]>
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> |
9a11b49a | 54 | #include <linux/debug_locks.h> |
3ac7fe5a | 55 | #include <linux/debugobjects.h> |
fbb9ce95 | 56 | #include <linux/lockdep.h> |
3c7b4e6b | 57 | #include <linux/kmemleak.h> |
84d73786 | 58 | #include <linux/pid_namespace.h> |
1f21782e | 59 | #include <linux/device.h> |
73c27992 | 60 | #include <linux/kthread.h> |
e6fe6649 | 61 | #include <linux/sched.h> |
a1c9eea9 | 62 | #include <linux/signal.h> |
199f0ca5 | 63 | #include <linux/idr.h> |
0b4b3827 | 64 | #include <linux/kgdb.h> |
68bf21aa | 65 | #include <linux/ftrace.h> |
22a9d645 | 66 | #include <linux/async.h> |
dfec072e | 67 | #include <linux/kmemcheck.h> |
6ae6996a | 68 | #include <linux/sfi.h> |
2b2af54a | 69 | #include <linux/shmem_fs.h> |
5a0e3ad6 | 70 | #include <linux/slab.h> |
24a24bb6 | 71 | #include <linux/perf_event.h> |
4a9d4b02 | 72 | #include <linux/file.h> |
a74fb73c | 73 | #include <linux/ptrace.h> |
bb813f4c TH |
74 | #include <linux/blkdev.h> |
75 | #include <linux/elevator.h> | |
38ff87f7 | 76 | #include <linux/sched_clock.h> |
65f382fd | 77 | #include <linux/context_tracking.h> |
47d06e53 | 78 | #include <linux/random.h> |
7b0b73d7 | 79 | #include <linux/list.h> |
1da177e4 LT |
80 | |
81 | #include <asm/io.h> | |
82 | #include <asm/bugs.h> | |
83 | #include <asm/setup.h> | |
a940199f | 84 | #include <asm/sections.h> |
37b73c82 | 85 | #include <asm/cacheflush.h> |
1da177e4 | 86 | |
1da177e4 LT |
87 | #ifdef CONFIG_X86_LOCAL_APIC |
88 | #include <asm/smp.h> | |
89 | #endif | |
90 | ||
aae5f662 | 91 | static int kernel_init(void *); |
1da177e4 LT |
92 | |
93 | extern void init_IRQ(void); | |
1da177e4 | 94 | extern void fork_init(unsigned long); |
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. | |
147 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
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 LT |
166 | |
167 | static int __init obsolete_checksetup(char *line) | |
168 | { | |
914dcaa8 | 169 | const struct obs_kernel_param *p; |
33df0d19 | 170 | int had_early_param = 0; |
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] == '=') |
33df0d19 | 182 | had_early_param = 1; |
1da177e4 | 183 | } else if (!p->setup_func) { |
ea676e84 AM |
184 | pr_warn("Parameter %s is obsolete, ignored\n", |
185 | p->str); | |
1da177e4 LT |
186 | return 1; |
187 | } else if (p->setup_func(line + n)) | |
188 | return 1; | |
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. */ |
eb157427 | 238 | static int __init repair_env_string(char *param, char *val, const char *unused) |
1da177e4 | 239 | { |
1da177e4 LT |
240 | if (val) { |
241 | /* param=val or param="val"? */ | |
242 | if (val == param+strlen(param)+1) | |
243 | val[-1] = '='; | |
244 | else if (val == param+strlen(param)+2) { | |
245 | val[-2] = '='; | |
246 | memmove(val-1, val, strlen(val)+1); | |
247 | val--; | |
248 | } else | |
249 | BUG(); | |
250 | } | |
a99cd112 CM |
251 | return 0; |
252 | } | |
253 | ||
51e158c1 RR |
254 | /* Anything after -- gets handed straight to init. */ |
255 | static int __init set_init_arg(char *param, char *val, const char *unused) | |
256 | { | |
257 | unsigned int i; | |
258 | ||
259 | if (panic_later) | |
260 | return 0; | |
261 | ||
262 | repair_env_string(param, val, unused); | |
263 | ||
264 | for (i = 0; argv_init[i]; i++) { | |
265 | if (i == MAX_INIT_ARGS) { | |
266 | panic_later = "init"; | |
267 | panic_param = param; | |
268 | return 0; | |
269 | } | |
270 | } | |
271 | argv_init[i] = param; | |
272 | return 0; | |
273 | } | |
274 | ||
a99cd112 CM |
275 | /* |
276 | * Unknown boot options get handed to init, unless they look like | |
277 | * unused parameters (modprobe will find them in /proc/cmdline). | |
278 | */ | |
eb157427 | 279 | static int __init unknown_bootoption(char *param, char *val, const char *unused) |
a99cd112 | 280 | { |
eb157427 | 281 | repair_env_string(param, val, unused); |
1da177e4 LT |
282 | |
283 | /* Handle obsolete-style parameters */ | |
284 | if (obsolete_checksetup(param)) | |
285 | return 0; | |
286 | ||
f066a4f6 RR |
287 | /* Unused module parameter. */ |
288 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 289 | return 0; |
1da177e4 LT |
290 | |
291 | if (panic_later) | |
292 | return 0; | |
293 | ||
294 | if (val) { | |
295 | /* Environment option */ | |
296 | unsigned int i; | |
297 | for (i = 0; envp_init[i]; i++) { | |
298 | if (i == MAX_INIT_ENVS) { | |
499a4584 | 299 | panic_later = "env"; |
1da177e4 LT |
300 | panic_param = param; |
301 | } | |
302 | if (!strncmp(param, envp_init[i], val - param)) | |
303 | break; | |
304 | } | |
305 | envp_init[i] = param; | |
306 | } else { | |
307 | /* Command line option */ | |
308 | unsigned int i; | |
309 | for (i = 0; argv_init[i]; i++) { | |
310 | if (i == MAX_INIT_ARGS) { | |
499a4584 | 311 | panic_later = "init"; |
1da177e4 LT |
312 | panic_param = param; |
313 | } | |
314 | } | |
315 | argv_init[i] = param; | |
316 | } | |
317 | return 0; | |
318 | } | |
319 | ||
320 | static int __init init_setup(char *str) | |
321 | { | |
322 | unsigned int i; | |
323 | ||
324 | execute_command = str; | |
325 | /* | |
326 | * In case LILO is going to boot us with default command line, | |
327 | * it prepends "auto" before the whole cmdline which makes | |
328 | * the shell think it should execute a script with such name. | |
329 | * So we ignore all arguments entered _before_ init=... [MJ] | |
330 | */ | |
331 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
332 | argv_init[i] = NULL; | |
333 | return 1; | |
334 | } | |
335 | __setup("init=", init_setup); | |
336 | ||
ffdfc409 OJ |
337 | static int __init rdinit_setup(char *str) |
338 | { | |
339 | unsigned int i; | |
340 | ||
341 | ramdisk_execute_command = str; | |
342 | /* See "auto" comment in init_setup */ | |
343 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
344 | argv_init[i] = NULL; | |
345 | return 1; | |
346 | } | |
347 | __setup("rdinit=", rdinit_setup); | |
348 | ||
1da177e4 | 349 | #ifndef CONFIG_SMP |
34db18a0 | 350 | static const unsigned int setup_max_cpus = NR_CPUS; |
1da177e4 LT |
351 | #ifdef CONFIG_X86_LOCAL_APIC |
352 | static void __init smp_init(void) | |
353 | { | |
354 | APIC_init_uniprocessor(); | |
355 | } | |
356 | #else | |
357 | #define smp_init() do { } while (0) | |
358 | #endif | |
359 | ||
e0982e90 | 360 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 361 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
362 | #endif |
363 | ||
30d7e0d4 ABL |
364 | /* |
365 | * We need to store the untouched command line for future reference. | |
366 | * We also need to store the touched command line since the parameter | |
367 | * parsing is performed in place, and we should allow a component to | |
368 | * store reference of name/value for future reference. | |
369 | */ | |
370 | static void __init setup_command_line(char *command_line) | |
371 | { | |
098b081b SS |
372 | saved_command_line = |
373 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
374 | initcall_command_line = | |
375 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
376 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
dd4d9fec FF |
377 | strcpy(saved_command_line, boot_command_line); |
378 | strcpy(static_command_line, command_line); | |
30d7e0d4 ABL |
379 | } |
380 | ||
1da177e4 LT |
381 | /* |
382 | * We need to finalize in a non-__init function or else race conditions | |
383 | * between the root thread and the init thread may cause start_kernel to | |
384 | * be reaped by free_initmem before the root thread has proceeded to | |
385 | * cpu_idle. | |
386 | * | |
387 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
388 | */ | |
389 | ||
b433c3d4 PZ |
390 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
391 | ||
f99ebf0a | 392 | static noinline void __init_refok rest_init(void) |
1da177e4 | 393 | { |
73c27992 EB |
394 | int pid; |
395 | ||
7db905e6 | 396 | rcu_scheduler_starting(); |
b433c3d4 | 397 | /* |
97158569 | 398 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
399 | * the init task will end up wanting to create kthreads, which, if |
400 | * we schedule it before we create kthreadd, will OOPS. | |
401 | */ | |
34a1b723 | 402 | kernel_thread(kernel_init, NULL, CLONE_FS); |
1da177e4 | 403 | numa_default_policy(); |
73c27992 | 404 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 405 | rcu_read_lock(); |
5cd20455 | 406 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 407 | rcu_read_unlock(); |
b433c3d4 | 408 | complete(&kthreadd_done); |
f340c0d1 IM |
409 | |
410 | /* | |
411 | * The boot idle thread must execute schedule() | |
1df21055 | 412 | * at least once to get things moving: |
f340c0d1 | 413 | */ |
1df21055 | 414 | init_idle_bootup_task(current); |
bd2f5536 | 415 | schedule_preempt_disabled(); |
5bfb5d69 | 416 | /* Call into cpu_idle with preempt disabled */ |
a1a04ec3 | 417 | cpu_startup_entry(CPUHP_ONLINE); |
1df21055 | 418 | } |
1da177e4 LT |
419 | |
420 | /* Check for early params. */ | |
9fb48c74 | 421 | static int __init do_early_param(char *param, char *val, const char *unused) |
1da177e4 | 422 | { |
914dcaa8 | 423 | const struct obs_kernel_param *p; |
1da177e4 LT |
424 | |
425 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 426 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
427 | (strcmp(param, "console") == 0 && |
428 | strcmp(p->str, "earlycon") == 0) | |
429 | ) { | |
1da177e4 | 430 | if (p->setup_func(val) != 0) |
ea676e84 | 431 | pr_warn("Malformed early option '%s'\n", param); |
1da177e4 LT |
432 | } |
433 | } | |
434 | /* We accept everything at this stage. */ | |
435 | return 0; | |
436 | } | |
437 | ||
13977091 MD |
438 | void __init parse_early_options(char *cmdline) |
439 | { | |
026cee00 | 440 | parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); |
13977091 MD |
441 | } |
442 | ||
1da177e4 LT |
443 | /* Arch code calls this early on, or if not, just before other parsing. */ |
444 | void __init parse_early_param(void) | |
445 | { | |
dd4d9fec FF |
446 | static int done __initdata; |
447 | static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata; | |
1da177e4 LT |
448 | |
449 | if (done) | |
450 | return; | |
451 | ||
452 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 453 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 454 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
455 | done = 1; |
456 | } | |
457 | ||
458 | /* | |
459 | * Activate the first processor. | |
460 | */ | |
461 | ||
44fd2299 SS |
462 | static void __init boot_cpu_init(void) |
463 | { | |
464 | int cpu = smp_processor_id(); | |
465 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 466 | set_cpu_online(cpu, true); |
933b0618 | 467 | set_cpu_active(cpu, true); |
915441b6 RR |
468 | set_cpu_present(cpu, true); |
469 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
470 | } |
471 | ||
839ad62e | 472 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
473 | { |
474 | } | |
475 | ||
eded09cc | 476 | # if THREAD_SIZE >= PAGE_SIZE |
8c9843e5 BH |
477 | void __init __weak thread_info_cache_init(void) |
478 | { | |
479 | } | |
eded09cc | 480 | #endif |
8c9843e5 | 481 | |
444f478f PE |
482 | /* |
483 | * Set up kernel memory allocators | |
484 | */ | |
485 | static void __init mm_init(void) | |
486 | { | |
487 | mem_init(); | |
488 | kmem_cache_init(); | |
099a19d9 | 489 | percpu_init_late(); |
b35f1819 | 490 | pgtable_init(); |
444f478f PE |
491 | vmalloc_init(); |
492 | } | |
493 | ||
722a9f92 | 494 | asmlinkage __visible void __init start_kernel(void) |
1da177e4 | 495 | { |
dd4d9fec FF |
496 | char *command_line; |
497 | char *after_dashes; | |
033ab7f8 | 498 | |
fbb9ce95 IM |
499 | /* |
500 | * Need to run as early as possible, to initialize the | |
501 | * lockdep hash: | |
502 | */ | |
503 | lockdep_init(); | |
d4311ff1 | 504 | set_task_stack_end_magic(&init_task); |
73839c5b | 505 | smp_setup_processor_id(); |
3ac7fe5a | 506 | debug_objects_early_init(); |
42059429 IM |
507 | |
508 | /* | |
509 | * Set up the the initial canary ASAP: | |
510 | */ | |
511 | boot_init_stack_canary(); | |
512 | ||
ddbcc7e8 | 513 | cgroup_init_early(); |
fbb9ce95 IM |
514 | |
515 | local_irq_disable(); | |
2ce802f6 | 516 | early_boot_irqs_disabled = true; |
fbb9ce95 | 517 | |
1da177e4 LT |
518 | /* |
519 | * Interrupts are still disabled. Do necessary setups, then | |
520 | * enable them | |
521 | */ | |
44fd2299 | 522 | boot_cpu_init(); |
1da177e4 | 523 | page_address_init(); |
ea676e84 | 524 | pr_notice("%s", linux_banner); |
1da177e4 | 525 | setup_arch(&command_line); |
6345d24d | 526 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 527 | setup_command_line(command_line); |
e0982e90 | 528 | setup_nr_cpu_ids(); |
d6647bdf | 529 | setup_per_cpu_areas(); |
44fd2299 | 530 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 531 | |
9adb62a5 | 532 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
533 | page_alloc_init(); |
534 | ||
ea676e84 | 535 | pr_notice("Kernel command line: %s\n", boot_command_line); |
83b519e8 | 536 | parse_early_param(); |
51e158c1 RR |
537 | after_dashes = parse_args("Booting kernel", |
538 | static_command_line, __start___param, | |
539 | __stop___param - __start___param, | |
540 | -1, -1, &unknown_bootoption); | |
3438cf54 | 541 | if (!IS_ERR_OR_NULL(after_dashes)) |
51e158c1 RR |
542 | parse_args("Setting init args", after_dashes, NULL, 0, -1, -1, |
543 | set_init_arg); | |
97ce2c88 JF |
544 | |
545 | jump_label_init(); | |
546 | ||
83b519e8 PE |
547 | /* |
548 | * These use large bootmem allocations and must precede | |
549 | * kmem_cache_init() | |
550 | */ | |
162a7e75 | 551 | setup_log_buf(0); |
83b519e8 | 552 | pidhash_init(); |
83b519e8 PE |
553 | vfs_caches_init_early(); |
554 | sort_main_extable(); | |
555 | trap_init(); | |
444f478f | 556 | mm_init(); |
de03c72c | 557 | |
1da177e4 LT |
558 | /* |
559 | * Set up the scheduler prior starting any interrupts (such as the | |
560 | * timer interrupt). Full topology setup happens at smp_init() | |
561 | * time - but meanwhile we still have a functioning scheduler. | |
562 | */ | |
563 | sched_init(); | |
564 | /* | |
565 | * Disable preemption - early bootup scheduling is extremely | |
566 | * fragile until we cpu_idle() for the first time. | |
567 | */ | |
568 | preempt_disable(); | |
dd4d9fec FF |
569 | if (WARN(!irqs_disabled(), |
570 | "Interrupts were enabled *very* early, fixing it\n")) | |
c4a68306 | 571 | local_irq_disable(); |
9f58a205 | 572 | idr_init_cache(); |
1da177e4 | 573 | rcu_init(); |
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 | |
3ac7fe5a | 624 | debug_objects_mem_init(); |
9b090f2d | 625 | kmemleak_init(); |
e7c8d5c9 | 626 | setup_per_cpu_pageset(); |
1da177e4 LT |
627 | numa_policy_init(); |
628 | if (late_time_init) | |
629 | late_time_init(); | |
fa84e9ee | 630 | sched_clock_init(); |
1da177e4 LT |
631 | calibrate_delay(); |
632 | pidmap_init(); | |
1da177e4 | 633 | anon_vma_init(); |
c4e1acbb | 634 | acpi_early_init(); |
11520e5e | 635 | #ifdef CONFIG_X86 |
83e68189 | 636 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
11520e5e | 637 | efi_enter_virtual_mode(); |
3891a04a | 638 | #endif |
197725de | 639 | #ifdef CONFIG_X86_ESPFIX64 |
3891a04a PA |
640 | /* Should be run before the first non-init thread is created */ |
641 | init_espfix_bsp(); | |
11520e5e | 642 | #endif |
8c9843e5 | 643 | thread_info_cache_init(); |
d84f4f99 | 644 | cred_init(); |
4481374c | 645 | fork_init(totalram_pages); |
1da177e4 LT |
646 | proc_caches_init(); |
647 | buffer_init(); | |
1da177e4 LT |
648 | key_init(); |
649 | security_init(); | |
0b4b3827 | 650 | dbg_late_init(); |
4481374c | 651 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
652 | signals_init(); |
653 | /* rootfs populating might need page-writeback */ | |
654 | page_writeback_init(); | |
1da177e4 | 655 | proc_root_init(); |
ddbcc7e8 | 656 | cgroup_init(); |
1da177e4 | 657 | cpuset_init(); |
c757249a | 658 | taskstats_init_early(); |
ca74e92b | 659 | delayacct_init(); |
1da177e4 LT |
660 | |
661 | check_bugs(); | |
662 | ||
6ae6996a | 663 | sfi_init_late(); |
1da177e4 | 664 | |
83e68189 | 665 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
2223af38 | 666 | efi_late_init(); |
78510792 | 667 | efi_free_boot_services(); |
2223af38 | 668 | } |
78510792 | 669 | |
68bf21aa SR |
670 | ftrace_init(); |
671 | ||
1da177e4 LT |
672 | /* Do the rest non-__init'ed, we're now alive */ |
673 | rest_init(); | |
674 | } | |
675 | ||
b99b87f7 PO |
676 | /* Call all constructor functions linked into the kernel. */ |
677 | static void __init do_ctors(void) | |
678 | { | |
679 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 680 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 681 | |
196a15b4 HS |
682 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
683 | (*fn)(); | |
b99b87f7 PO |
684 | #endif |
685 | } | |
686 | ||
2329abfa | 687 | bool initcall_debug; |
d0ea3d7d | 688 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 689 | |
7b0b73d7 PB |
690 | #ifdef CONFIG_KALLSYMS |
691 | struct blacklist_entry { | |
692 | struct list_head next; | |
693 | char *buf; | |
694 | }; | |
695 | ||
696 | static __initdata_or_module LIST_HEAD(blacklisted_initcalls); | |
697 | ||
698 | static int __init initcall_blacklist(char *str) | |
699 | { | |
700 | char *str_entry; | |
701 | struct blacklist_entry *entry; | |
702 | ||
703 | /* str argument is a comma-separated list of functions */ | |
704 | do { | |
705 | str_entry = strsep(&str, ","); | |
706 | if (str_entry) { | |
707 | pr_debug("blacklisting initcall %s\n", str_entry); | |
708 | entry = alloc_bootmem(sizeof(*entry)); | |
709 | entry->buf = alloc_bootmem(strlen(str_entry) + 1); | |
710 | strcpy(entry->buf, str_entry); | |
711 | list_add(&entry->next, &blacklisted_initcalls); | |
712 | } | |
713 | } while (str_entry); | |
714 | ||
715 | return 0; | |
716 | } | |
717 | ||
718 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
719 | { | |
720 | struct list_head *tmp; | |
721 | struct blacklist_entry *entry; | |
722 | char *fn_name; | |
723 | ||
724 | fn_name = kasprintf(GFP_KERNEL, "%pf", fn); | |
725 | if (!fn_name) | |
726 | return false; | |
727 | ||
728 | list_for_each(tmp, &blacklisted_initcalls) { | |
729 | entry = list_entry(tmp, struct blacklist_entry, next); | |
730 | if (!strcmp(fn_name, entry->buf)) { | |
731 | pr_debug("initcall %s blacklisted\n", fn_name); | |
732 | kfree(fn_name); | |
733 | return true; | |
734 | } | |
735 | } | |
736 | ||
737 | kfree(fn_name); | |
738 | return false; | |
739 | } | |
740 | #else | |
741 | static int __init initcall_blacklist(char *str) | |
742 | { | |
743 | pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n"); | |
744 | return 0; | |
745 | } | |
746 | ||
747 | static bool __init_or_module initcall_blacklisted(initcall_t fn) | |
748 | { | |
749 | return false; | |
750 | } | |
751 | #endif | |
752 | __setup("initcall_blacklist=", initcall_blacklist); | |
753 | ||
e4461271 | 754 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 755 | { |
74239072 | 756 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
757 | unsigned long long duration; |
758 | int ret; | |
1da177e4 | 759 | |
d62cf815 | 760 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
22c5c03b | 761 | calltime = ktime_get(); |
30dbb20e | 762 | ret = fn(); |
22c5c03b KW |
763 | rettime = ktime_get(); |
764 | delta = ktime_sub(rettime, calltime); | |
765 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
d62cf815 | 766 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", |
ea676e84 | 767 | fn, ret, duration); |
1da177e4 | 768 | |
22c5c03b KW |
769 | return ret; |
770 | } | |
771 | ||
e4461271 | 772 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
773 | { |
774 | int count = preempt_count(); | |
775 | int ret; | |
ff1c8fac | 776 | char msgbuf[64]; |
22c5c03b | 777 | |
7b0b73d7 PB |
778 | if (initcall_blacklisted(fn)) |
779 | return -EPERM; | |
780 | ||
22c5c03b KW |
781 | if (initcall_debug) |
782 | ret = do_one_initcall_debug(fn); | |
783 | else | |
784 | ret = fn(); | |
8f0c45cd | 785 | |
e0df154f | 786 | msgbuf[0] = 0; |
e662e1cf | 787 | |
e0df154f | 788 | if (preempt_count() != count) { |
bf5d770b | 789 | sprintf(msgbuf, "preemption imbalance "); |
4a2b4b22 | 790 | preempt_count_set(count); |
1da177e4 | 791 | } |
e0df154f | 792 | if (irqs_disabled()) { |
a76bfd0d | 793 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
794 | local_irq_enable(); |
795 | } | |
f91eb62f | 796 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
59f9415f | 797 | |
30dbb20e | 798 | return ret; |
e0df154f LT |
799 | } |
800 | ||
801 | ||
026cee00 PM |
802 | extern initcall_t __initcall_start[]; |
803 | extern initcall_t __initcall0_start[]; | |
804 | extern initcall_t __initcall1_start[]; | |
805 | extern initcall_t __initcall2_start[]; | |
806 | extern initcall_t __initcall3_start[]; | |
807 | extern initcall_t __initcall4_start[]; | |
808 | extern initcall_t __initcall5_start[]; | |
809 | extern initcall_t __initcall6_start[]; | |
810 | extern initcall_t __initcall7_start[]; | |
811 | extern initcall_t __initcall_end[]; | |
812 | ||
813 | static initcall_t *initcall_levels[] __initdata = { | |
814 | __initcall0_start, | |
815 | __initcall1_start, | |
816 | __initcall2_start, | |
817 | __initcall3_start, | |
818 | __initcall4_start, | |
819 | __initcall5_start, | |
820 | __initcall6_start, | |
821 | __initcall7_start, | |
822 | __initcall_end, | |
823 | }; | |
824 | ||
96263d28 | 825 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 826 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
827 | "early", |
828 | "core", | |
829 | "postcore", | |
830 | "arch", | |
831 | "subsys", | |
832 | "fs", | |
833 | "device", | |
834 | "late", | |
026cee00 PM |
835 | }; |
836 | ||
026cee00 | 837 | static void __init do_initcall_level(int level) |
e0df154f | 838 | { |
196a15b4 | 839 | initcall_t *fn; |
e0df154f | 840 | |
08746a65 | 841 | strcpy(initcall_command_line, saved_command_line); |
026cee00 | 842 | parse_args(initcall_level_names[level], |
08746a65 | 843 | initcall_command_line, __start___param, |
026cee00 PM |
844 | __stop___param - __start___param, |
845 | level, level, | |
eb157427 | 846 | &repair_env_string); |
026cee00 PM |
847 | |
848 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 849 | do_one_initcall(*fn); |
1da177e4 LT |
850 | } |
851 | ||
026cee00 PM |
852 | static void __init do_initcalls(void) |
853 | { | |
854 | int level; | |
855 | ||
19efb72f | 856 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
857 | do_initcall_level(level); |
858 | } | |
859 | ||
1da177e4 LT |
860 | /* |
861 | * Ok, the machine is now initialized. None of the devices | |
862 | * have been touched yet, but the CPU subsystem is up and | |
863 | * running, and memory and process management works. | |
864 | * | |
865 | * Now we can finally start doing some real work.. | |
866 | */ | |
867 | static void __init do_basic_setup(void) | |
868 | { | |
759ee091 | 869 | cpuset_init_smp(); |
1da177e4 | 870 | usermodehelper_init(); |
41ffe5d5 | 871 | shmem_init(); |
1da177e4 | 872 | driver_init(); |
b04c3afb | 873 | init_irq_proc(); |
b99b87f7 | 874 | do_ctors(); |
d5767c53 | 875 | usermodehelper_enable(); |
b0f84374 | 876 | do_initcalls(); |
47d06e53 | 877 | random_int_secret_init(); |
1da177e4 LT |
878 | } |
879 | ||
7babe8db | 880 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 881 | { |
196a15b4 | 882 | initcall_t *fn; |
c2147a50 | 883 | |
026cee00 | 884 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 885 | do_one_initcall(*fn); |
c2147a50 EGM |
886 | } |
887 | ||
bb813f4c TH |
888 | /* |
889 | * This function requests modules which should be loaded by default and is | |
890 | * called twice right after initrd is mounted and right before init is | |
891 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
892 | * loaded before control is passed to userland. | |
893 | */ | |
894 | void __init load_default_modules(void) | |
895 | { | |
896 | load_default_elevator_module(); | |
897 | } | |
898 | ||
a74fb73c | 899 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
900 | { |
901 | argv_init[0] = init_filename; | |
c4ad8f98 | 902 | return do_execve(getname_kernel(init_filename), |
ae903caa AV |
903 | (const char __user *const __user *)argv_init, |
904 | (const char __user *const __user *)envp_init); | |
1da177e4 LT |
905 | } |
906 | ||
ba24762b MO |
907 | static int try_to_run_init_process(const char *init_filename) |
908 | { | |
909 | int ret; | |
910 | ||
911 | ret = run_init_process(init_filename); | |
912 | ||
913 | if (ret && ret != -ENOENT) { | |
914 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
915 | init_filename, ret); | |
916 | } | |
917 | ||
918 | return ret; | |
919 | } | |
920 | ||
f80b0c90 | 921 | static noinline void __init kernel_init_freeable(void); |
d6b21238 AV |
922 | |
923 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 924 | { |
ba24762b MO |
925 | int ret; |
926 | ||
d6b21238 | 927 | kernel_init_freeable(); |
22a9d645 AV |
928 | /* need to finish all async __init code before freeing the memory */ |
929 | async_synchronize_full(); | |
ee5bfa64 | 930 | free_initmem(); |
ee5bfa64 VG |
931 | mark_rodata_ro(); |
932 | system_state = SYSTEM_RUNNING; | |
933 | numa_default_policy(); | |
934 | ||
4a9d4b02 | 935 | flush_delayed_fput(); |
fae5fa44 | 936 | |
ee5bfa64 | 937 | if (ramdisk_execute_command) { |
ba24762b MO |
938 | ret = run_init_process(ramdisk_execute_command); |
939 | if (!ret) | |
a74fb73c | 940 | return 0; |
ba24762b MO |
941 | pr_err("Failed to execute %s (error %d)\n", |
942 | ramdisk_execute_command, ret); | |
ee5bfa64 VG |
943 | } |
944 | ||
945 | /* | |
946 | * We try each of these until one succeeds. | |
947 | * | |
948 | * The Bourne shell can be used instead of init if we are | |
949 | * trying to recover a really broken machine. | |
950 | */ | |
951 | if (execute_command) { | |
ba24762b MO |
952 | ret = run_init_process(execute_command); |
953 | if (!ret) | |
a74fb73c | 954 | return 0; |
6ef4536e AL |
955 | #ifndef CONFIG_INIT_FALLBACK |
956 | panic("Requested init %s failed (error %d).", | |
957 | execute_command, ret); | |
958 | #else | |
ba24762b | 959 | pr_err("Failed to execute %s (error %d). Attempting defaults...\n", |
6ef4536e AL |
960 | execute_command, ret); |
961 | #endif | |
ee5bfa64 | 962 | } |
ba24762b MO |
963 | if (!try_to_run_init_process("/sbin/init") || |
964 | !try_to_run_init_process("/etc/init") || | |
965 | !try_to_run_init_process("/bin/init") || | |
966 | !try_to_run_init_process("/bin/sh")) | |
a74fb73c | 967 | return 0; |
ee5bfa64 | 968 | |
ba24762b | 969 | panic("No working init found. Try passing init= option to kernel. " |
9a85b8d6 | 970 | "See Linux Documentation/init.txt for guidance."); |
ee5bfa64 VG |
971 | } |
972 | ||
f80b0c90 | 973 | static noinline void __init kernel_init_freeable(void) |
1da177e4 | 974 | { |
b433c3d4 PZ |
975 | /* |
976 | * Wait until kthreadd is all set-up. | |
977 | */ | |
978 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
979 | |
980 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
981 | gfp_allowed_mask = __GFP_BITS_MASK; | |
982 | ||
58568d2a MX |
983 | /* |
984 | * init can allocate pages on any node | |
985 | */ | |
3c466d46 | 986 | set_mems_allowed(node_states[N_MEMORY]); |
1da177e4 LT |
987 | /* |
988 | * init can run on any cpu. | |
989 | */ | |
1a2142af | 990 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 991 | |
9ec52099 CLG |
992 | cad_pid = task_pid(current); |
993 | ||
ca74a6f8 | 994 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
995 | |
996 | do_pre_smp_initcalls(); | |
004417a6 | 997 | lockup_detector_init(); |
1da177e4 | 998 | |
1da177e4 LT |
999 | smp_init(); |
1000 | sched_init_smp(); | |
1001 | ||
1da177e4 LT |
1002 | do_basic_setup(); |
1003 | ||
2bd3a997 EB |
1004 | /* Open the /dev/console on the rootfs, this should never fail */ |
1005 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
ea676e84 | 1006 | pr_err("Warning: unable to open an initial console.\n"); |
2bd3a997 EB |
1007 | |
1008 | (void) sys_dup(0); | |
1009 | (void) sys_dup(0); | |
1da177e4 LT |
1010 | /* |
1011 | * check if there is an early userspace init. If yes, let it do all | |
1012 | * the work | |
1013 | */ | |
ffdfc409 OJ |
1014 | |
1015 | if (!ramdisk_execute_command) | |
1016 | ramdisk_execute_command = "/init"; | |
1017 | ||
1018 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
1019 | ramdisk_execute_command = NULL; | |
1da177e4 | 1020 | prepare_namespace(); |
ffdfc409 | 1021 | } |
1da177e4 LT |
1022 | |
1023 | /* | |
1024 | * Ok, we have completed the initial bootup, and | |
1025 | * we're essentially up and running. Get rid of the | |
1026 | * initmem segments and start the user-mode stuff.. | |
1027 | */ | |
bb813f4c TH |
1028 | |
1029 | /* rootfs is available now, try loading default modules */ | |
1030 | load_default_modules(); | |
1da177e4 | 1031 | } |