignore all arguments entered _before_ init=... [MJ] | |
329 | */ | |
330 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
331 | argv_init[i] = NULL; | |
332 | return 1; | |
333 | } | |
334 | __setup("init=", init_setup); | |
335 | ||
ffdfc409 OJ |
336 | static int __init rdinit_setup(char *str) |
337 | { | |
338 | unsigned int i; | |
339 | ||
340 | ramdisk_execute_command = str; | |
341 | /* See "auto" comment in init_setup */ | |
342 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
343 | argv_init[i] = NULL; | |
344 | return 1; | |
345 | } | |
346 | __setup("rdinit=", rdinit_setup); | |
347 | ||
1da177e4 | 348 | #ifndef CONFIG_SMP |
34db18a0 | 349 | static const unsigned int setup_max_cpus = NR_CPUS; |
e0982e90 | 350 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 351 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
352 | #endif |
353 | ||
30d7e0d4 ABL |
354 | /* |
355 | * We need to store the untouched command line for future reference. | |
356 | * We also need to store the touched command line since the parameter | |
357 | * parsing is performed in place, and we should allow a component to | |
358 | * store reference of name/value for future reference. | |
359 | */ | |
360 | static void __init setup_command_line(char *command_line) | |
361 | { | |
098b081b SS |
362 | saved_command_line = |
363 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
364 | initcall_command_line = | |
365 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
366 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
dd4d9fec FF |
367 | strcpy(saved_command_line, boot_command_line); |
368 | strcpy(static_command_line, command_line); | |
30d7e0d4 ABL |
369 | } |
370 | ||
1da177e4 LT |
371 | /* |
372 | * We need to finalize in a non-__init function or else race conditions | |
373 | * between the root thread and the init thread may cause start_kernel to | |
374 | * be reaped by free_initmem before the root thread has proceeded to | |
375 | * cpu_idle. | |
376 | * | |
377 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
378 | */ | |
379 | ||
b433c3d4 PZ |
380 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
381 | ||
f99ebf0a | 382 | static noinline void __init_refok rest_init(void) |
1da177e4 | 383 | { |
73c27992 EB |
384 | int pid; |
385 | ||
7db905e6 | 386 | rcu_scheduler_starting(); |
00df35f9 | 387 | smpboot_thread_init(); |
b433c3d4 | 388 | /* |
97158569 | 389 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
390 | * the init task will end up wanting to create kthreads, which, if |
391 | * we schedule it before we create kthreadd, will OOPS. | |
392 | */ | |
34a1b723 | 393 | kernel_thread(kernel_init, NULL, CLONE_FS); |
1da177e4 | 394 | numa_default_policy(); |
73c27992 | 395 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 396 | rcu_read_lock(); |
5cd20455 | 397 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 398 | rcu_read_unlock(); |
b433c3d4 | 399 | complete(&kthreadd_done); |
f340c0d1 IM |
400 | |
401 | /* | |
402 | * The boot idle thread must execute schedule() | |
1df21055 | 403 | * at least once to get things moving: |
f340c0d1 | 404 | */ |
1df21055 | 405 | init_idle_bootup_task(current); |
bd2f5536 | 406 | schedule_preempt_disabled(); |
5bfb5d69 | 407 | /* Call into cpu_idle with preempt disabled */ |
a1a04ec3 | 408 | cpu_startup_entry(CPUHP_ONLINE); |
1df21055 | 409 | } |
1da177e4 LT |
410 | |
411 | /* Check for early params. */ | |
9fb48c74 | 412 | static int __init do_early_param(char *param, char *val, const char *unused) |
1da177e4 | 413 | { |
914dcaa8 | 414 | const struct obs_kernel_param *p; |
1da177e4 LT |
415 | |
416 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 417 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
418 | (strcmp(param, "console") == 0 && |
419 | strcmp(p->str, "earlycon") == 0) | |
420 | ) { | |
1da177e4 | 421 | if (p->setup_func(val) != 0) |
ea676e84 | 422 | pr_warn("Malformed early option '%s'\n", param); |
1da177e4 LT |
423 | } |
424 | } | |
425 | /* We accept everything at this stage. */ | |
426 | return 0; | |
427 | } | |
428 | ||
13977091 MD |
429 | void __init parse_early_options(char *cmdline) |
430 | { | |
026cee00 | 431 | parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); |
13977091 MD |
432 | } |
433 | ||
1da177e4 LT |
434 | /* Arch code calls this early on, or if not, just before other parsing. */ |
435 | void __init parse_early_param(void) | |
436 | { | |
dd4d9fec FF |
437 | static int done __initdata; |
438 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; | |
1da177e4 LT |
439 | |
440 | if (done) | |
441 | return; | |
442 | ||
443 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 444 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 445 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
446 | done = 1; |
447 | } | |
448 | ||
449 | /* | |
450 | * Activate the first processor. | |
451 | */ | |
452 | ||
44fd2299 SS |
453 | static void __init boot_cpu_init(void) |
454 | { | |
455 | int cpu = smp_processor_id(); | |
456 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 457 | set_cpu_online(cpu, true); |
933b0618 | 458 | set_cpu_active(cpu, true); |
915441b6 RR |
459 | set_cpu_present(cpu, true); |
460 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
461 | } |
462 | ||
839ad62e | 463 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
464 | { |
465 | } | |
466 | ||
eded09cc | 467 | # if THREAD_SIZE >= PAGE_SIZE |
8c9843e5 BH |
468 | void __init __weak thread_info_cache_init(void) |
469 | { | |
470 | } | |
eded09cc | 471 | #endif |
8c9843e5 | 472 | |
444f478f PE |
473 | /* |
474 | * Set up kernel memory allocators | |
475 | */ | |
476 | static void __init mm_init(void) | |
477 | { | |
eefa864b JK |
478 | /* |
479 | * page_ext requires contiguous pages, | |
480 | * bigger than MAX_ORDER unless SPARSEMEM. | |
481 | */ | |
482 | page_ext_init_flatmem(); | |
444f478f PE |
483 | mem_init(); |
484 | kmem_cache_init(); | |
099a19d9 | 485 | percpu_init_late(); |
b35f1819 | 486 | pgtable_init(); |
444f478f PE |
487 | vmalloc_init(); |
488 | } | |
489 | ||
722a9f92 | 490 | asmlinkage __visible void __init start_kernel(void) |
1da177e4 | 491 | { |
dd4d9fec FF |
492 | char *command_line; |
493 | char *after_dashes; | |
033ab7f8 | 494 | |
fbb9ce95 IM |
495 | /* |
496 | * Need to run as early as possible, to initialize the | |
497 | * lockdep hash: | |
498 | */ | |
499 | lockdep_init(); | |
d4311ff1 | 500 | set_task_stack_end_magic(&init_task); |
73839c5b | 501 | smp_setup_processor_id(); |
3ac7fe5a | 502 | debug_objects_early_init(); |
42059429 IM |
503 | |
504 | /* | |
505 | * Set up the the initial canary ASAP: | |
506 | */ | |
507 | boot_init_stack_canary(); | |
508 | ||
ddbcc7e8 | 509 | cgroup_init_early(); |
fbb9ce95 IM |
510 | |
511 | local_irq_disable(); | |
2ce802f6 | 512 | early_boot_irqs_disabled = true; |
fbb9ce95 | 513 | |
1da177e4 LT |
514 | /* |
515 | * Interrupts are still disabled. Do necessary setups, then | |
516 | * enable them | |
517 | */ | |
44fd2299 | 518 | boot_cpu_init(); |
1da177e4 | 519 | page_address_init(); |
ea676e84 | 520 | pr_notice("%s", linux_banner); |
1da177e4 | 521 | setup_arch(&command_line); |
6345d24d | 522 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 523 | setup_command_line(command_line); |
e0982e90 | 524 | setup_nr_cpu_ids(); |
d6647bdf | 525 | setup_per_cpu_areas(); |
44fd2299 | 526 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 527 | |
9adb62a5 | 528 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
529 | page_alloc_init(); |
530 | ||
ea676e84 | 531 | pr_notice("Kernel command line: %s\n", boot_command_line); |
83b519e8 | 532 | parse_early_param(); |
51e158c1 RR |
533 | after_dashes = parse_args("Booting kernel", |
534 | static_command_line, __start___param, | |
535 | __stop___param - __start___param, | |
536 | -1, -1, &unknown_bootoption); | |
3438cf54 | 537 | if (!IS_ERR_OR_NULL(after_dashes)) |
51e158c1 RR |
538 | parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, |
539 | set_init_arg); | |
97ce2c88 JF |
540 | |
541 | jump_label_init(); | |
542 | ||
83b519e8 PE |
543 | /* |
544 | * These use large bootmem allocations and must precede | |
545 | * kmem_cache_init() | |
546 | */ | |
162a7e75 | 547 | setup_log_buf(0); |
83b519e8 | 548 | pidhash_init(); |
83b519e8 PE |
549 | vfs_caches_init_early(); |
550 | sort_main_extable(); | |
551 | trap_init(); | |
444f478f | 552 | mm_init(); |
de03c72c | 553 | |
1da177e4 LT |
554 | /* |
555 | * Set up the scheduler prior starting any interrupts (such as the | |
556 | * timer interrupt). Full topology setup happens at smp_init() | |
557 | * time - but meanwhile we still have a functioning scheduler. | |
558 | */ | |
559 | sched_init(); | |
560 | /* | |
561 | * Disable preemption - early bootup scheduling is extremely | |
562 | * fragile until we cpu_idle() for the first time. | |
563 | */ | |
564 | preempt_disable(); | |
dd4d9fec FF |
565 | if (WARN(!irqs_disabled(), |
566 | "Interrupts were enabled *very* early, fixing it\n")) | |
c4a68306 | 567 | local_irq_disable(); |
9f58a205 | 568 | idr_init_cache(); |
1da177e4 | 569 | rcu_init(); |
5f893b26 SRRH |
570 | |
571 | /* trace_printk() and trace points may be used after this */ | |
572 | trace_init(); | |
573 | ||
65f382fd | 574 | context_tracking_init(); |
773e3eb7 | 575 | radix_tree_init(); |
0b8f1efa YL |
576 | /* init some links before init_ISA_irqs() */ |
577 | early_irq_init(); | |
1da177e4 | 578 | init_IRQ(); |
ad2b1353 | 579 | tick_init(); |
d6dd50e0 | 580 | rcu_init_nohz(); |
1da177e4 | 581 | init_timers(); |
c0a31329 | 582 | hrtimers_init(); |
1da177e4 | 583 | softirq_init(); |
ad596171 | 584 | timekeeping_init(); |
88fecaa2 | 585 | time_init(); |
38ff87f7 | 586 | sched_clock_postinit(); |
9e630205 | 587 | perf_event_init(); |
93e02814 | 588 | profile_init(); |
d8ad7d11 | 589 | call_function_init(); |
f91eb62f | 590 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); |
2ce802f6 | 591 | early_boot_irqs_disabled = false; |
93e02814 | 592 | local_irq_enable(); |
dcce284a | 593 | |
7e85ee0c | 594 | kmem_cache_init_late(); |
1da177e4 LT |
595 | |
596 | /* | |
597 | * HACK ALERT! This is early. We're enabling the console before | |
598 | * we've done PCI setups etc, and console_init() must be aware of | |
599 | * this. But we do want output early, in case something goes wrong. | |
600 | */ | |
601 | console_init(); | |
602 | if (panic_later) | |
499a4584 TH |
603 | panic("Too many boot %s vars at `%s'", panic_later, |
604 | panic_param); | |
fbb9ce95 IM |
605 | |
606 | lockdep_info(); | |
607 | ||
9a11b49a IM |
608 | /* |
609 | * Need to run this when irqs are enabled, because it wants | |
610 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
611 | * too: | |
612 | */ | |
613 | locking_selftest(); | |
614 | ||
1da177e4 LT |
615 | #ifdef CONFIG_BLK_DEV_INITRD |
616 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 617 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
ea676e84 | 618 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", |
bd673c7c GU |
619 | page_to_pfn(virt_to_page((void *)initrd_start)), |
620 | min_low_pfn); | |
1da177e4 LT |
621 | initrd_start = 0; |
622 | } | |
623 | #endif | |
eefa864b | 624 | page_ext_init(); |
3ac7fe5a | 625 | debug_objects_mem_init(); |
9b090f2d | 626 | kmemleak_init(); |
e7c8d5c9 | 627 | setup_per_cpu_pageset(); |
1da177e4 LT |
628 | numa_policy_init(); |
629 | if (late_time_init) | |
630 | late_time_init(); | |
fa84e9ee | 631 | sched_clock_init(); |
1da177e4 LT |
632 | calibrate_delay(); |
633 | pidmap_init(); | |
1da177e4 | 634 | anon_vma_init(); |
c4e1acbb | 635 | acpi_early_init(); |
11520e5e | 636 | #ifdef CONFIG_X86 |
83e68189 | 637 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
11520e5e | 638 | efi_enter_virtual_mode(); |
3891a04a | 639 | #endif |
197725de | 640 | #ifdef CONFIG_X86_ESPFIX64 |
3891a04a PA |
641 | /* Should be run before the first non-init thread is created */ |
642 | init_espfix_bsp(); | |
11520e5e | 643 | #endif |
8c9843e5 | 644 | thread_info_cache_init(); |
d84f4f99 | 645 | cred_init(); |
4481374c | 646 | fork_init(totalram_pages); |
1da177e4 LT |
647 | proc_caches_init(); |
648 | buffer_init(); | |
1da177e4 LT |
649 | key_init(); |
650 | security_init(); | |
0b4b3827 | 651 | dbg_late_init(); |
4481374c | 652 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
653 | signals_init(); |
654 | /* rootfs populating might need page-writeback */ | |
655 | page_writeback_init(); | |
1da177e4 | 656 | proc_root_init(); |
e149ed2b | 657 | nsfs_init(); |
1da177e4 | 658 | cpuset_init(); |
695df213 | 659 | cgroup_init(); |
c757249a | 660 | taskstats_init_early(); |
ca74e92b | 661 | delayacct_init(); |
1da177e4 LT |
662 | |
663 | check_bugs(); | |
664 | ||
6ae6996a | 665 | sfi_init_late(); |
1da177e4 | 666 | |
83e68189 | 667 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
2223af38 | 668 | efi_late_init(); |
78510792 | 669 | efi_free_boot_services(); |
2223af38 | 670 | } |
78510792 | 671 | |
68bf21aa SR |
672 | ftrace_init(); |
673 | ||
1da177e4 LT |
674 | /* Do the rest non-__init'ed, we're now alive */ |
675 | rest_init(); | |
676 | } | |
677 | ||
b99b87f7 PO |
678 | /* Call all constructor functions linked into the kernel. */ |
679 | static void __init do_ctors(void) | |
680 | { | |
681 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 682 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 683 | |
196a15b4 HS |
684 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
685 | (*fn)(); | |
b99b87f7 PO |
686 | #endif |
687 | } | |
688 | ||
2329abfa | 689 | bool initcall_debug; |
d0ea3d7d | 690 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 691 | |
7b0b73d7 PB |
692 | #ifdef CONFIG_KALLSYMS |
693 | struct blacklist_entry { | |
694 | struct list_head next; | |
695 | char *buf; | |
696 | }; | |
697 | ||
698 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); | |
699 | ||
700 | static int __init initcall_blacklist(char *str) | |
701 | { | |
702 | char *str_entry; | |
703 | struct blacklist_entry *entry; | |
704 | ||
705 | /* str argument is a comma-separated list of functions */ | |
706 | do { | |
707 | str_entry = strsep(&str, ","); | |
708 | if (str_entry) { | |
709 | pr_debug("blacklisting initcall %s\n", str_entry); | |
710 | entry = alloc_bootmem(sizeof(*entry)); | |
711 | entry->buf = alloc_bootmem(strlen(str_entry) + 1); | |
712 | strcpy(entry->buf, str_entry); | |
713 | list_add(&entry->next, &blacklisted_initcalls); | |
714 | } | |
715 | } while (str_entry); | |
716 | ||
717 | return 0; | |
718 | } | |
719 | ||
720 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
721 | { | |
722 | struct list_head *tmp; | |
723 | struct blacklist_entry *entry; | |
724 | char *fn_name; | |
725 | ||
726 | fn_name = kasprintf(GFP_KERNEL, "%pf", fn); | |
727 | if (!fn_name) | |
728 | return false; | |
729 | ||
730 | list_for_each(tmp, &blacklisted_initcalls) { | |
731 | entry = list_entry(tmp, struct blacklist_entry, next); | |
732 | if (!strcmp(fn_name, entry->buf)) { | |
733 | pr_debug("initcall %s blacklisted\n", fn_name); | |
734 | kfree(fn_name); | |
735 | return true; | |
736 | } | |
737 | } | |
738 | ||
739 | kfree(fn_name); | |
740 | return false; | |
741 | } | |
742 | #else | |
743 | static int __init initcall_blacklist(char *str) | |
744 | { | |
745 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); | |
746 | return 0; | |
747 | } | |
748 | ||
749 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
750 | { | |
751 | return false; | |
752 | } | |
753 | #endif | |
754 | __setup("initcall_blacklist=", initcall_blacklist); | |
755 | ||
e4461271 | 756 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 757 | { |
74239072 | 758 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
759 | unsigned long long duration; |
760 | int ret; | |
1da177e4 | 761 | |
d62cf815 | 762 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
22c5c03b | 763 | calltime = ktime_get(); |
30dbb20e | 764 | ret = fn(); |
22c5c03b KW |
765 | rettime = ktime_get(); |
766 | delta = ktime_sub(rettime, calltime); | |
767 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
d62cf815 | 768 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", |
ea676e84 | 769 | fn, ret, duration); |
1da177e4 | 770 | |
22c5c03b KW |
771 | return ret; |
772 | } | |
773 | ||
e4461271 | 774 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
775 | { |
776 | int count = preempt_count(); | |
777 | int ret; | |
ff1c8fac | 778 | char msgbuf[64]; |
22c5c03b | 779 | |
7b0b73d7 PB |
780 | if (initcall_blacklisted(fn)) |
781 | return -EPERM; | |
782 | ||
22c5c03b KW |
783 | if (initcall_debug) |
784 | ret = do_one_initcall_debug(fn); | |
785 | else | |
786 | ret = fn(); | |
8f0c45cd | 787 | |
e0df154f | 788 | msgbuf[0] = 0; |
e662e1cf | 789 | |
e0df154f | 790 | if (preempt_count() != count) { |
bf5d770b | 791 | sprintf(msgbuf, "preemption imbalance "); |
4a2b4b22 | 792 | preempt_count_set(count); |
1da177e4 | 793 | } |
e0df154f | 794 | if (irqs_disabled()) { |
a76bfd0d | 795 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
796 | local_irq_enable(); |
797 | } | |
f91eb62f | 798 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
59f9415f | 799 | |
30dbb20e | 800 | return ret; |
e0df154f LT |
801 | } |
802 | ||
803 | ||
026cee00 PM |
804 | extern initcall_t __initcall_start[]; |
805 | extern initcall_t __initcall0_start[]; | |
806 | extern initcall_t __initcall1_start[]; | |
807 | extern initcall_t __initcall2_start[]; | |
808 | extern initcall_t __initcall3_start[]; | |
809 | extern initcall_t __initcall4_start[]; | |
810 | extern initcall_t __initcall5_start[]; | |
811 | extern initcall_t __initcall6_start[]; | |
812 | extern initcall_t __initcall7_start[]; | |
813 | extern initcall_t __initcall_end[]; | |
814 | ||
815 | static initcall_t *initcall_levels[] __initdata = { | |
816 | __initcall0_start, | |
817 | __initcall1_start, | |
818 | __initcall2_start, | |
819 | __initcall3_start, | |
820 | __initcall4_start, | |
821 | __initcall5_start, | |
822 | __initcall6_start, | |
823 | __initcall7_start, | |
824 | __initcall_end, | |
825 | }; | |
826 | ||
96263d28 | 827 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 828 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
829 | "early", |
830 | "core", | |
831 | "postcore", | |
832 | "arch", | |
833 | "subsys", | |
834 | "fs", | |
835 | "device", | |
836 | "late", | |
026cee00 PM |
837 | }; |
838 | ||
026cee00 | 839 | static void __init do_initcall_level(int level) |
e0df154f | 840 | { |
196a15b4 | 841 | initcall_t *fn; |
e0df154f | 842 | |
08746a65 | 843 | strcpy(initcall_command_line, saved_command_line); |
026cee00 | 844 | parse_args(initcall_level_names[level], |
08746a65 | 845 | initcall_command_line, __start___param, |
026cee00 PM |
846 | __stop___param - __start___param, |
847 | level, level, | |
eb157427 | 848 | &repair_env_string); |
026cee00 PM |
849 | |
850 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 851 | do_one_initcall(*fn); |
1da177e4 LT |
852 | } |
853 | ||
026cee00 PM |
854 | static void __init do_initcalls(void) |
855 | { | |
856 | int level; | |
857 | ||
19efb72f | 858 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
859 | do_initcall_level(level); |
860 | } | |
861 | ||
1da177e4 LT |
862 | /* |
863 | * Ok, the machine is now initialized. None of the devices | |
864 | * have been touched yet, but the CPU subsystem is up and | |
865 | * running, and memory and process management works. | |
866 | * | |
867 | * Now we can finally start doing some real work.. | |
868 | */ | |
869 | static void __init do_basic_setup(void) | |
870 | { | |
759ee091 | 871 | cpuset_init_smp(); |
1da177e4 | 872 | usermodehelper_init(); |
41ffe5d5 | 873 | shmem_init(); |
1da177e4 | 874 | driver_init(); |
b04c3afb | 875 | init_irq_proc(); |
b99b87f7 | 876 | do_ctors(); |
d5767c53 | 877 | usermodehelper_enable(); |
b0f84374 | 878 | do_initcalls(); |
47d06e53 | 879 | random_int_secret_init(); |
1da177e4 LT |
880 | } |
881 | ||
7babe8db | 882 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 883 | { |
196a15b4 | 884 | initcall_t *fn; |
c2147a50 | 885 | |
026cee00 | 886 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 887 | do_one_initcall(*fn); |
c2147a50 EGM |
888 | } |
889 | ||
bb813f4c TH |
890 | /* |
891 | * This function requests modules which should be loaded by default and is | |
892 | * called twice right after initrd is mounted and right before init is | |
893 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
894 | * loaded before control is passed to userland. | |
895 | */ | |
896 | void __init load_default_modules(void) | |
897 | { | |
898 | load_default_elevator_module(); | |
899 | } | |
900 | ||
a74fb73c | 901 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
902 | { |
903 | argv_init[0] = init_filename; | |
c4ad8f98 | 904 | return do_execve(getname_kernel(init_filename), |
ae903caa AV |
905 | (const char __user *const __user *)argv_init, |
906 | (const char __user *const __user *)envp_init); | |
1da177e4 LT |
907 | } |
908 | ||
ba24762b MO |
909 | static int try_to_run_init_process(const char *init_filename) |
910 | { | |
911 | int ret; | |
912 | ||
913 | ret = run_init_process(init_filename); | |
914 | ||
915 | if (ret && ret != -ENOENT) { | |
916 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
917 | init_filename, ret); | |
918 | } | |
919 | ||
920 | return ret; | |
921 | } | |
922 | ||
f80b0c90 | 923 | static noinline void __init kernel_init_freeable(void); |
d6b21238 AV |
924 | |
925 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 926 | { |
ba24762b MO |
927 | int ret; |
928 | ||
d6b21238 | 929 | kernel_init_freeable(); |
22a9d645 AV |
930 | /* need to finish all async __init code before freeing the memory */ |
931 | async_synchronize_full(); | |
ee5bfa64 | 932 | free_initmem(); |
ee5bfa64 VG |
933 | mark_rodata_ro(); |
934 | system_state = SYSTEM_RUNNING; | |
935 | numa_default_policy(); | |
936 | ||
4a9d4b02 | 937 | flush_delayed_fput(); |
fae5fa44 | 938 | |
ee5bfa64 | 939 | if (ramdisk_execute_command) { |
ba24762b MO |
940 | ret = run_init_process(ramdisk_execute_command); |
941 | if (!ret) | |
a74fb73c | 942 | return 0; |
ba24762b MO |
943 | pr_err("Failed to execute %s (error %d)\n", |
944 | ramdisk_execute_command, ret); | |
ee5bfa64 VG |
945 | } |
946 | ||
947 | /* | |
948 | * We try each of these until one succeeds. | |
949 | * | |
950 | * The Bourne shell can be used instead of init if we are | |
951 | * trying to recover a really broken machine. | |
952 | */ | |
953 | if (execute_command) { | |
ba24762b MO |
954 | ret = run_init_process(execute_command); |
955 | if (!ret) | |
a74fb73c | 956 | return 0; |
6ef4536e AL |
957 | panic("Requested init %s failed (error %d).", |
958 | execute_command, ret); | |
ee5bfa64 | 959 | } |
ba24762b MO |
960 | if (!try_to_run_init_process("/sbin/init") || |
961 | !try_to_run_init_process("/etc/init") || | |
962 | !try_to_run_init_process("/bin/init") || | |
963 | !try_to_run_init_process("/bin/sh")) | |
a74fb73c | 964 | return 0; |
ee5bfa64 | 965 | |
ba24762b | 966 | panic("No working init found. Try passing init= option to kernel. " |
9a85b8d6 | 967 | "See Linux Documentation/init.txt for guidance."); |
ee5bfa64 VG |
968 | } |
969 | ||
f80b0c90 | 970 | static noinline void __init kernel_init_freeable(void) |
1da177e4 | 971 | { |
b433c3d4 PZ |
972 | /* |
973 | * Wait until kthreadd is all set-up. | |
974 | */ | |
975 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
976 | |
977 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
978 | gfp_allowed_mask = __GFP_BITS_MASK; | |
979 | ||
58568d2a MX |
980 | /* |
981 | * init can allocate pages on any node | |
982 | */ | |
3c466d46 | 983 | set_mems_allowed(node_states[N_MEMORY]); |
1da177e4 LT |
984 | /* |
985 | * init can run on any cpu. | |
986 | */ | |
1a2142af | 987 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 988 | |
9ec52099 CLG |
989 | cad_pid = task_pid(current); |
990 | ||
ca74a6f8 | 991 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
992 | |
993 | do_pre_smp_initcalls(); | |
004417a6 | 994 | lockup_detector_init(); |
1da177e4 | 995 | |
1da177e4 LT |
996 | smp_init(); |
997 | sched_init_smp(); | |
998 | ||
1da177e4 LT |
999 | do_basic_setup(); |
1000 | ||
2bd3a997 EB |
1001 | /* Open the /dev/console on the rootfs, this should never fail */ |
1002 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
ea676e84 | 1003 | pr_err("Warning: unable to open an initial console.\n"); |
2bd3a997 EB |
1004 | |
1005 | (void) sys_dup(0); | |
1006 | (void) sys_dup(0); | |
1da177e4 LT |
1007 | /* |
1008 | * check if there is an early userspace init. If yes, let it do all | |
1009 | * the work | |
1010 | */ | |
ffdfc409 OJ |
1011 | |
1012 | if (!ramdisk_execute_command) | |
1013 | ramdisk_execute_command = "/init"; | |
1014 | ||
1015 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
1016 | ramdisk_execute_command = NULL; | |
1da177e4 | 1017 | prepare_namespace(); |
ffdfc409 | 1018 | } |
1da177e4 LT |
1019 | |
1020 | /* | |
1021 | * Ok, we have completed the initial bootup, and | |
1022 | * we're essentially up and running. Get rid of the | |
1023 | * initmem segments and start the user-mode stuff.. | |
c9cd2ce2 DK |
1024 | * |
1025 | * rootfs is available now, try loading the public keys | |
1026 | * and default modules | |
1da177e4 | 1027 | */ |
bb813f4c | 1028 | |
c9cd2ce2 | 1029 | integrity_load_keys(); |
bb813f4c | 1030 | load_default_modules(); |
1da177e4 | 1031 | } |