]>
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 | |
9 | * Simplified starting of init: Michael A. Griffith <grif@acm.org> | |
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> |
94b6da5a | 54 | #include <linux/page_cgroup.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> |
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 | */ | |
138 | bool static_key_initialized __read_mostly = false; | |
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 | ||
d7627467 DH |
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); | |
201 | ||
202 | EXPORT_SYMBOL(loops_per_jiffy); | |
203 | ||
204 | static int __init debug_kernel(char *str) | |
205 | { | |
a8fe19eb | 206 | console_loglevel = CONSOLE_LOGLEVEL_DEBUG; |
f6f21c81 | 207 | return 0; |
1da177e4 LT |
208 | } |
209 | ||
210 | static int __init quiet_kernel(char *str) | |
211 | { | |
a8fe19eb | 212 | console_loglevel = CONSOLE_LOGLEVEL_QUIET; |
f6f21c81 | 213 | return 0; |
1da177e4 LT |
214 | } |
215 | ||
f6f21c81 YL |
216 | early_param("debug", debug_kernel); |
217 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
218 | |
219 | static int __init loglevel(char *str) | |
220 | { | |
808bf29b AS |
221 | int newlevel; |
222 | ||
223 | /* | |
224 | * Only update loglevel value when a correct setting was passed, | |
225 | * to prevent blind crashes (when loglevel being set to 0) that | |
226 | * are quite hard to debug | |
227 | */ | |
228 | if (get_option(&str, &newlevel)) { | |
229 | console_loglevel = newlevel; | |
230 | return 0; | |
231 | } | |
232 | ||
233 | return -EINVAL; | |
1da177e4 LT |
234 | } |
235 | ||
f6f21c81 | 236 | early_param("loglevel", loglevel); |
1da177e4 | 237 | |
a99cd112 | 238 | /* Change NUL term back to "=", to make "param" the whole string. */ |
eb157427 | 239 | static int __init repair_env_string(char *param, char *val, const char *unused) |
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 | ||
255 | /* | |
256 | * Unknown boot options get handed to init, unless they look like | |
257 | * unused parameters (modprobe will find them in /proc/cmdline). | |
258 | */ | |
eb157427 | 259 | static int __init unknown_bootoption(char *param, char *val, const char *unused) |
a99cd112 | 260 | { |
eb157427 | 261 | repair_env_string(param, val, unused); |
1da177e4 LT |
262 | |
263 | /* Handle obsolete-style parameters */ | |
264 | if (obsolete_checksetup(param)) | |
265 | return 0; | |
266 | ||
f066a4f6 RR |
267 | /* Unused module parameter. */ |
268 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 269 | return 0; |
1da177e4 LT |
270 | |
271 | if (panic_later) | |
272 | return 0; | |
273 | ||
274 | if (val) { | |
275 | /* Environment option */ | |
276 | unsigned int i; | |
277 | for (i = 0; envp_init[i]; i++) { | |
278 | if (i == MAX_INIT_ENVS) { | |
499a4584 | 279 | panic_later = "env"; |
1da177e4 LT |
280 | panic_param = param; |
281 | } | |
282 | if (!strncmp(param, envp_init[i], val - param)) | |
283 | break; | |
284 | } | |
285 | envp_init[i] = param; | |
286 | } else { | |
287 | /* Command line option */ | |
288 | unsigned int i; | |
289 | for (i = 0; argv_init[i]; i++) { | |
290 | if (i == MAX_INIT_ARGS) { | |
499a4584 | 291 | panic_later = "init"; |
1da177e4 LT |
292 | panic_param = param; |
293 | } | |
294 | } | |
295 | argv_init[i] = param; | |
296 | } | |
297 | return 0; | |
298 | } | |
299 | ||
300 | static int __init init_setup(char *str) | |
301 | { | |
302 | unsigned int i; | |
303 | ||
304 | execute_command = str; | |
305 | /* | |
306 | * In case LILO is going to boot us with default command line, | |
307 | * it prepends "auto" before the whole cmdline which makes | |
308 | * the shell think it should execute a script with such name. | |
309 | * So we ignore all arguments entered _before_ init=... [MJ] | |
310 | */ | |
311 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
312 | argv_init[i] = NULL; | |
313 | return 1; | |
314 | } | |
315 | __setup("init=", init_setup); | |
316 | ||
ffdfc409 OJ |
317 | static int __init rdinit_setup(char *str) |
318 | { | |
319 | unsigned int i; | |
320 | ||
321 | ramdisk_execute_command = str; | |
322 | /* See "auto" comment in init_setup */ | |
323 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
324 | argv_init[i] = NULL; | |
325 | return 1; | |
326 | } | |
327 | __setup("rdinit=", rdinit_setup); | |
328 | ||
1da177e4 | 329 | #ifndef CONFIG_SMP |
34db18a0 | 330 | static const unsigned int setup_max_cpus = NR_CPUS; |
1da177e4 LT |
331 | #ifdef CONFIG_X86_LOCAL_APIC |
332 | static void __init smp_init(void) | |
333 | { | |
334 | APIC_init_uniprocessor(); | |
335 | } | |
336 | #else | |
337 | #define smp_init() do { } while (0) | |
338 | #endif | |
339 | ||
e0982e90 | 340 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 341 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
342 | #endif |
343 | ||
30d7e0d4 ABL |
344 | /* |
345 | * We need to store the untouched command line for future reference. | |
346 | * We also need to store the touched command line since the parameter | |
347 | * parsing is performed in place, and we should allow a component to | |
348 | * store reference of name/value for future reference. | |
349 | */ | |
350 | static void __init setup_command_line(char *command_line) | |
351 | { | |
098b081b SS |
352 | saved_command_line = |
353 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
354 | initcall_command_line = | |
355 | memblock_virt_alloc(strlen(boot_command_line) + 1, 0); | |
356 | static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0); | |
30d7e0d4 ABL |
357 | strcpy (saved_command_line, boot_command_line); |
358 | strcpy (static_command_line, command_line); | |
359 | } | |
360 | ||
1da177e4 LT |
361 | /* |
362 | * We need to finalize in a non-__init function or else race conditions | |
363 | * between the root thread and the init thread may cause start_kernel to | |
364 | * be reaped by free_initmem before the root thread has proceeded to | |
365 | * cpu_idle. | |
366 | * | |
367 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
368 | */ | |
369 | ||
b433c3d4 PZ |
370 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
371 | ||
f99ebf0a | 372 | static noinline void __init_refok rest_init(void) |
1da177e4 | 373 | { |
73c27992 EB |
374 | int pid; |
375 | ||
7db905e6 | 376 | rcu_scheduler_starting(); |
b433c3d4 | 377 | /* |
97158569 | 378 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
379 | * the init task will end up wanting to create kthreads, which, if |
380 | * we schedule it before we create kthreadd, will OOPS. | |
381 | */ | |
aae5f662 | 382 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 383 | numa_default_policy(); |
73c27992 | 384 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 385 | rcu_read_lock(); |
5cd20455 | 386 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 387 | rcu_read_unlock(); |
b433c3d4 | 388 | complete(&kthreadd_done); |
f340c0d1 IM |
389 | |
390 | /* | |
391 | * The boot idle thread must execute schedule() | |
1df21055 | 392 | * at least once to get things moving: |
f340c0d1 | 393 | */ |
1df21055 | 394 | init_idle_bootup_task(current); |
bd2f5536 | 395 | schedule_preempt_disabled(); |
5bfb5d69 | 396 | /* Call into cpu_idle with preempt disabled */ |
a1a04ec3 | 397 | cpu_startup_entry(CPUHP_ONLINE); |
1df21055 | 398 | } |
1da177e4 LT |
399 | |
400 | /* Check for early params. */ | |
9fb48c74 | 401 | static int __init do_early_param(char *param, char *val, const char *unused) |
1da177e4 | 402 | { |
914dcaa8 | 403 | const struct obs_kernel_param *p; |
1da177e4 LT |
404 | |
405 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 406 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
407 | (strcmp(param, "console") == 0 && |
408 | strcmp(p->str, "earlycon") == 0) | |
409 | ) { | |
1da177e4 | 410 | if (p->setup_func(val) != 0) |
ea676e84 | 411 | pr_warn("Malformed early option '%s'\n", param); |
1da177e4 LT |
412 | } |
413 | } | |
414 | /* We accept everything at this stage. */ | |
415 | return 0; | |
416 | } | |
417 | ||
13977091 MD |
418 | void __init parse_early_options(char *cmdline) |
419 | { | |
026cee00 | 420 | parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param); |
13977091 MD |
421 | } |
422 | ||
1da177e4 LT |
423 | /* Arch code calls this early on, or if not, just before other parsing. */ |
424 | void __init parse_early_param(void) | |
425 | { | |
426 | static __initdata int done = 0; | |
427 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
428 | ||
429 | if (done) | |
430 | return; | |
431 | ||
432 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 433 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 434 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
435 | done = 1; |
436 | } | |
437 | ||
438 | /* | |
439 | * Activate the first processor. | |
440 | */ | |
441 | ||
44fd2299 SS |
442 | static void __init boot_cpu_init(void) |
443 | { | |
444 | int cpu = smp_processor_id(); | |
445 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 446 | set_cpu_online(cpu, true); |
933b0618 | 447 | set_cpu_active(cpu, true); |
915441b6 RR |
448 | set_cpu_present(cpu, true); |
449 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
450 | } |
451 | ||
839ad62e | 452 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
453 | { |
454 | } | |
455 | ||
eded09cc | 456 | # if THREAD_SIZE >= PAGE_SIZE |
8c9843e5 BH |
457 | void __init __weak thread_info_cache_init(void) |
458 | { | |
459 | } | |
eded09cc | 460 | #endif |
8c9843e5 | 461 | |
444f478f PE |
462 | /* |
463 | * Set up kernel memory allocators | |
464 | */ | |
465 | static void __init mm_init(void) | |
466 | { | |
ca371c0d | 467 | /* |
7fa87ce7 JC |
468 | * page_cgroup requires contiguous pages, |
469 | * bigger than MAX_ORDER unless SPARSEMEM. | |
ca371c0d KH |
470 | */ |
471 | page_cgroup_init_flatmem(); | |
444f478f PE |
472 | mem_init(); |
473 | kmem_cache_init(); | |
099a19d9 | 474 | percpu_init_late(); |
b35f1819 | 475 | pgtable_init(); |
444f478f PE |
476 | vmalloc_init(); |
477 | } | |
478 | ||
722a9f92 | 479 | asmlinkage __visible void __init start_kernel(void) |
1da177e4 LT |
480 | { |
481 | char * command_line; | |
914dcaa8 | 482 | extern const struct kernel_param __start___param[], __stop___param[]; |
033ab7f8 | 483 | |
fbb9ce95 IM |
484 | /* |
485 | * Need to run as early as possible, to initialize the | |
486 | * lockdep hash: | |
487 | */ | |
488 | lockdep_init(); | |
73839c5b | 489 | smp_setup_processor_id(); |
3ac7fe5a | 490 | debug_objects_early_init(); |
42059429 IM |
491 | |
492 | /* | |
493 | * Set up the the initial canary ASAP: | |
494 | */ | |
495 | boot_init_stack_canary(); | |
496 | ||
ddbcc7e8 | 497 | cgroup_init_early(); |
fbb9ce95 IM |
498 | |
499 | local_irq_disable(); | |
2ce802f6 | 500 | early_boot_irqs_disabled = true; |
fbb9ce95 | 501 | |
1da177e4 LT |
502 | /* |
503 | * Interrupts are still disabled. Do necessary setups, then | |
504 | * enable them | |
505 | */ | |
44fd2299 | 506 | boot_cpu_init(); |
1da177e4 | 507 | page_address_init(); |
ea676e84 | 508 | pr_notice("%s", linux_banner); |
1da177e4 | 509 | setup_arch(&command_line); |
6345d24d | 510 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 511 | setup_command_line(command_line); |
e0982e90 | 512 | setup_nr_cpu_ids(); |
d6647bdf | 513 | setup_per_cpu_areas(); |
44fd2299 | 514 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 515 | |
9adb62a5 | 516 | build_all_zonelists(NULL, NULL); |
83b519e8 PE |
517 | page_alloc_init(); |
518 | ||
ea676e84 | 519 | pr_notice("Kernel command line: %s\n", boot_command_line); |
83b519e8 PE |
520 | parse_early_param(); |
521 | parse_args("Booting kernel", static_command_line, __start___param, | |
522 | __stop___param - __start___param, | |
ae82fdb1 | 523 | -1, -1, &unknown_bootoption); |
97ce2c88 JF |
524 | |
525 | jump_label_init(); | |
526 | ||
83b519e8 PE |
527 | /* |
528 | * These use large bootmem allocations and must precede | |
529 | * kmem_cache_init() | |
530 | */ | |
162a7e75 | 531 | setup_log_buf(0); |
83b519e8 | 532 | pidhash_init(); |
83b519e8 PE |
533 | vfs_caches_init_early(); |
534 | sort_main_extable(); | |
535 | trap_init(); | |
444f478f | 536 | mm_init(); |
de03c72c | 537 | |
1da177e4 LT |
538 | /* |
539 | * Set up the scheduler prior starting any interrupts (such as the | |
540 | * timer interrupt). Full topology setup happens at smp_init() | |
541 | * time - but meanwhile we still have a functioning scheduler. | |
542 | */ | |
543 | sched_init(); | |
544 | /* | |
545 | * Disable preemption - early bootup scheduling is extremely | |
546 | * fragile until we cpu_idle() for the first time. | |
547 | */ | |
548 | preempt_disable(); | |
f91eb62f | 549 | if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n")) |
c4a68306 | 550 | local_irq_disable(); |
9f58a205 | 551 | idr_init_cache(); |
1da177e4 | 552 | rcu_init(); |
d1e43fa5 | 553 | tick_nohz_init(); |
65f382fd | 554 | context_tracking_init(); |
773e3eb7 | 555 | radix_tree_init(); |
0b8f1efa YL |
556 | /* init some links before init_ISA_irqs() */ |
557 | early_irq_init(); | |
1da177e4 | 558 | init_IRQ(); |
ad2b1353 | 559 | tick_init(); |
1da177e4 | 560 | init_timers(); |
c0a31329 | 561 | hrtimers_init(); |
1da177e4 | 562 | softirq_init(); |
ad596171 | 563 | timekeeping_init(); |
88fecaa2 | 564 | time_init(); |
38ff87f7 | 565 | sched_clock_postinit(); |
9e630205 | 566 | perf_event_init(); |
93e02814 | 567 | profile_init(); |
d8ad7d11 | 568 | call_function_init(); |
f91eb62f | 569 | WARN(!irqs_disabled(), "Interrupts were enabled early\n"); |
2ce802f6 | 570 | early_boot_irqs_disabled = false; |
93e02814 | 571 | local_irq_enable(); |
dcce284a | 572 | |
7e85ee0c | 573 | kmem_cache_init_late(); |
1da177e4 LT |
574 | |
575 | /* | |
576 | * HACK ALERT! This is early. We're enabling the console before | |
577 | * we've done PCI setups etc, and console_init() must be aware of | |
578 | * this. But we do want output early, in case something goes wrong. | |
579 | */ | |
580 | console_init(); | |
581 | if (panic_later) | |
499a4584 TH |
582 | panic("Too many boot %s vars at `%s'", panic_later, |
583 | panic_param); | |
fbb9ce95 IM |
584 | |
585 | lockdep_info(); | |
586 | ||
9a11b49a IM |
587 | /* |
588 | * Need to run this when irqs are enabled, because it wants | |
589 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
590 | * too: | |
591 | */ | |
592 | locking_selftest(); | |
593 | ||
1da177e4 LT |
594 | #ifdef CONFIG_BLK_DEV_INITRD |
595 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 596 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
ea676e84 | 597 | pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n", |
bd673c7c GU |
598 | page_to_pfn(virt_to_page((void *)initrd_start)), |
599 | min_low_pfn); | |
1da177e4 LT |
600 | initrd_start = 0; |
601 | } | |
602 | #endif | |
94b6da5a | 603 | page_cgroup_init(); |
3ac7fe5a | 604 | debug_objects_mem_init(); |
9b090f2d | 605 | kmemleak_init(); |
e7c8d5c9 | 606 | setup_per_cpu_pageset(); |
1da177e4 LT |
607 | numa_policy_init(); |
608 | if (late_time_init) | |
609 | late_time_init(); | |
fa84e9ee | 610 | sched_clock_init(); |
1da177e4 LT |
611 | calibrate_delay(); |
612 | pidmap_init(); | |
1da177e4 | 613 | anon_vma_init(); |
c4e1acbb | 614 | acpi_early_init(); |
11520e5e | 615 | #ifdef CONFIG_X86 |
83e68189 | 616 | if (efi_enabled(EFI_RUNTIME_SERVICES)) |
11520e5e LT |
617 | efi_enter_virtual_mode(); |
618 | #endif | |
8c9843e5 | 619 | thread_info_cache_init(); |
d84f4f99 | 620 | cred_init(); |
4481374c | 621 | fork_init(totalram_pages); |
1da177e4 LT |
622 | proc_caches_init(); |
623 | buffer_init(); | |
1da177e4 LT |
624 | key_init(); |
625 | security_init(); | |
0b4b3827 | 626 | dbg_late_init(); |
4481374c | 627 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
628 | signals_init(); |
629 | /* rootfs populating might need page-writeback */ | |
630 | page_writeback_init(); | |
631 | #ifdef CONFIG_PROC_FS | |
632 | proc_root_init(); | |
633 | #endif | |
ddbcc7e8 | 634 | cgroup_init(); |
1da177e4 | 635 | cpuset_init(); |
c757249a | 636 | taskstats_init_early(); |
ca74e92b | 637 | delayacct_init(); |
1da177e4 LT |
638 | |
639 | check_bugs(); | |
640 | ||
6ae6996a | 641 | sfi_init_late(); |
1da177e4 | 642 | |
83e68189 | 643 | if (efi_enabled(EFI_RUNTIME_SERVICES)) { |
2223af38 | 644 | efi_late_init(); |
78510792 | 645 | efi_free_boot_services(); |
2223af38 | 646 | } |
78510792 | 647 | |
68bf21aa SR |
648 | ftrace_init(); |
649 | ||
1da177e4 LT |
650 | /* Do the rest non-__init'ed, we're now alive */ |
651 | rest_init(); | |
652 | } | |
653 | ||
b99b87f7 PO |
654 | /* Call all constructor functions linked into the kernel. */ |
655 | static void __init do_ctors(void) | |
656 | { | |
657 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 658 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 659 | |
196a15b4 HS |
660 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
661 | (*fn)(); | |
b99b87f7 PO |
662 | #endif |
663 | } | |
664 | ||
2329abfa | 665 | bool initcall_debug; |
d0ea3d7d | 666 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 667 | |
e4461271 | 668 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 669 | { |
74239072 | 670 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
671 | unsigned long long duration; |
672 | int ret; | |
1da177e4 | 673 | |
d62cf815 | 674 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
22c5c03b | 675 | calltime = ktime_get(); |
30dbb20e | 676 | ret = fn(); |
22c5c03b KW |
677 | rettime = ktime_get(); |
678 | delta = ktime_sub(rettime, calltime); | |
679 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
d62cf815 | 680 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", |
ea676e84 | 681 | fn, ret, duration); |
1da177e4 | 682 | |
22c5c03b KW |
683 | return ret; |
684 | } | |
685 | ||
e4461271 | 686 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
687 | { |
688 | int count = preempt_count(); | |
689 | int ret; | |
ff1c8fac | 690 | char msgbuf[64]; |
22c5c03b KW |
691 | |
692 | if (initcall_debug) | |
693 | ret = do_one_initcall_debug(fn); | |
694 | else | |
695 | ret = fn(); | |
8f0c45cd | 696 | |
e0df154f | 697 | msgbuf[0] = 0; |
e662e1cf | 698 | |
e0df154f | 699 | if (preempt_count() != count) { |
bf5d770b | 700 | sprintf(msgbuf, "preemption imbalance "); |
4a2b4b22 | 701 | preempt_count_set(count); |
1da177e4 | 702 | } |
e0df154f | 703 | if (irqs_disabled()) { |
a76bfd0d | 704 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
705 | local_irq_enable(); |
706 | } | |
f91eb62f | 707 | WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf); |
59f9415f | 708 | |
30dbb20e | 709 | return ret; |
e0df154f LT |
710 | } |
711 | ||
712 | ||
026cee00 PM |
713 | extern initcall_t __initcall_start[]; |
714 | extern initcall_t __initcall0_start[]; | |
715 | extern initcall_t __initcall1_start[]; | |
716 | extern initcall_t __initcall2_start[]; | |
717 | extern initcall_t __initcall3_start[]; | |
718 | extern initcall_t __initcall4_start[]; | |
719 | extern initcall_t __initcall5_start[]; | |
720 | extern initcall_t __initcall6_start[]; | |
721 | extern initcall_t __initcall7_start[]; | |
722 | extern initcall_t __initcall_end[]; | |
723 | ||
724 | static initcall_t *initcall_levels[] __initdata = { | |
725 | __initcall0_start, | |
726 | __initcall1_start, | |
727 | __initcall2_start, | |
728 | __initcall3_start, | |
729 | __initcall4_start, | |
730 | __initcall5_start, | |
731 | __initcall6_start, | |
732 | __initcall7_start, | |
733 | __initcall_end, | |
734 | }; | |
735 | ||
96263d28 | 736 | /* Keep these in sync with initcalls in include/linux/init.h */ |
026cee00 | 737 | static char *initcall_level_names[] __initdata = { |
9fb48c74 JC |
738 | "early", |
739 | "core", | |
740 | "postcore", | |
741 | "arch", | |
742 | "subsys", | |
743 | "fs", | |
744 | "device", | |
745 | "late", | |
026cee00 PM |
746 | }; |
747 | ||
026cee00 | 748 | static void __init do_initcall_level(int level) |
e0df154f | 749 | { |
026cee00 | 750 | extern const struct kernel_param __start___param[], __stop___param[]; |
196a15b4 | 751 | initcall_t *fn; |
e0df154f | 752 | |
08746a65 | 753 | strcpy(initcall_command_line, saved_command_line); |
026cee00 | 754 | parse_args(initcall_level_names[level], |
08746a65 | 755 | initcall_command_line, __start___param, |
026cee00 PM |
756 | __stop___param - __start___param, |
757 | level, level, | |
eb157427 | 758 | &repair_env_string); |
026cee00 PM |
759 | |
760 | for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++) | |
196a15b4 | 761 | do_one_initcall(*fn); |
1da177e4 LT |
762 | } |
763 | ||
026cee00 PM |
764 | static void __init do_initcalls(void) |
765 | { | |
766 | int level; | |
767 | ||
19efb72f | 768 | for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) |
026cee00 PM |
769 | do_initcall_level(level); |
770 | } | |
771 | ||
1da177e4 LT |
772 | /* |
773 | * Ok, the machine is now initialized. None of the devices | |
774 | * have been touched yet, but the CPU subsystem is up and | |
775 | * running, and memory and process management works. | |
776 | * | |
777 | * Now we can finally start doing some real work.. | |
778 | */ | |
779 | static void __init do_basic_setup(void) | |
780 | { | |
759ee091 | 781 | cpuset_init_smp(); |
1da177e4 | 782 | usermodehelper_init(); |
41ffe5d5 | 783 | shmem_init(); |
1da177e4 | 784 | driver_init(); |
b04c3afb | 785 | init_irq_proc(); |
b99b87f7 | 786 | do_ctors(); |
d5767c53 | 787 | usermodehelper_enable(); |
b0f84374 | 788 | do_initcalls(); |
47d06e53 | 789 | random_int_secret_init(); |
1da177e4 LT |
790 | } |
791 | ||
7babe8db | 792 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 793 | { |
196a15b4 | 794 | initcall_t *fn; |
c2147a50 | 795 | |
026cee00 | 796 | for (fn = __initcall_start; fn < __initcall0_start; fn++) |
196a15b4 | 797 | do_one_initcall(*fn); |
c2147a50 EGM |
798 | } |
799 | ||
bb813f4c TH |
800 | /* |
801 | * This function requests modules which should be loaded by default and is | |
802 | * called twice right after initrd is mounted and right before init is | |
803 | * exec'd. If such modules are on either initrd or rootfs, they will be | |
804 | * loaded before control is passed to userland. | |
805 | */ | |
806 | void __init load_default_modules(void) | |
807 | { | |
808 | load_default_elevator_module(); | |
809 | } | |
810 | ||
a74fb73c | 811 | static int run_init_process(const char *init_filename) |
1da177e4 LT |
812 | { |
813 | argv_init[0] = init_filename; | |
c4ad8f98 | 814 | return do_execve(getname_kernel(init_filename), |
ae903caa AV |
815 | (const char __user *const __user *)argv_init, |
816 | (const char __user *const __user *)envp_init); | |
1da177e4 LT |
817 | } |
818 | ||
ba24762b MO |
819 | static int try_to_run_init_process(const char *init_filename) |
820 | { | |
821 | int ret; | |
822 | ||
823 | ret = run_init_process(init_filename); | |
824 | ||
825 | if (ret && ret != -ENOENT) { | |
826 | pr_err("Starting init: %s exists but couldn't execute it (error %d)\n", | |
827 | init_filename, ret); | |
828 | } | |
829 | ||
830 | return ret; | |
831 | } | |
832 | ||
f80b0c90 | 833 | static noinline void __init kernel_init_freeable(void); |
d6b21238 AV |
834 | |
835 | static int __ref kernel_init(void *unused) | |
ee5bfa64 | 836 | { |
ba24762b MO |
837 | int ret; |
838 | ||
d6b21238 | 839 | kernel_init_freeable(); |
22a9d645 AV |
840 | /* need to finish all async __init code before freeing the memory */ |
841 | async_synchronize_full(); | |
ee5bfa64 | 842 | free_initmem(); |
ee5bfa64 VG |
843 | mark_rodata_ro(); |
844 | system_state = SYSTEM_RUNNING; | |
845 | numa_default_policy(); | |
846 | ||
4a9d4b02 | 847 | flush_delayed_fput(); |
fae5fa44 | 848 | |
ee5bfa64 | 849 | if (ramdisk_execute_command) { |
ba24762b MO |
850 | ret = run_init_process(ramdisk_execute_command); |
851 | if (!ret) | |
a74fb73c | 852 | return 0; |
ba24762b MO |
853 | pr_err("Failed to execute %s (error %d)\n", |
854 | ramdisk_execute_command, ret); | |
ee5bfa64 VG |
855 | } |
856 | ||
857 | /* | |
858 | * We try each of these until one succeeds. | |
859 | * | |
860 | * The Bourne shell can be used instead of init if we are | |
861 | * trying to recover a really broken machine. | |
862 | */ | |
863 | if (execute_command) { | |
ba24762b MO |
864 | ret = run_init_process(execute_command); |
865 | if (!ret) | |
a74fb73c | 866 | return 0; |
ba24762b MO |
867 | pr_err("Failed to execute %s (error %d). Attempting defaults...\n", |
868 | execute_command, ret); | |
ee5bfa64 | 869 | } |
ba24762b MO |
870 | if (!try_to_run_init_process("/sbin/init") || |
871 | !try_to_run_init_process("/etc/init") || | |
872 | !try_to_run_init_process("/bin/init") || | |
873 | !try_to_run_init_process("/bin/sh")) | |
a74fb73c | 874 | return 0; |
ee5bfa64 | 875 | |
ba24762b | 876 | panic("No working init found. Try passing init= option to kernel. " |
9a85b8d6 | 877 | "See Linux Documentation/init.txt for guidance."); |
ee5bfa64 VG |
878 | } |
879 | ||
f80b0c90 | 880 | static noinline void __init kernel_init_freeable(void) |
1da177e4 | 881 | { |
b433c3d4 PZ |
882 | /* |
883 | * Wait until kthreadd is all set-up. | |
884 | */ | |
885 | wait_for_completion(&kthreadd_done); | |
31a67102 LT |
886 | |
887 | /* Now the scheduler is fully set up and can do blocking allocations */ | |
888 | gfp_allowed_mask = __GFP_BITS_MASK; | |
889 | ||
58568d2a MX |
890 | /* |
891 | * init can allocate pages on any node | |
892 | */ | |
3c466d46 | 893 | set_mems_allowed(node_states[N_MEMORY]); |
1da177e4 LT |
894 | /* |
895 | * init can run on any cpu. | |
896 | */ | |
1a2142af | 897 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 898 | |
9ec52099 CLG |
899 | cad_pid = task_pid(current); |
900 | ||
ca74a6f8 | 901 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
902 | |
903 | do_pre_smp_initcalls(); | |
004417a6 | 904 | lockup_detector_init(); |
1da177e4 | 905 | |
1da177e4 LT |
906 | smp_init(); |
907 | sched_init_smp(); | |
908 | ||
1da177e4 LT |
909 | do_basic_setup(); |
910 | ||
2bd3a997 EB |
911 | /* Open the /dev/console on the rootfs, this should never fail */ |
912 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
ea676e84 | 913 | pr_err("Warning: unable to open an initial console.\n"); |
2bd3a997 EB |
914 | |
915 | (void) sys_dup(0); | |
916 | (void) sys_dup(0); | |
1da177e4 LT |
917 | /* |
918 | * check if there is an early userspace init. If yes, let it do all | |
919 | * the work | |
920 | */ | |
ffdfc409 OJ |
921 | |
922 | if (!ramdisk_execute_command) | |
923 | ramdisk_execute_command = "/init"; | |
924 | ||
925 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
926 | ramdisk_execute_command = NULL; | |
1da177e4 | 927 | prepare_namespace(); |
ffdfc409 | 928 | } |
1da177e4 LT |
929 | |
930 | /* | |
931 | * Ok, we have completed the initial bootup, and | |
932 | * we're essentially up and running. Get rid of the | |
933 | * initmem segments and start the user-mode stuff.. | |
934 | */ | |
bb813f4c TH |
935 | |
936 | /* rootfs is available now, try loading default modules */ | |
937 | load_default_modules(); | |
1da177e4 | 938 | } |