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 | ||
1da177e4 LT |
12 | #include <linux/types.h> |
13 | #include <linux/module.h> | |
14 | #include <linux/proc_fs.h> | |
1da177e4 LT |
15 | #include <linux/kernel.h> |
16 | #include <linux/syscalls.h> | |
9b5609fd | 17 | #include <linux/stackprotector.h> |
1da177e4 LT |
18 | #include <linux/string.h> |
19 | #include <linux/ctype.h> | |
20 | #include <linux/delay.h> | |
1da177e4 LT |
21 | #include <linux/ioport.h> |
22 | #include <linux/init.h> | |
1da177e4 | 23 | #include <linux/initrd.h> |
1da177e4 | 24 | #include <linux/bootmem.h> |
4a7a16dc | 25 | #include <linux/acpi.h> |
1da177e4 | 26 | #include <linux/tty.h> |
1da177e4 LT |
27 | #include <linux/percpu.h> |
28 | #include <linux/kmod.h> | |
db64fe02 | 29 | #include <linux/vmalloc.h> |
1da177e4 | 30 | #include <linux/kernel_stat.h> |
d7cd5611 | 31 | #include <linux/start_kernel.h> |
1da177e4 | 32 | #include <linux/security.h> |
3d442233 | 33 | #include <linux/smp.h> |
1da177e4 LT |
34 | #include <linux/profile.h> |
35 | #include <linux/rcupdate.h> | |
36 | #include <linux/moduleparam.h> | |
37 | #include <linux/kallsyms.h> | |
38 | #include <linux/writeback.h> | |
39 | #include <linux/cpu.h> | |
40 | #include <linux/cpuset.h> | |
ddbcc7e8 | 41 | #include <linux/cgroup.h> |
1da177e4 | 42 | #include <linux/efi.h> |
906568c9 | 43 | #include <linux/tick.h> |
6168a702 | 44 | #include <linux/interrupt.h> |
c757249a | 45 | #include <linux/taskstats_kern.h> |
ca74e92b | 46 | #include <linux/delayacct.h> |
1da177e4 LT |
47 | #include <linux/unistd.h> |
48 | #include <linux/rmap.h> | |
49 | #include <linux/mempolicy.h> | |
50 | #include <linux/key.h> | |
b6cd0b77 | 51 | #include <linux/buffer_head.h> |
94b6da5a | 52 | #include <linux/page_cgroup.h> |
9a11b49a | 53 | #include <linux/debug_locks.h> |
3ac7fe5a | 54 | #include <linux/debugobjects.h> |
fbb9ce95 | 55 | #include <linux/lockdep.h> |
3c7b4e6b | 56 | #include <linux/kmemleak.h> |
84d73786 | 57 | #include <linux/pid_namespace.h> |
1f21782e | 58 | #include <linux/device.h> |
73c27992 | 59 | #include <linux/kthread.h> |
e6fe6649 | 60 | #include <linux/sched.h> |
a1c9eea9 | 61 | #include <linux/signal.h> |
199f0ca5 | 62 | #include <linux/idr.h> |
0b4b3827 | 63 | #include <linux/kgdb.h> |
68bf21aa | 64 | #include <linux/ftrace.h> |
22a9d645 | 65 | #include <linux/async.h> |
dfec072e | 66 | #include <linux/kmemcheck.h> |
6ae6996a | 67 | #include <linux/sfi.h> |
2b2af54a | 68 | #include <linux/shmem_fs.h> |
5a0e3ad6 | 69 | #include <linux/slab.h> |
24a24bb6 | 70 | #include <linux/perf_event.h> |
1da177e4 LT |
71 | |
72 | #include <asm/io.h> | |
73 | #include <asm/bugs.h> | |
74 | #include <asm/setup.h> | |
a940199f | 75 | #include <asm/sections.h> |
37b73c82 | 76 | #include <asm/cacheflush.h> |
1da177e4 | 77 | |
1da177e4 LT |
78 | #ifdef CONFIG_X86_LOCAL_APIC |
79 | #include <asm/smp.h> | |
80 | #endif | |
81 | ||
aae5f662 | 82 | static int kernel_init(void *); |
1da177e4 LT |
83 | |
84 | extern void init_IRQ(void); | |
1da177e4 LT |
85 | extern void fork_init(unsigned long); |
86 | extern void mca_init(void); | |
87 | extern void sbus_init(void); | |
1da177e4 LT |
88 | extern void prio_tree_init(void); |
89 | extern void radix_tree_init(void); | |
90 | extern void free_initmem(void); | |
37b73c82 AV |
91 | #ifndef CONFIG_DEBUG_RODATA |
92 | static inline void mark_rodata_ro(void) { } | |
93 | #endif | |
1da177e4 LT |
94 | |
95 | #ifdef CONFIG_TC | |
96 | extern void tc_init(void); | |
97 | #endif | |
98 | ||
2ce802f6 TH |
99 | /* |
100 | * Debug helper: via this flag we know that we are in 'early bootup code' | |
101 | * where only the boot processor is running with IRQ disabled. This means | |
102 | * two things - IRQ must not be enabled before the flag is cleared and some | |
103 | * operations which are not allowed with IRQ disabled are allowed while the | |
104 | * flag is set. | |
105 | */ | |
106 | bool early_boot_irqs_disabled __read_mostly; | |
107 | ||
a6826048 | 108 | enum system_states system_state __read_mostly; |
1da177e4 LT |
109 | EXPORT_SYMBOL(system_state); |
110 | ||
111 | /* | |
112 | * Boot command-line arguments | |
113 | */ | |
114 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
115 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
116 | ||
117 | extern void time_init(void); | |
118 | /* Default late time init is NULL. archs can override this later. */ | |
d2e3192b | 119 | void (*__initdata late_time_init)(void); |
1da177e4 LT |
120 | extern void softirq_init(void); |
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; | |
1da177e4 LT |
128 | |
129 | static char *execute_command; | |
ffdfc409 | 130 | static char *ramdisk_execute_command; |
1da177e4 | 131 | |
8b3b2955 JB |
132 | /* |
133 | * If set, this is an indication to the drivers that reset the underlying | |
134 | * device before going ahead with the initialization otherwise driver might | |
135 | * rely on the BIOS and skip the reset operation. | |
136 | * | |
137 | * This is useful if kernel is booting in an unreliable environment. | |
138 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
139 | * skipped and devices will be in unknown state. | |
140 | */ | |
141 | unsigned int reset_devices; | |
142 | EXPORT_SYMBOL(reset_devices); | |
1da177e4 | 143 | |
7e96287d VG |
144 | static int __init set_reset_devices(char *str) |
145 | { | |
146 | reset_devices = 1; | |
147 | return 1; | |
148 | } | |
149 | ||
150 | __setup("reset_devices", set_reset_devices); | |
151 | ||
d7627467 DH |
152 | static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; |
153 | const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
1da177e4 LT |
154 | static const char *panic_later, *panic_param; |
155 | ||
914dcaa8 | 156 | extern const struct obs_kernel_param __setup_start[], __setup_end[]; |
1da177e4 LT |
157 | |
158 | static int __init obsolete_checksetup(char *line) | |
159 | { | |
914dcaa8 | 160 | const struct obs_kernel_param *p; |
33df0d19 | 161 | int had_early_param = 0; |
1da177e4 LT |
162 | |
163 | p = __setup_start; | |
164 | do { | |
165 | int n = strlen(p->str); | |
b1e4d20c | 166 | if (parameqn(line, p->str, n)) { |
1da177e4 | 167 | if (p->early) { |
33df0d19 RR |
168 | /* Already done in parse_early_param? |
169 | * (Needs exact match on param part). | |
170 | * Keep iterating, as we can have early | |
171 | * params and __setups of same names 8( */ | |
1da177e4 | 172 | if (line[n] == '\0' || line[n] == '=') |
33df0d19 | 173 | had_early_param = 1; |
1da177e4 LT |
174 | } else if (!p->setup_func) { |
175 | printk(KERN_WARNING "Parameter %s is obsolete," | |
176 | " ignored\n", p->str); | |
177 | return 1; | |
178 | } else if (p->setup_func(line + n)) | |
179 | return 1; | |
180 | } | |
181 | p++; | |
182 | } while (p < __setup_end); | |
33df0d19 RR |
183 | |
184 | return had_early_param; | |
1da177e4 LT |
185 | } |
186 | ||
187 | /* | |
188 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
189 | * still work even if initially too large, it will just take slightly longer | |
190 | */ | |
191 | unsigned long loops_per_jiffy = (1<<12); | |
192 | ||
193 | EXPORT_SYMBOL(loops_per_jiffy); | |
194 | ||
195 | static int __init debug_kernel(char *str) | |
196 | { | |
1da177e4 | 197 | console_loglevel = 10; |
f6f21c81 | 198 | return 0; |
1da177e4 LT |
199 | } |
200 | ||
201 | static int __init quiet_kernel(char *str) | |
202 | { | |
1da177e4 | 203 | console_loglevel = 4; |
f6f21c81 | 204 | return 0; |
1da177e4 LT |
205 | } |
206 | ||
f6f21c81 YL |
207 | early_param("debug", debug_kernel); |
208 | early_param("quiet", quiet_kernel); | |
1da177e4 LT |
209 | |
210 | static int __init loglevel(char *str) | |
211 | { | |
808bf29b AS |
212 | int newlevel; |
213 | ||
214 | /* | |
215 | * Only update loglevel value when a correct setting was passed, | |
216 | * to prevent blind crashes (when loglevel being set to 0) that | |
217 | * are quite hard to debug | |
218 | */ | |
219 | if (get_option(&str, &newlevel)) { | |
220 | console_loglevel = newlevel; | |
221 | return 0; | |
222 | } | |
223 | ||
224 | return -EINVAL; | |
1da177e4 LT |
225 | } |
226 | ||
f6f21c81 | 227 | early_param("loglevel", loglevel); |
1da177e4 LT |
228 | |
229 | /* | |
230 | * Unknown boot options get handed to init, unless they look like | |
f066a4f6 | 231 | * unused parameters (modprobe will find them in /proc/cmdline). |
1da177e4 LT |
232 | */ |
233 | static int __init unknown_bootoption(char *param, char *val) | |
234 | { | |
235 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
236 | if (val) { | |
237 | /* param=val or param="val"? */ | |
238 | if (val == param+strlen(param)+1) | |
239 | val[-1] = '='; | |
240 | else if (val == param+strlen(param)+2) { | |
241 | val[-2] = '='; | |
242 | memmove(val-1, val, strlen(val)+1); | |
243 | val--; | |
244 | } else | |
245 | BUG(); | |
246 | } | |
247 | ||
248 | /* Handle obsolete-style parameters */ | |
249 | if (obsolete_checksetup(param)) | |
250 | return 0; | |
251 | ||
f066a4f6 RR |
252 | /* Unused module parameter. */ |
253 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
1da177e4 | 254 | return 0; |
1da177e4 LT |
255 | |
256 | if (panic_later) | |
257 | return 0; | |
258 | ||
259 | if (val) { | |
260 | /* Environment option */ | |
261 | unsigned int i; | |
262 | for (i = 0; envp_init[i]; i++) { | |
263 | if (i == MAX_INIT_ENVS) { | |
264 | panic_later = "Too many boot env vars at `%s'"; | |
265 | panic_param = param; | |
266 | } | |
267 | if (!strncmp(param, envp_init[i], val - param)) | |
268 | break; | |
269 | } | |
270 | envp_init[i] = param; | |
271 | } else { | |
272 | /* Command line option */ | |
273 | unsigned int i; | |
274 | for (i = 0; argv_init[i]; i++) { | |
275 | if (i == MAX_INIT_ARGS) { | |
276 | panic_later = "Too many boot init vars at `%s'"; | |
277 | panic_param = param; | |
278 | } | |
279 | } | |
280 | argv_init[i] = param; | |
281 | } | |
282 | return 0; | |
283 | } | |
284 | ||
12d6f21e IM |
285 | #ifdef CONFIG_DEBUG_PAGEALLOC |
286 | int __read_mostly debug_pagealloc_enabled = 0; | |
287 | #endif | |
288 | ||
1da177e4 LT |
289 | static int __init init_setup(char *str) |
290 | { | |
291 | unsigned int i; | |
292 | ||
293 | execute_command = str; | |
294 | /* | |
295 | * In case LILO is going to boot us with default command line, | |
296 | * it prepends "auto" before the whole cmdline which makes | |
297 | * the shell think it should execute a script with such name. | |
298 | * So we ignore all arguments entered _before_ init=... [MJ] | |
299 | */ | |
300 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
301 | argv_init[i] = NULL; | |
302 | return 1; | |
303 | } | |
304 | __setup("init=", init_setup); | |
305 | ||
ffdfc409 OJ |
306 | static int __init rdinit_setup(char *str) |
307 | { | |
308 | unsigned int i; | |
309 | ||
310 | ramdisk_execute_command = str; | |
311 | /* See "auto" comment in init_setup */ | |
312 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
313 | argv_init[i] = NULL; | |
314 | return 1; | |
315 | } | |
316 | __setup("rdinit=", rdinit_setup); | |
317 | ||
1da177e4 | 318 | #ifndef CONFIG_SMP |
34db18a0 | 319 | static const unsigned int setup_max_cpus = NR_CPUS; |
1da177e4 LT |
320 | #ifdef CONFIG_X86_LOCAL_APIC |
321 | static void __init smp_init(void) | |
322 | { | |
323 | APIC_init_uniprocessor(); | |
324 | } | |
325 | #else | |
326 | #define smp_init() do { } while (0) | |
327 | #endif | |
328 | ||
e0982e90 | 329 | static inline void setup_nr_cpu_ids(void) { } |
1da177e4 | 330 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } |
1da177e4 LT |
331 | #endif |
332 | ||
30d7e0d4 ABL |
333 | /* |
334 | * We need to store the untouched command line for future reference. | |
335 | * We also need to store the touched command line since the parameter | |
336 | * parsing is performed in place, and we should allow a component to | |
337 | * store reference of name/value for future reference. | |
338 | */ | |
339 | static void __init setup_command_line(char *command_line) | |
340 | { | |
341 | saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); | |
342 | static_command_line = alloc_bootmem(strlen (command_line)+1); | |
343 | strcpy (saved_command_line, boot_command_line); | |
344 | strcpy (static_command_line, command_line); | |
345 | } | |
346 | ||
1da177e4 LT |
347 | /* |
348 | * We need to finalize in a non-__init function or else race conditions | |
349 | * between the root thread and the init thread may cause start_kernel to | |
350 | * be reaped by free_initmem before the root thread has proceeded to | |
351 | * cpu_idle. | |
352 | * | |
353 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
354 | */ | |
355 | ||
b433c3d4 PZ |
356 | static __initdata DECLARE_COMPLETION(kthreadd_done); |
357 | ||
f99ebf0a | 358 | static noinline void __init_refok rest_init(void) |
1da177e4 | 359 | { |
73c27992 EB |
360 | int pid; |
361 | ||
7db905e6 | 362 | rcu_scheduler_starting(); |
b433c3d4 | 363 | /* |
97158569 | 364 | * We need to spawn init first so that it obtains pid 1, however |
b433c3d4 PZ |
365 | * the init task will end up wanting to create kthreads, which, if |
366 | * we schedule it before we create kthreadd, will OOPS. | |
367 | */ | |
aae5f662 | 368 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); |
1da177e4 | 369 | numa_default_policy(); |
73c27992 | 370 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); |
d11c563d | 371 | rcu_read_lock(); |
5cd20455 | 372 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); |
d11c563d | 373 | rcu_read_unlock(); |
b433c3d4 | 374 | complete(&kthreadd_done); |
f340c0d1 IM |
375 | |
376 | /* | |
377 | * The boot idle thread must execute schedule() | |
1df21055 | 378 | * at least once to get things moving: |
f340c0d1 | 379 | */ |
1df21055 | 380 | init_idle_bootup_task(current); |
5bfb5d69 | 381 | preempt_enable_no_resched(); |
f340c0d1 | 382 | schedule(); |
288d5abe | 383 | |
5bfb5d69 | 384 | /* Call into cpu_idle with preempt disabled */ |
288d5abe | 385 | preempt_disable(); |
1da177e4 | 386 | cpu_idle(); |
1df21055 | 387 | } |
1da177e4 LT |
388 | |
389 | /* Check for early params. */ | |
390 | static int __init do_early_param(char *param, char *val) | |
391 | { | |
914dcaa8 | 392 | const struct obs_kernel_param *p; |
1da177e4 LT |
393 | |
394 | for (p = __setup_start; p < __setup_end; p++) { | |
b1e4d20c | 395 | if ((p->early && parameq(param, p->str)) || |
18a8bd94 YL |
396 | (strcmp(param, "console") == 0 && |
397 | strcmp(p->str, "earlycon") == 0) | |
398 | ) { | |
1da177e4 LT |
399 | if (p->setup_func(val) != 0) |
400 | printk(KERN_WARNING | |
401 | "Malformed early option '%s'\n", param); | |
402 | } | |
403 | } | |
404 | /* We accept everything at this stage. */ | |
405 | return 0; | |
406 | } | |
407 | ||
13977091 MD |
408 | void __init parse_early_options(char *cmdline) |
409 | { | |
410 | parse_args("early options", cmdline, NULL, 0, do_early_param); | |
411 | } | |
412 | ||
1da177e4 LT |
413 | /* Arch code calls this early on, or if not, just before other parsing. */ |
414 | void __init parse_early_param(void) | |
415 | { | |
416 | static __initdata int done = 0; | |
417 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
418 | ||
419 | if (done) | |
420 | return; | |
421 | ||
422 | /* All fall through to do_early_param. */ | |
30d7e0d4 | 423 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); |
13977091 | 424 | parse_early_options(tmp_cmdline); |
1da177e4 LT |
425 | done = 1; |
426 | } | |
427 | ||
428 | /* | |
429 | * Activate the first processor. | |
430 | */ | |
431 | ||
44fd2299 SS |
432 | static void __init boot_cpu_init(void) |
433 | { | |
434 | int cpu = smp_processor_id(); | |
435 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
915441b6 | 436 | set_cpu_online(cpu, true); |
933b0618 | 437 | set_cpu_active(cpu, true); |
915441b6 RR |
438 | set_cpu_present(cpu, true); |
439 | set_cpu_possible(cpu, true); | |
44fd2299 SS |
440 | } |
441 | ||
839ad62e | 442 | void __init __weak smp_setup_processor_id(void) |
033ab7f8 AM |
443 | { |
444 | } | |
445 | ||
8c9843e5 BH |
446 | void __init __weak thread_info_cache_init(void) |
447 | { | |
448 | } | |
449 | ||
444f478f PE |
450 | /* |
451 | * Set up kernel memory allocators | |
452 | */ | |
453 | static void __init mm_init(void) | |
454 | { | |
ca371c0d KH |
455 | /* |
456 | * page_cgroup requires countinous pages as memmap | |
457 | * and it's bigger than MAX_ORDER unless SPARSEMEM. | |
458 | */ | |
459 | page_cgroup_init_flatmem(); | |
444f478f PE |
460 | mem_init(); |
461 | kmem_cache_init(); | |
099a19d9 | 462 | percpu_init_late(); |
c868d550 | 463 | pgtable_cache_init(); |
444f478f PE |
464 | vmalloc_init(); |
465 | } | |
466 | ||
1da177e4 LT |
467 | asmlinkage void __init start_kernel(void) |
468 | { | |
469 | char * command_line; | |
914dcaa8 | 470 | extern const struct kernel_param __start___param[], __stop___param[]; |
033ab7f8 | 471 | |
fbb9ce95 IM |
472 | /* |
473 | * Need to run as early as possible, to initialize the | |
474 | * lockdep hash: | |
475 | */ | |
476 | lockdep_init(); | |
73839c5b | 477 | smp_setup_processor_id(); |
3ac7fe5a | 478 | debug_objects_early_init(); |
42059429 IM |
479 | |
480 | /* | |
481 | * Set up the the initial canary ASAP: | |
482 | */ | |
483 | boot_init_stack_canary(); | |
484 | ||
ddbcc7e8 | 485 | cgroup_init_early(); |
fbb9ce95 IM |
486 | |
487 | local_irq_disable(); | |
2ce802f6 | 488 | early_boot_irqs_disabled = true; |
fbb9ce95 | 489 | |
1da177e4 LT |
490 | /* |
491 | * Interrupts are still disabled. Do necessary setups, then | |
492 | * enable them | |
493 | */ | |
906568c9 | 494 | tick_init(); |
44fd2299 | 495 | boot_cpu_init(); |
1da177e4 | 496 | page_address_init(); |
657cafa6 | 497 | printk(KERN_NOTICE "%s", linux_banner); |
1da177e4 | 498 | setup_arch(&command_line); |
cf475ad2 | 499 | mm_init_owner(&init_mm, &init_task); |
6345d24d | 500 | mm_init_cpumask(&init_mm); |
30d7e0d4 | 501 | setup_command_line(command_line); |
e0982e90 | 502 | setup_nr_cpu_ids(); |
d6647bdf | 503 | setup_per_cpu_areas(); |
44fd2299 | 504 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ |
1da177e4 | 505 | |
1f522509 | 506 | build_all_zonelists(NULL); |
83b519e8 PE |
507 | page_alloc_init(); |
508 | ||
509 | printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); | |
510 | parse_early_param(); | |
511 | parse_args("Booting kernel", static_command_line, __start___param, | |
512 | __stop___param - __start___param, | |
513 | &unknown_bootoption); | |
97ce2c88 JF |
514 | |
515 | jump_label_init(); | |
516 | ||
83b519e8 PE |
517 | /* |
518 | * These use large bootmem allocations and must precede | |
519 | * kmem_cache_init() | |
520 | */ | |
162a7e75 | 521 | setup_log_buf(0); |
83b519e8 | 522 | pidhash_init(); |
83b519e8 PE |
523 | vfs_caches_init_early(); |
524 | sort_main_extable(); | |
525 | trap_init(); | |
444f478f | 526 | mm_init(); |
de03c72c | 527 | |
1da177e4 LT |
528 | /* |
529 | * Set up the scheduler prior starting any interrupts (such as the | |
530 | * timer interrupt). Full topology setup happens at smp_init() | |
531 | * time - but meanwhile we still have a functioning scheduler. | |
532 | */ | |
533 | sched_init(); | |
534 | /* | |
535 | * Disable preemption - early bootup scheduling is extremely | |
536 | * fragile until we cpu_idle() for the first time. | |
537 | */ | |
538 | preempt_disable(); | |
c4a68306 AB |
539 | if (!irqs_disabled()) { |
540 | printk(KERN_WARNING "start_kernel(): bug: interrupts were " | |
541 | "enabled *very* early, fixing it\n"); | |
542 | local_irq_disable(); | |
543 | } | |
9f58a205 | 544 | idr_init_cache(); |
24a24bb6 | 545 | perf_event_init(); |
1da177e4 | 546 | rcu_init(); |
773e3eb7 | 547 | radix_tree_init(); |
0b8f1efa YL |
548 | /* init some links before init_ISA_irqs() */ |
549 | early_irq_init(); | |
1da177e4 | 550 | init_IRQ(); |
3c7b4e6b | 551 | prio_tree_init(); |
1da177e4 | 552 | init_timers(); |
c0a31329 | 553 | hrtimers_init(); |
1da177e4 | 554 | softirq_init(); |
ad596171 | 555 | timekeeping_init(); |
88fecaa2 | 556 | time_init(); |
93e02814 | 557 | profile_init(); |
d8ad7d11 | 558 | call_function_init(); |
93e02814 | 559 | if (!irqs_disabled()) |
24d431d0 RL |
560 | printk(KERN_CRIT "start_kernel(): bug: interrupts were " |
561 | "enabled early\n"); | |
2ce802f6 | 562 | early_boot_irqs_disabled = false; |
93e02814 | 563 | local_irq_enable(); |
dcce284a BH |
564 | |
565 | /* Interrupts are enabled now so all GFP allocations are safe. */ | |
452aa699 | 566 | gfp_allowed_mask = __GFP_BITS_MASK; |
dcce284a | 567 | |
7e85ee0c | 568 | kmem_cache_init_late(); |
1da177e4 LT |
569 | |
570 | /* | |
571 | * HACK ALERT! This is early. We're enabling the console before | |
572 | * we've done PCI setups etc, and console_init() must be aware of | |
573 | * this. But we do want output early, in case something goes wrong. | |
574 | */ | |
575 | console_init(); | |
576 | if (panic_later) | |
577 | panic(panic_later, panic_param); | |
fbb9ce95 IM |
578 | |
579 | lockdep_info(); | |
580 | ||
9a11b49a IM |
581 | /* |
582 | * Need to run this when irqs are enabled, because it wants | |
583 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
584 | * too: | |
585 | */ | |
586 | locking_selftest(); | |
587 | ||
1da177e4 LT |
588 | #ifdef CONFIG_BLK_DEV_INITRD |
589 | if (initrd_start && !initrd_below_start_ok && | |
bd673c7c | 590 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { |
1da177e4 | 591 | printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " |
fb6624eb | 592 | "disabling it.\n", |
bd673c7c GU |
593 | page_to_pfn(virt_to_page((void *)initrd_start)), |
594 | min_low_pfn); | |
1da177e4 LT |
595 | initrd_start = 0; |
596 | } | |
597 | #endif | |
94b6da5a | 598 | page_cgroup_init(); |
a03c2a48 | 599 | enable_debug_pagealloc(); |
3ac7fe5a | 600 | debug_objects_mem_init(); |
9b090f2d | 601 | kmemleak_init(); |
e7c8d5c9 | 602 | setup_per_cpu_pageset(); |
1da177e4 LT |
603 | numa_policy_init(); |
604 | if (late_time_init) | |
605 | late_time_init(); | |
fa84e9ee | 606 | sched_clock_init(); |
1da177e4 LT |
607 | calibrate_delay(); |
608 | pidmap_init(); | |
1da177e4 LT |
609 | anon_vma_init(); |
610 | #ifdef CONFIG_X86 | |
611 | if (efi_enabled) | |
612 | efi_enter_virtual_mode(); | |
613 | #endif | |
8c9843e5 | 614 | thread_info_cache_init(); |
d84f4f99 | 615 | cred_init(); |
4481374c | 616 | fork_init(totalram_pages); |
1da177e4 LT |
617 | proc_caches_init(); |
618 | buffer_init(); | |
1da177e4 LT |
619 | key_init(); |
620 | security_init(); | |
0b4b3827 | 621 | dbg_late_init(); |
4481374c | 622 | vfs_caches_init(totalram_pages); |
1da177e4 LT |
623 | signals_init(); |
624 | /* rootfs populating might need page-writeback */ | |
625 | page_writeback_init(); | |
626 | #ifdef CONFIG_PROC_FS | |
627 | proc_root_init(); | |
628 | #endif | |
ddbcc7e8 | 629 | cgroup_init(); |
1da177e4 | 630 | cpuset_init(); |
c757249a | 631 | taskstats_init_early(); |
ca74e92b | 632 | delayacct_init(); |
1da177e4 LT |
633 | |
634 | check_bugs(); | |
635 | ||
636 | acpi_early_init(); /* before LAPIC and SMP init */ | |
6ae6996a | 637 | sfi_init_late(); |
1da177e4 | 638 | |
68bf21aa SR |
639 | ftrace_init(); |
640 | ||
1da177e4 LT |
641 | /* Do the rest non-__init'ed, we're now alive */ |
642 | rest_init(); | |
643 | } | |
644 | ||
b99b87f7 PO |
645 | /* Call all constructor functions linked into the kernel. */ |
646 | static void __init do_ctors(void) | |
647 | { | |
648 | #ifdef CONFIG_CONSTRUCTORS | |
196a15b4 | 649 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; |
b99b87f7 | 650 | |
196a15b4 HS |
651 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) |
652 | (*fn)(); | |
b99b87f7 PO |
653 | #endif |
654 | } | |
655 | ||
22a9d645 | 656 | int initcall_debug; |
d0ea3d7d | 657 | core_param(initcall_debug, initcall_debug, bool, 0644); |
1da177e4 | 658 | |
4a683bf9 | 659 | static char msgbuf[64]; |
4a683bf9 | 660 | |
e4461271 | 661 | static int __init_or_module do_one_initcall_debug(initcall_t fn) |
1da177e4 | 662 | { |
74239072 | 663 | ktime_t calltime, delta, rettime; |
30dbb20e AW |
664 | unsigned long long duration; |
665 | int ret; | |
1da177e4 | 666 | |
22c5c03b KW |
667 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); |
668 | calltime = ktime_get(); | |
30dbb20e | 669 | ret = fn(); |
22c5c03b KW |
670 | rettime = ktime_get(); |
671 | delta = ktime_sub(rettime, calltime); | |
672 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
673 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn, | |
674 | ret, duration); | |
1da177e4 | 675 | |
22c5c03b KW |
676 | return ret; |
677 | } | |
678 | ||
e4461271 | 679 | int __init_or_module do_one_initcall(initcall_t fn) |
22c5c03b KW |
680 | { |
681 | int count = preempt_count(); | |
682 | int ret; | |
683 | ||
684 | if (initcall_debug) | |
685 | ret = do_one_initcall_debug(fn); | |
686 | else | |
687 | ret = fn(); | |
8f0c45cd | 688 | |
e0df154f | 689 | msgbuf[0] = 0; |
e662e1cf | 690 | |
30dbb20e AW |
691 | if (ret && ret != -ENODEV && initcall_debug) |
692 | sprintf(msgbuf, "error code %d ", ret); | |
e662e1cf | 693 | |
e0df154f | 694 | if (preempt_count() != count) { |
a76bfd0d | 695 | strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); |
e0df154f | 696 | preempt_count() = count; |
1da177e4 | 697 | } |
e0df154f | 698 | if (irqs_disabled()) { |
a76bfd0d | 699 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); |
e0df154f LT |
700 | local_irq_enable(); |
701 | } | |
702 | if (msgbuf[0]) { | |
96d746c6 | 703 | printk("initcall %pF returned with %s\n", fn, msgbuf); |
e0df154f | 704 | } |
59f9415f | 705 | |
30dbb20e | 706 | return ret; |
e0df154f LT |
707 | } |
708 | ||
709 | ||
c2147a50 | 710 | extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[]; |
e0df154f LT |
711 | |
712 | static void __init do_initcalls(void) | |
713 | { | |
196a15b4 | 714 | initcall_t *fn; |
e0df154f | 715 | |
196a15b4 HS |
716 | for (fn = __early_initcall_end; fn < __initcall_end; fn++) |
717 | do_one_initcall(*fn); | |
1da177e4 LT |
718 | } |
719 | ||
720 | /* | |
721 | * Ok, the machine is now initialized. None of the devices | |
722 | * have been touched yet, but the CPU subsystem is up and | |
723 | * running, and memory and process management works. | |
724 | * | |
725 | * Now we can finally start doing some real work.. | |
726 | */ | |
727 | static void __init do_basic_setup(void) | |
728 | { | |
759ee091 | 729 | cpuset_init_smp(); |
1da177e4 | 730 | usermodehelper_init(); |
41ffe5d5 | 731 | shmem_init(); |
1da177e4 | 732 | driver_init(); |
b04c3afb | 733 | init_irq_proc(); |
b99b87f7 | 734 | do_ctors(); |
d5767c53 | 735 | usermodehelper_enable(); |
b0f84374 | 736 | do_initcalls(); |
1da177e4 LT |
737 | } |
738 | ||
7babe8db | 739 | static void __init do_pre_smp_initcalls(void) |
c2147a50 | 740 | { |
196a15b4 | 741 | initcall_t *fn; |
c2147a50 | 742 | |
196a15b4 HS |
743 | for (fn = __initcall_start; fn < __early_initcall_end; fn++) |
744 | do_one_initcall(*fn); | |
c2147a50 EGM |
745 | } |
746 | ||
d7627467 | 747 | static void run_init_process(const char *init_filename) |
1da177e4 LT |
748 | { |
749 | argv_init[0] = init_filename; | |
67608567 | 750 | kernel_execve(init_filename, argv_init, envp_init); |
1da177e4 LT |
751 | } |
752 | ||
ee5bfa64 VG |
753 | /* This is a non __init function. Force it to be noinline otherwise gcc |
754 | * makes it inline to init() and it becomes part of init.text section | |
755 | */ | |
f99ebf0a | 756 | static noinline int init_post(void) |
ee5bfa64 | 757 | { |
22a9d645 AV |
758 | /* need to finish all async __init code before freeing the memory */ |
759 | async_synchronize_full(); | |
ee5bfa64 | 760 | free_initmem(); |
ee5bfa64 VG |
761 | mark_rodata_ro(); |
762 | system_state = SYSTEM_RUNNING; | |
763 | numa_default_policy(); | |
764 | ||
ee5bfa64 | 765 | |
fae5fa44 ON |
766 | current->signal->flags |= SIGNAL_UNKILLABLE; |
767 | ||
ee5bfa64 VG |
768 | if (ramdisk_execute_command) { |
769 | run_init_process(ramdisk_execute_command); | |
770 | printk(KERN_WARNING "Failed to execute %s\n", | |
771 | ramdisk_execute_command); | |
772 | } | |
773 | ||
774 | /* | |
775 | * We try each of these until one succeeds. | |
776 | * | |
777 | * The Bourne shell can be used instead of init if we are | |
778 | * trying to recover a really broken machine. | |
779 | */ | |
780 | if (execute_command) { | |
781 | run_init_process(execute_command); | |
782 | printk(KERN_WARNING "Failed to execute %s. Attempting " | |
783 | "defaults...\n", execute_command); | |
784 | } | |
785 | run_init_process("/sbin/init"); | |
786 | run_init_process("/etc/init"); | |
787 | run_init_process("/bin/init"); | |
788 | run_init_process("/bin/sh"); | |
789 | ||
9a85b8d6 AM |
790 | panic("No init found. Try passing init= option to kernel. " |
791 | "See Linux Documentation/init.txt for guidance."); | |
ee5bfa64 VG |
792 | } |
793 | ||
aae5f662 | 794 | static int __init kernel_init(void * unused) |
1da177e4 | 795 | { |
b433c3d4 PZ |
796 | /* |
797 | * Wait until kthreadd is all set-up. | |
798 | */ | |
799 | wait_for_completion(&kthreadd_done); | |
58568d2a MX |
800 | /* |
801 | * init can allocate pages on any node | |
802 | */ | |
5ab116c9 | 803 | set_mems_allowed(node_states[N_HIGH_MEMORY]); |
1da177e4 LT |
804 | /* |
805 | * init can run on any cpu. | |
806 | */ | |
1a2142af | 807 | set_cpus_allowed_ptr(current, cpu_all_mask); |
1da177e4 | 808 | |
9ec52099 CLG |
809 | cad_pid = task_pid(current); |
810 | ||
ca74a6f8 | 811 | smp_prepare_cpus(setup_max_cpus); |
1da177e4 LT |
812 | |
813 | do_pre_smp_initcalls(); | |
004417a6 | 814 | lockup_detector_init(); |
1da177e4 | 815 | |
1da177e4 LT |
816 | smp_init(); |
817 | sched_init_smp(); | |
818 | ||
1da177e4 LT |
819 | do_basic_setup(); |
820 | ||
2bd3a997 EB |
821 | /* Open the /dev/console on the rootfs, this should never fail */ |
822 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
823 | printk(KERN_WARNING "Warning: unable to open an initial console.\n"); | |
824 | ||
825 | (void) sys_dup(0); | |
826 | (void) sys_dup(0); | |
1da177e4 LT |
827 | /* |
828 | * check if there is an early userspace init. If yes, let it do all | |
829 | * the work | |
830 | */ | |
ffdfc409 OJ |
831 | |
832 | if (!ramdisk_execute_command) | |
833 | ramdisk_execute_command = "/init"; | |
834 | ||
835 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
836 | ramdisk_execute_command = NULL; | |
1da177e4 | 837 | prepare_namespace(); |
ffdfc409 | 838 | } |
1da177e4 LT |
839 | |
840 | /* | |
841 | * Ok, we have completed the initial bootup, and | |
842 | * we're essentially up and running. Get rid of the | |
843 | * initmem segments and start the user-mode stuff.. | |
844 | */ | |
71566a0d | 845 | |
ee5bfa64 VG |
846 | init_post(); |
847 | return 0; | |
1da177e4 | 848 | } |