]>
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> |
0ddab1d2 | 83 | #include <linux/io.h> |
1da177e4 LT |
84 | |
85 | #include <asm/io.h> | |
86 | #include <asm/bugs.h> | |
87 | #include <asm/setup.h> | |
a940199f | 88 | #include <asm/sections.h> |
37b73c82 | 89 | #include <asm/cacheflush.h> |
1da177e4 | 90 | |
aae5f662 | 91 | static int kernel_init(void *); |
1da177e4 LT |
92 | |
93 | extern void init_IRQ(void); | |
ff691f6e | 94 | extern void fork_init(void); |
1da177e4 | 95 | extern void radix_tree_init(void); |
37b73c82 AV |
96 | #ifndef CONFIG_DEBUG_RODATA |
97 | static inline void mark_rodata_ro(void) { } | |
98 | #endif | |
1da177e4 | 99 | |
2ce802f6 TH |
100 | /* |
101 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
102 | * where only the boot processor is running with IRQ disabled. This means | |
103 | * two things - IRQ must not be enabled before the flag is cleared and some | |
104 | * operations which are not allowed with IRQ disabled are allowed while the | |
105 | * flag is set. | |
106 | */ | |
107 | bool early_boot_irqs_disabled __read_mostly; | |
108 | ||
a6826048 | 109 | enum system_states system_state __read_mostly; |
1da177e4 LT |
110 | EXPORT_SYMBOL(system_state); |
111 | ||
112 | /* | |
113 | * Boot command-line arguments | |
114 | */ | |
115 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
116 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
117 | ||
118 | extern void time_init(void); | |
119 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 120 | void (*__initdata late_time_init)(void); |
1da177e4 | 121 | |
30d7e0d4 ABL |
122 | /* Untouched command line saved by arch-specific code. */ |
123 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
124 | /* Untouched saved command line (eg. for /proc) */ | |
125 | char *saved_command_line; | |
126 | /* Command line for parameter parsing */ | |
127 | static char *static_command_line; | |
08746a65 KM |
128 | /* Command line for per-initcall parameter parsing */ |
129 | static char *initcall_command_line; | |
1da177e4 LT |
130 | |
131 | static char *execute_command; | |
ffdfc409 | 132 | static char *ramdisk_execute_command; |
1da177e4 | 133 | |
c4b2c0c5 HFS |
134 | /* |
135 | * Used to generate warnings if static_key manipulation functions are used | |
136 | * before jump_label_init is called. | |
137 | */ | |
dd4d9fec | 138 | bool static_key_initialized __read_mostly; |
c4b2c0c5 HFS |
139 | EXPORT_SYMBOL_GPL(static_key_initialized); |
140 | ||
8b3b2955 JB |
141 | /* |
142 | * If set, this is an indication to the drivers that reset the underlying | |
143 | * device before going ahead with the initialization otherwise driver might | |
144 | * rely on the BIOS and skip the reset operation. | |
145 | * | |
146 | * This is useful if kernel is booting in an unreliable environment. | |
fc454fdc | 147 | * For ex. kdump situation where previous kernel has crashed, BIOS has been |
8b3b2955 JB |
148 | * skipped and devices will be in unknown state. |
149 | */ | |
150 | unsigned int reset_devices; | |
151 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 152 | |
7e96287d VG |
153 | static int __init set_reset_devices(char *str) |
154 | { | |
155 | reset_devices = 1; | |
156 | return 1; | |
157 | } | |
158 | ||
159 | __setup("reset_devices", set_reset_devices); | |
160 | ||
dd4d9fec FF |
161 | static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
162 | const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
1da177e4 LT |
163 | static const char *panic_later, *panic_param; |
164 | ||
914dcaa8 | 165 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
1da177e4 | 166 | |
31c025b5 | 167 | static bool __init obsolete_checksetup(char *line) |
1da177e4 | 168 | { |
914dcaa8 | 169 | const struct obs_kernel_param *p; |
31c025b5 | 170 | bool had_early_param = false; |
1da177e4 LT |
171 | |
172 | p = __setup_start; | |
173 | do { | |
174 | int n = strlen(p->str); | |
b1e4d20c | 175 | if (parameqn(line, p->str, n)) { |
1da177e4 | 176 | if (p->early) { |
33df0d19 RR |
177 | /* Already done in parse_early_param? |
178 | * (Needs exact match on param part). | |
179 | * Keep iterating, as we can have early | |
180 | * params and __setups of same names 8( */ | |
1da177e4 | 181 | if (line[n] == '\0' || line[n] == '=') |
31c025b5 | 182 | had_early_param = true; |
1da177e4 | 183 | } else if (!p->setup_func) { |
ea676e84 AM |
184 | pr_warn("Parameter %s is obsolete, ignored\n", |
185 | p->str); | |
31c025b5 | 186 | return true; |
1da177e4 | 187 | } else if (p->setup_func(line + n)) |
31c025b5 | 188 | return true; |
1da177e4 LT |
189 | } |
190 | p++; | |
191 | } while (p < __setup_end); | |
33df0d19 RR |
192 | |
193 | return had_early_param; | |
1da177e4 LT |
194 | } |
195 | ||
196 | /* | |
197 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
198 | * still work even if initially too large, it will just take slightly longer | |
199 | */ | |
200 | unsigned long loops_per_jiffy = (1<<12); | |
1da177e4 LT |
201 | EXPORT_SYMBOL(loops_per_jiffy); |
202 | ||
203 | static int __init debug_kernel(char *str) | |
204 | { | |
a8fe19eb | 205 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; |
f6f21c81 | 206 | return 0; |
1da177e4 LT |
207 | } |
208 | ||
209 | static int __init quiet_kernel(char *str) | |
210 | { | |
a8fe19eb | 211 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; |
f6f21c81 | 212 | return 0; |
1da177e4 LT |
213 | } |
214 | ||
f6f21c81 YL |
215 | early_param("debug", debug_kernel); |
216 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
217 | |
218 | static int __init loglevel(char *str) | |
219 | { | |
808bf29b AS |
220 | int newlevel; |
221 | ||
222 | /* | |
223 | * Only update loglevel value when a correct setting was passed, | |
224 | * to prevent blind crashes (when loglevel being set to 0) that | |
225 | * are quite hard to debug | |
226 | */ | |
227 | if (get_option(&str, &newlevel)) { | |
228 | console_loglevel = newlevel; | |
229 | return 0; | |
230 | } | |
231 | ||
232 | return -EINVAL; | |
1da177e4 LT |
233 | } |
234 | ||
f6f21c81 | 235 | early_param("loglevel", loglevel); |
1da177e4 | 236 | |
a99cd112 | 237 | /* Change NUL term back to "=", to make "param" the whole string. */ |
ecc86170 LR |
238 | static int __init repair_env_string(char *param, char *val, |
239 | const char *unused, void *arg) | |
1da177e4 | 240 | { |
1da177e4 LT |
241 | if (val) { |
242 | /* param=val or param="val"? */ | |
243 | if (val == param+strlen(param)+1) | |
244 | val[-1] = '='; | |
245 | else if (val == param+strlen(param)+2) { | |
246 | val[-2] = '='; | |
247 | memmove(val-1, val, strlen(val)+1); | |
248 | val--; | |
249 | } else | |
250 | BUG(); | |
251 | } | |
a99cd112 CM |
252 | return 0; |
253 | } | |
254 | ||
51e158c1 | 255 | /* Anything after -- gets handed straight to init. */ |
ecc86170 LR |
256 | static int __init set_init_arg(char *param, char *val, |
257 | const char *unused, void *arg) | |
51e158c1 RR |
258 | { |
259 | unsigned int i; | |
260 | ||
261 | if (panic_later) | |
262 | return 0; | |
263 | ||
ecc86170 | 264 | repair_env_string(param, val, unused, NULL); |
51e158c1 RR |
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 | */ | |
ecc86170 LR |
281 | static int __init unknown_bootoption(char *param, char *val, |
282 | const char *unused, void *arg) | |
a99cd112 | 283 | { |
ecc86170 | 284 | repair_env_string(param, val, unused, NULL); |
1da177e4 LT |
285 | |
286 | /* Handle obsolete-style parameters */ | |
287 | if (obsolete_checksetup(param)) | |
288 | return 0; | |
289 | ||
f066a4f6 RR |
290 | /* Unused module parameter. */ |
291 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 292 | return 0; |
1da177e4 LT |
293 | |
294 | if (panic_later) | |
295 | return 0; | |
296 | ||
297 | if (val) { | |
298 | /* Environment option */ | |
299 | unsigned int i; | |
300 | for (i = 0; envp_init[i]; i++) { | |
301 | if (i == MAX_INIT_ENVS) { | |
499a4584 | 302 | panic_later = "env"; |
1da177e4 LT |
303 | panic_param = param; |
304 | } | |
305 | if (!strncmp(param, envp_init[i], val - param)) | |
306 | break; | |
307 | } | |
308 | envp_init[i] = param; | |
309 | } else { | |
310 | /* Command line option */ | |
311 | unsigned int i; | |
312 | for (i = 0; argv_init[i]; i++) { | |
313 | if (i == MAX_INIT_ARGS) { | |
499a4584 | 314 | panic_later = "init"; |
1da177e4 LT |
315 | panic_param = param; |
316 | } | |
317 | } | |
318 | argv_init[i] = param; | |
319 | } | |
320 | return 0; | |
321 | } | |
322 | ||
323 | static int __init init_setup(char *str) | |
324 | { | |
325 | unsigned int i; | |
326 | ||
327 | execute_command = str; | |
328 | /* | |
329 | * In case LILO is going to boot us with default command line, | |
330 | * it prepends "auto" before the whole cmdline which makes | |
331 | * the shell think it should execute a script with such name. | |
332 | * So we ignore all arguments entered _before_ init=... [MJ] | |
333 | */ | |
334 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
335 | argv_init[i] = NULL; | |
336 | return 1; | |
337 | } | |
338 | __setup("init=", init_setup); | |
339 | ||
ffdfc409 OJ |
340 | static int __init rdinit_setup(char *str) |
341 | { | |
342 | unsigned int i; | |
343 | ||
344 | ramdisk_execute_command = str; | |
345 | /* See "auto" comment in init_setup */ | |
346 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
347 | argv_init[i] = NULL; | |
348 | return 1; | |
349 | } | |
350 | __setup("rdinit=", rdinit_setup); | |
351 | ||
1da177e4 | 352 | #ifndef CONFIG_SMP |
34db18a0 | 353 | static const unsigned int setup_max_cpus = NR_CPUS; |
e0982e90 | 354 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 355 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
356 | #endif |
357 | ||
30d7e0d4 ABL |
358 | /* |
359 | * We need to store the untouched command line for future reference. | |
360 | * We also need to store the touched command line since the parameter | |
361 | * parsing is performed in place, and we should allow a component to | |
362 | * store reference of name/value for future reference. | |
363 | */ | |
364 | static void __init setup_command_line(char *command_line) | |
365 | { | |
098b081b SS |
366 | saved_command_line = |
367 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
368 | initcall_command_line = | |
369 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
370 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
dd4d9fec FF |
371 | strcpy(saved_command_line, boot_command_line); |
372 | strcpy(static_command_line, command_line); | |
30d7e0d4 ABL |
373 | } |
374 | ||
1da177e4 LT |
375 | /* |
376 | * We need to finalize in a non-__init function or else race conditions | |
377 | * between the root thread and the init thread may cause start_kernel to | |
378 | * be reaped by free_initmem before the root thread has proceeded to | |
379 | * cpu_idle. | |
380 | * | |
381 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
382 | */ | |
383 | ||
b433c3d4 PZ |
384 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
385 | ||
f99ebf0a | 386 | static noinline void __init_refok rest_init(void) |
1da177e4 | 387 | { |
73c27992 EB |
388 | int pid; |
389 | ||
7db905e6 | 390 | rcu_scheduler_starting(); |
b433c3d4 | 391 | /* |
97158569 | 392 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
393 | * the init task will end up wanting to create kthreads, which, if |
394 | * we schedule it before we create kthreadd, will OOPS. | |
395 | */ | |
34a1b723 | 396 | kernel_thread(kernel_init, NULL, CLONE_FS); |
1da177e4 | 397 | numa_default_policy(); |
73c27992 | 398 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 399 | rcu_read_lock(); |
5cd20455 | 400 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 401 | rcu_read_unlock(); |
b433c3d4 | 402 | complete(&kthreadd_done); |
f340c0d1 IM |
403 | |
404 | /* | |
405 | * The boot idle thread must execute schedule() | |
1df21055 | 406 | * at least once to get things moving: |
f340c0d1 | 407 | */ |
1df21055 | 408 | init_idle_bootup_task(current); |
bd2f5536 | 409 | schedule_preempt_disabled(); |
5bfb5d69 | 410 | /* Call into cpu_idle with preempt disabled */ |
a1a04ec3 | 411 | cpu_startup_entry(CPUHP_ONLINE); |
1df21055 | 412 | } |
1da177e4 LT |
413 | |
414 | /* Check for early params. */ | |
ecc86170 LR |
415 | static int __init do_early_param(char *param, char *val, |
416 | const char *unused, void *arg) | |
1da177e4 | 417 | { |
914dcaa8 | 418 | const struct obs_kernel_param *p; |
1da177e4 LT |
419 | |
420 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 421 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
422 | (strcmp(param, "console") == 0 && |
423 | strcmp(p->str, "earlycon") == 0) | |
424 | ) { | |
1da177e4 | 425 | if (p->setup_func(val) != 0) |
ea676e84 | 426 | pr_warn("Malformed early option '%s'\n", param); |
1da177e4 LT |
427 | } |
428 | } | |
429 | /* We accept everything at this stage. */ | |
430 | return 0; | |
431 | } | |
432 | ||
13977091 MD |
433 | void __init parse_early_options(char *cmdline) |
434 | { | |
ecc86170 LR |
435 | parse_args("early options", cmdline, NULL, 0, 0, 0, NULL, |
436 | do_early_param); | |
13977091 MD |
437 | } |
438 | ||
1da177e4 LT |
439 | /* Arch code calls this early on, or if not, just before other parsing. */ |
440 | void __init parse_early_param(void) | |
441 | { | |
dd4d9fec FF |
442 | static int done __initdata; |
443 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; | |
1da177e4 LT |
444 | |
445 | if (done) | |
446 | return; | |
447 | ||
448 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 449 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 450 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
451 | done = 1; |
452 | } | |
453 | ||
839ad62e | 454 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
455 | { |
456 | } | |
457 | ||
eded09cc | 458 | # if THREAD_SIZE >= PAGE_SIZE |
8c9843e5 BH |
459 | void __init __weak thread_info_cache_init(void) |
460 | { | |
461 | } | |
eded09cc | 462 | #endif |
8c9843e5 | 463 | |
444f478f PE |
464 | /* |
465 | * Set up kernel memory allocators | |
466 | */ | |
467 | static void __init mm_init(void) | |
468 | { | |
eefa864b JK |
469 | /* |
470 | * page_ext requires contiguous pages, | |
471 | * bigger than MAX_ORDER unless SPARSEMEM. | |
472 | */ | |
473 | page_ext_init_flatmem(); | |
444f478f PE |
474 | mem_init(); |
475 | kmem_cache_init(); | |
099a19d9 | 476 | percpu_init_late(); |
b35f1819 | 477 | pgtable_init(); |
444f478f | 478 | vmalloc_init(); |
0ddab1d2 | 479 | ioremap_huge_init(); |
444f478f PE |
480 | } |
481 | ||
722a9f92 | 482 | asmlinkage __visible void __init start_kernel(void) |
1da177e4 | 483 | { |
dd4d9fec FF |
484 | char *command_line; |
485 | char *after_dashes; | |
033ab7f8 | 486 | |
fbb9ce95 IM |
487 | /* |
488 | * Need to run as early as possible, to initialize the | |
489 | * lockdep hash: | |
490 | */ | |
491 | lockdep_init(); | |
d4311ff1 | 492 | set_task_stack_end_magic(&init_task); |
73839c5b | 493 | smp_setup_processor_id(); |
3ac7fe5a | 494 | debug_objects_early_init(); |
42059429 IM |
495 | |
496 | /* | |
497 | * Set up the the initial canary ASAP: | |
498 | */ | |
499 | boot_init_stack_canary(); | |
500 | ||
ddbcc7e8 | 501 | cgroup_init_early(); |
fbb9ce95 IM |
502 | |
503 | local_irq_disable(); | |
2ce802f6 | 504 | early_boot_irqs_disabled = true; |
fbb9ce95 | 505 | |
1da177e4 LT |
506 | /* |
507 | * Interrupts are still disabled. Do necessary setups, then | |
508 | * enable them | |
509 | */ | |
44fd2299 | 510 | boot_cpu_init(); |
1da177e4 | 511 | page_address_init(); |
ea676e84 | 512 | pr_notice("%s", linux_banner); |
1da177e4 | 513 | setup_arch(&command_line); |
6345d24d | 514 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 515 | setup_command_line(command_line); |
e0982e90 | 516 | setup_nr_cpu_ids(); |
d6647bdf | 517 | setup_per_cpu_areas(); |
cff7d378 | 518 | boot_cpu_state_init(); |
44fd2299 | 519 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 520 | |
9adb62a5 | 521 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
522 | page_alloc_init(); |
523 | ||
ea676e84 | 524 | pr_notice("Kernel command line: %s\n", boot_command_line); |
83b519e8 | 525 | parse_early_param(); |
51e158c1 RR |
526 | after_dashes = parse_args("Booting kernel", |
527 | static_command_line, __start___param, | |
528 | __stop___param - __start___param, | |
ecc86170 | 529 | -1, -1, NULL, &unknown_bootoption); |
3438cf54 | 530 | if (!IS_ERR_OR_NULL(after_dashes)) |
51e158c1 | 531 | parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, |
ecc86170 | 532 | NULL, set_init_arg); |
97ce2c88 JF |
533 | |
534 | jump_label_init(); | |
535 | ||
83b519e8 PE |
536 | /* |
537 | * These use large bootmem allocations and must precede | |
538 | * kmem_cache_init() | |
539 | */ | |
162a7e75 | 540 | setup_log_buf(0); |
83b519e8 | 541 | pidhash_init(); |
83b519e8 PE |
542 | vfs_caches_init_early(); |
543 | sort_main_extable(); | |
544 | trap_init(); | |
444f478f | 545 | mm_init(); |
de03c72c | 546 | |
1da177e4 LT |
547 | /* |
548 | * Set up the scheduler prior starting any interrupts (such as the | |
549 | * timer interrupt). Full topology setup happens at smp_init() | |
550 | * time - but meanwhile we still have a functioning scheduler. | |
551 | */ | |
552 | sched_init(); | |
553 | /* | |
554 | * Disable preemption - early bootup scheduling is extremely | |
555 | * fragile until we cpu_idle() for the first time. | |
556 | */ | |
557 | preempt_disable(); | |
dd4d9fec FF |
558 | if (WARN(!irqs_disabled(), |
559 | "Interrupts were enabled *very* early, fixing it\n")) | |
c4a68306 | 560 | local_irq_disable(); |
9f58a205 | 561 | idr_init_cache(); |
1da177e4 | 562 | rcu_init(); |
5f893b26 SRRH |
563 | |
564 | /* trace_printk() and trace points may be used after this */ | |
565 | trace_init(); | |
566 | ||
65f382fd | 567 | context_tracking_init(); |
773e3eb7 | 568 | radix_tree_init(); |
0b8f1efa YL |
569 | /* init some links before init_ISA_irqs() */ |
570 | early_irq_init(); | |
1da177e4 | 571 | init_IRQ(); |
ad2b1353 | 572 | tick_init(); |
d6dd50e0 | 573 | rcu_init_nohz(); |
1da177e4 | 574 | init_timers(); |
c0a31329 | 575 | hrtimers_init(); |
1da177e4 | 576 | softirq_init(); |
ad596171 | 577 | timekeeping_init(); |
88fecaa2 | 578 | time_init(); |
38ff87f7 | 579 | sched_clock_postinit(); |
9e630205 | 580 | perf_event_init(); |
93e02814 | 581 | profile_init(); |
d8ad7d11 | 582 | call_function_init(); |
f91eb62f | 583 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); |
2ce802f6 | 584 | early_boot_irqs_disabled = false; |
93e02814 | 585 | local_irq_enable(); |
dcce284a | 586 | |
7e85ee0c | 587 | kmem_cache_init_late(); |
1da177e4 LT |
588 | |
589 | /* | |
590 | * HACK ALERT! This is early. We're enabling the console before | |
591 | * we've done PCI setups etc, and console_init() must be aware of | |
592 | * this. But we do want output early, in case something goes wrong. | |
593 | */ | |
594 | console_init(); | |
595 | if (panic_later) | |
499a4584 TH |
596 | panic("Too many boot %s vars at `%s'", panic_later, |
597 | panic_param); | |
fbb9ce95 IM |
598 | |
599 | lockdep_info(); | |
600 | ||
9a11b49a IM |
601 | /* |
602 | * Need to run this when irqs are enabled, because it wants | |
603 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
604 | * too: | |
605 | */ | |
606 | locking_selftest(); | |
607 | ||
1da177e4 LT |
608 | #ifdef CONFIG_BLK_DEV_INITRD |
609 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 610 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
ea676e84 | 611 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", |
bd673c7c GU |
612 | page_to_pfn(virt_to_page((void *)initrd_start)), |
613 | min_low_pfn); | |
1da177e4 LT |
614 | initrd_start = 0; |
615 | } | |
616 | #endif | |
eefa864b | 617 | page_ext_init(); |
3ac7fe5a | 618 | debug_objects_mem_init(); |
9b090f2d | 619 | kmemleak_init(); |
e7c8d5c9 | 620 | setup_per_cpu_pageset(); |
1da177e4 LT |
621 | numa_policy_init(); |
622 | if (late_time_init) | |
623 | late_time_init(); | |
fa84e9ee | 624 | sched_clock_init(); |
1da177e4 LT |
625 | calibrate_delay(); |
626 | pidmap_init(); | |
1da177e4 | 627 | anon_vma_init(); |
c4e1acbb | 628 | acpi_early_init(); |
11520e5e | 629 | #ifdef CONFIG_X86 |
83e68189 | 630 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
11520e5e | 631 | efi_enter_virtual_mode(); |
3891a04a | 632 | #endif |
197725de | 633 | #ifdef CONFIG_X86_ESPFIX64 |
3891a04a PA |
634 | /* Should be run before the first non-init thread is created */ |
635 | init_espfix_bsp(); | |
11520e5e | 636 | #endif |
8c9843e5 | 637 | thread_info_cache_init(); |
d84f4f99 | 638 | cred_init(); |
ff691f6e | 639 | fork_init(); |
1da177e4 LT |
640 | proc_caches_init(); |
641 | buffer_init(); | |
1da177e4 LT |
642 | key_init(); |
643 | security_init(); | |
0b4b3827 | 644 | dbg_late_init(); |
4248b0da | 645 | vfs_caches_init(); |
1da177e4 LT |
646 | signals_init(); |
647 | /* rootfs populating might need page-writeback */ | |
648 | page_writeback_init(); | |
1da177e4 | 649 | proc_root_init(); |
e149ed2b | 650 | nsfs_init(); |
1da177e4 | 651 | cpuset_init(); |
695df213 | 652 | cgroup_init(); |
c757249a | 653 | taskstats_init_early(); |
ca74e92b | 654 | delayacct_init(); |
1da177e4 LT |
655 | |
656 | check_bugs(); | |
657 | ||
b064a8fa | 658 | acpi_subsystem_init(); |
6ae6996a | 659 | sfi_init_late(); |
1da177e4 | 660 | |
83e68189 | 661 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
2223af38 | 662 | efi_late_init(); |
78510792 | 663 | efi_free_boot_services(); |
2223af38 | 664 | } |
78510792 | 665 | |
68bf21aa SR |
666 | ftrace_init(); |
667 | ||
1da177e4 LT |
668 | /* Do the rest non-__init'ed, we're now alive */ |
669 | rest_init(); | |
670 | } | |
671 | ||
b99b87f7 PO |
672 | /* Call all constructor functions linked into the kernel. */ |
673 | static void __init do_ctors(void) | |
674 | { | |
675 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 676 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 677 | |
196a15b4 HS |
678 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
679 | (*fn)(); | |
b99b87f7 PO |
680 | #endif |
681 | } | |
682 | ||
2329abfa | 683 | bool initcall_debug; |
d0ea3d7d | 684 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 685 | |
7b0b73d7 PB |
686 | #ifdef CONFIG_KALLSYMS |
687 | struct blacklist_entry { | |
688 | struct list_head next; | |
689 | char *buf; | |
690 | }; | |
691 | ||
692 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); | |
693 | ||
694 | static int __init initcall_blacklist(char *str) | |
695 | { | |
696 | char *str_entry; | |
697 | struct blacklist_entry *entry; | |
698 | ||
699 | /* str argument is a comma-separated list of functions */ | |
700 | do { | |
701 | str_entry = strsep(&str, ","); | |
702 | if (str_entry) { | |
703 | pr_debug("blacklisting initcall %s\n", str_entry); | |
704 | entry = alloc_bootmem(sizeof(*entry)); | |
705 | entry->buf = alloc_bootmem(strlen(str_entry) + 1); | |
706 | strcpy(entry->buf, str_entry); | |
707 | list_add(&entry->next, &blacklisted_initcalls); | |
708 | } | |
709 | } while (str_entry); | |
710 | ||
711 | return 0; | |
712 | } | |
713 | ||
714 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
715 | { | |
716 | struct list_head *tmp; | |
717 | struct blacklist_entry *entry; | |
718 | char *fn_name; | |
719 | ||
720 | fn_name = kasprintf(GFP_KERNEL, "%pf", fn); | |
721 | if (!fn_name) | |
722 | return false; | |
723 | ||
724 | list_for_each(tmp, &blacklisted_initcalls) { | |
725 | entry = list_entry(tmp, struct blacklist_entry, next); | |
726 | if (!strcmp(fn_name, entry->buf)) { | |
727 | pr_debug("initcall %s blacklisted\n", fn_name); | |
728 | kfree(fn_name); | |
729 | return true; | |
730 | } | |
731 | } | |
732 | ||
733 | kfree(fn_name); | |
734 | return false; | |
735 | } | |
736 | #else | |
737 | static int __init initcall_blacklist(char *str) | |
738 | { | |
739 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); | |
740 | return 0; | |
741 | } | |
742 | ||
743 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
744 | { | |
745 | return false; | |
746 | } | |
747 | #endif | |
748 | __setup("initcall_blacklist=", initcall_blacklist); | |
749 | ||
e4461271 | 750 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 751 | { |
74239072 | 752 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
753 | unsigned long long duration; |
754 | int ret; | |
1da177e4 | 755 | |
d62cf815 | 756 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
22c5c03b | 757 | calltime = ktime_get(); |
30dbb20e | 758 | ret = fn(); |
22c5c03b KW |
759 | rettime = ktime_get(); |
760 | delta = ktime_sub(rettime, calltime); | |
761 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
d62cf815 | 762 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", |
ea676e84 | 763 | fn, ret, duration); |
1da177e4 | 764 | |
22c5c03b KW |
765 | return ret; |
766 | } | |
767 | ||
e4461271 | 768 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
769 | { |
770 | int count = preempt_count(); | |
771 | int ret; | |
ff1c8fac | 772 | char msgbuf[64]; |
22c5c03b | 773 | |
7b0b73d7 PB |
774 | if (initcall_blacklisted(fn)) |
775 | return -EPERM; | |
776 | ||
22c5c03b KW |
777 | if (initcall_debug) |
778 | ret = do_one_initcall_debug(fn); | |
779 | else | |
780 | ret = fn(); | |
8f0c45cd | 781 | |
e0df154f | 782 | msgbuf[0] = 0; |
e662e1cf | 783 | |
e0df154f | 784 | if (preempt_count() != count) { |
bf5d770b | 785 | sprintf(msgbuf, "preemption imbalance "); |
4a2b4b22 | 786 | preempt_count_set(count); |
1da177e4 | 787 | } |
e0df154f | 788 | if (irqs_disabled()) { |
a76bfd0d | 789 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
790 | local_irq_enable(); |
791 | } | |
f91eb62f | 792 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
59f9415f | 793 | |
30dbb20e | 794 | return ret; |
e0df154f LT |
795 | } |
796 | ||
797 | ||
026cee00 PM |
798 | extern initcall_t __initcall_start[]; |
799 | extern initcall_t __initcall0_start[]; | |
800 | extern initcall_t __initcall1_start[]; | |
801 | extern initcall_t __initcall2_start[]; | |
802 | extern initcall_t __initcall3_start[]; | |
803 | extern initcall_t __initcall4_start[]; | |
804 | extern initcall_t __initcall5_start[]; | |
805 | extern initcall_t __initcall6_start[]; | |
806 | extern initcall_t __initcall7_start[]; | |
807 | extern initcall_t __initcall_end[]; | |
808 | ||
809 | static initcall_t *initcall_levels[] __initdata = { | |
810 | __initcall0_start, | |
811 | __initcall1_start, | |
812 | __initcall2_start, | |
813 | __initcall3_start, | |
814 | __initcall4_start, | |
815 | __initcall5_start, | |
816 | __initcall6_start, | |
817 | __initcall7_start, | |
818 | __initcall_end, | |
819 | }; | |
820 | ||
96263d28 | 821 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 822 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
823 | "early", |
824 | "core", | |
825 | "postcore", | |
826 | "arch", | |
827 | "subsys", | |
828 | "fs", | |
829 | "device", | |
830 | "late", | |
026cee00 PM |
831 | }; |
832 | ||
026cee00 | 833 | static void __init do_initcall_level(int level) |
e0df154f | 834 | { |
196a15b4 | 835 | initcall_t *fn; |
e0df154f | 836 | |
08746a65 | 837 | strcpy(initcall_command_line, saved_command_line); |
026cee00 | 838 | parse_args(initcall_level_names[level], |
08746a65 | 839 | initcall_command_line, __start___param, |
026cee00 PM |
840 | __stop___param - __start___param, |
841 | level, level, | |
ecc86170 | 842 | NULL, &repair_env_string); |
026cee00 PM |
843 | |
844 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 845 | do_one_initcall(*fn); |
1da177e4 LT |
846 | } |
847 | ||
026cee00 PM |
848 | static void __init do_initcalls(void) |
849 | { | |
850 | int level; | |
851 | ||
19efb72f | 852 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
853 | do_initcall_level(level); |
854 | } | |
855 | ||
1da177e4 LT |
856 | /* |
857 | * Ok, the machine is now initialized. None of the devices | |
858 | * have been touched yet, but the CPU subsystem is up and | |
859 | * running, and memory and process management works. | |
860 | * | |
861 | * Now we can finally start doing some real work.. | |
862 | */ | |
863 | static void __init do_basic_setup(void) | |
864 | { | |
759ee091 | 865 | cpuset_init_smp(); |
41ffe5d5 | 866 | shmem_init(); |
1da177e4 | 867 | driver_init(); |
b04c3afb | 868 | init_irq_proc(); |
b99b87f7 | 869 | do_ctors(); |
d5767c53 | 870 | usermodehelper_enable(); |
b0f84374 | 871 | do_initcalls(); |
47d06e53 | 872 | random_int_secret_init(); |
1da177e4 LT |
873 | } |
874 | ||
7babe8db | 875 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 876 | { |
196a15b4 | 877 | initcall_t *fn; |
c2147a50 | 878 | |
026cee00 | 879 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 880 | do_one_initcall(*fn); |
c2147a50 EGM |
881 | } |
882 | ||
bb813f4c TH |
883 | /* |
884 | * This function requests modules which should be loaded by default and is | |
885 | * called twice right after initrd is mounted and right before init is | |
886 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
887 | * loaded before control is passed to userland. | |
888 | */ | |
889 | void __init load_default_modules(void) | |
890 | { | |
891 | load_default_elevator_module(); | |
892 | } | |
893 | ||
a74fb73c | 894 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
895 | { |
896 | argv_init[0] = init_filename; | |
c4ad8f98 | 897 | return do_execve(getname_kernel(init_filename), |
ae903caa AV |
898 | (const char __user *const __user *)argv_init, |
899 | (const char __user *const __user *)envp_init); | |
1da177e4 LT |
900 | } |
901 | ||
ba24762b MO |
902 | static int try_to_run_init_process(const char *init_filename) |
903 | { | |
904 | int ret; | |
905 | ||
906 | ret = run_init_process(init_filename); | |
907 | ||
908 | if (ret && ret != -ENOENT) { | |
909 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
910 | init_filename, ret); | |
911 | } | |
912 | ||
913 | return ret; | |
914 | } | |
915 | ||
f80b0c90 | 916 | static noinline void __init kernel_init_freeable(void); |
d6b21238 AV |
917 | |
918 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 919 | { |
ba24762b MO |
920 | int ret; |
921 | ||
d6b21238 | 922 | kernel_init_freeable(); |
22a9d645 AV |
923 | /* need to finish all async __init code before freeing the memory */ |
924 | async_synchronize_full(); | |
ee5bfa64 | 925 | free_initmem(); |
ee5bfa64 VG |
926 | mark_rodata_ro(); |
927 | system_state = SYSTEM_RUNNING; | |
928 | numa_default_policy(); | |
929 | ||
4a9d4b02 | 930 | flush_delayed_fput(); |
fae5fa44 | 931 | |
967dcb8f PM |
932 | rcu_end_inkernel_boot(); |
933 | ||
ee5bfa64 | 934 | if (ramdisk_execute_command) { |
ba24762b MO |
935 | ret = run_init_process(ramdisk_execute_command); |
936 | if (!ret) | |
a74fb73c | 937 | return 0; |
ba24762b MO |
938 | pr_err("Failed to execute %s (error %d)\n", |
939 | ramdisk_execute_command, ret); | |
ee5bfa64 VG |
940 | } |
941 | ||
942 | /* | |
943 | * We try each of these until one succeeds. | |
944 | * | |
945 | * The Bourne shell can be used instead of init if we are | |
946 | * trying to recover a really broken machine. | |
947 | */ | |
948 | if (execute_command) { | |
ba24762b MO |
949 | ret = run_init_process(execute_command); |
950 | if (!ret) | |
a74fb73c | 951 | return 0; |
6ef4536e AL |
952 | panic("Requested init %s failed (error %d).", |
953 | execute_command, ret); | |
ee5bfa64 | 954 | } |
ba24762b MO |
955 | if (!try_to_run_init_process("/sbin/init") || |
956 | !try_to_run_init_process("/etc/init") || | |
957 | !try_to_run_init_process("/bin/init") || | |
958 | !try_to_run_init_process("/bin/sh")) | |
a74fb73c | 959 | return 0; |
ee5bfa64 | 960 | |
ba24762b | 961 | panic("No working init found. Try passing init= option to kernel. " |
9a85b8d6 | 962 | "See Linux Documentation/init.txt for guidance."); |
ee5bfa64 VG |
963 | } |
964 | ||
f80b0c90 | 965 | static noinline void __init kernel_init_freeable(void) |
1da177e4 | 966 | { |
b433c3d4 PZ |
967 | /* |
968 | * Wait until kthreadd is all set-up. | |
969 | */ | |
970 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
971 | |
972 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
973 | gfp_allowed_mask = __GFP_BITS_MASK; | |
974 | ||
58568d2a MX |
975 | /* |
976 | * init can allocate pages on any node | |
977 | */ | |
3c466d46 | 978 | set_mems_allowed(node_states[N_MEMORY]); |
1da177e4 LT |
979 | /* |
980 | * init can run on any cpu. | |
981 | */ | |
1a2142af | 982 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 983 | |
9ec52099 CLG |
984 | cad_pid = task_pid(current); |
985 | ||
ca74a6f8 | 986 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
987 | |
988 | do_pre_smp_initcalls(); | |
004417a6 | 989 | lockup_detector_init(); |
1da177e4 | 990 | |
1da177e4 LT |
991 | smp_init(); |
992 | sched_init_smp(); | |
993 | ||
0e1cc95b MG |
994 | page_alloc_init_late(); |
995 | ||
1da177e4 LT |
996 | do_basic_setup(); |
997 | ||
2bd3a997 EB |
998 | /* Open the /dev/console on the rootfs, this should never fail */ |
999 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
ea676e84 | 1000 | pr_err("Warning: unable to open an initial console.\n"); |
2bd3a997 EB |
1001 | |
1002 | (void) sys_dup(0); | |
1003 | (void) sys_dup(0); | |
1da177e4 LT |
1004 | /* |
1005 | * check if there is an early userspace init. If yes, let it do all | |
1006 | * the work | |
1007 | */ | |
ffdfc409 OJ |
1008 | |
1009 | if (!ramdisk_execute_command) | |
1010 | ramdisk_execute_command = "/init"; | |
1011 | ||
1012 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
1013 | ramdisk_execute_command = NULL; | |
1da177e4 | 1014 | prepare_namespace(); |
ffdfc409 | 1015 | } |
1da177e4 LT |
1016 | |
1017 | /* | |
1018 | * Ok, we have completed the initial bootup, and | |
1019 | * we're essentially up and running. Get rid of the | |
1020 | * initmem segments and start the user-mode stuff.. | |
c9cd2ce2 DK |
1021 | * |
1022 | * rootfs is available now, try loading the public keys | |
1023 | * and default modules | |
1da177e4 | 1024 | */ |
bb813f4c | 1025 | |
c9cd2ce2 | 1026 | integrity_load_keys(); |
bb813f4c | 1027 | load_default_modules(); |
1da177e4 | 1028 | } |