]>
Commit | Line | Data |
---|---|---|
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 | ||
12 | #include <linux/types.h> | |
13 | #include <linux/module.h> | |
14 | #include <linux/proc_fs.h> | |
15 | #include <linux/kernel.h> | |
16 | #include <linux/syscalls.h> | |
17 | #include <linux/stackprotector.h> | |
18 | #include <linux/string.h> | |
19 | #include <linux/ctype.h> | |
20 | #include <linux/delay.h> | |
21 | #include <linux/ioport.h> | |
22 | #include <linux/init.h> | |
23 | #include <linux/smp_lock.h> | |
24 | #include <linux/initrd.h> | |
25 | #include <linux/bootmem.h> | |
26 | #include <linux/acpi.h> | |
27 | #include <linux/tty.h> | |
28 | #include <linux/percpu.h> | |
29 | #include <linux/kmod.h> | |
30 | #include <linux/vmalloc.h> | |
31 | #include <linux/kernel_stat.h> | |
32 | #include <linux/start_kernel.h> | |
33 | #include <linux/security.h> | |
34 | #include <linux/smp.h> | |
35 | #include <linux/profile.h> | |
36 | #include <linux/rcupdate.h> | |
37 | #include <linux/moduleparam.h> | |
38 | #include <linux/kallsyms.h> | |
39 | #include <linux/writeback.h> | |
40 | #include <linux/cpu.h> | |
41 | #include <linux/cpuset.h> | |
42 | #include <linux/cgroup.h> | |
43 | #include <linux/efi.h> | |
44 | #include <linux/tick.h> | |
45 | #include <linux/interrupt.h> | |
46 | #include <linux/taskstats_kern.h> | |
47 | #include <linux/delayacct.h> | |
48 | #include <linux/unistd.h> | |
49 | #include <linux/rmap.h> | |
50 | #include <linux/mempolicy.h> | |
51 | #include <linux/key.h> | |
52 | #include <linux/buffer_head.h> | |
53 | #include <linux/page_cgroup.h> | |
54 | #include <linux/debug_locks.h> | |
55 | #include <linux/debugobjects.h> | |
56 | #include <linux/lockdep.h> | |
57 | #include <linux/kmemleak.h> | |
58 | #include <linux/pid_namespace.h> | |
59 | #include <linux/device.h> | |
60 | #include <linux/kthread.h> | |
61 | #include <linux/sched.h> | |
62 | #include <linux/signal.h> | |
63 | #include <linux/idr.h> | |
64 | #include <linux/kgdb.h> | |
65 | #include <linux/ftrace.h> | |
66 | #include <linux/async.h> | |
67 | #include <linux/kmemcheck.h> | |
68 | #include <linux/sfi.h> | |
69 | #include <linux/shmem_fs.h> | |
70 | #include <linux/slab.h> | |
71 | ||
72 | #include <asm/io.h> | |
73 | #include <asm/bugs.h> | |
74 | #include <asm/setup.h> | |
75 | #include <asm/sections.h> | |
76 | #include <asm/cacheflush.h> | |
77 | ||
78 | #ifdef CONFIG_X86_LOCAL_APIC | |
79 | #include <asm/smp.h> | |
80 | #endif | |
81 | ||
82 | static int kernel_init(void *); | |
83 | ||
84 | extern void init_IRQ(void); | |
85 | extern void fork_init(unsigned long); | |
86 | extern void mca_init(void); | |
87 | extern void sbus_init(void); | |
88 | extern void prio_tree_init(void); | |
89 | extern void radix_tree_init(void); | |
90 | extern void free_initmem(void); | |
91 | #ifndef CONFIG_DEBUG_RODATA | |
92 | static inline void mark_rodata_ro(void) { } | |
93 | #endif | |
94 | ||
95 | #ifdef CONFIG_TC | |
96 | extern void tc_init(void); | |
97 | #endif | |
98 | ||
99 | enum system_states system_state __read_mostly; | |
100 | EXPORT_SYMBOL(system_state); | |
101 | ||
102 | /* | |
103 | * Boot command-line arguments | |
104 | */ | |
105 | #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT | |
106 | #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT | |
107 | ||
108 | extern void time_init(void); | |
109 | /* Default late time init is NULL. archs can override this later. */ | |
110 | void (*__initdata late_time_init)(void); | |
111 | extern void softirq_init(void); | |
112 | ||
113 | /* Untouched command line saved by arch-specific code. */ | |
114 | char __initdata boot_command_line[COMMAND_LINE_SIZE]; | |
115 | /* Untouched saved command line (eg. for /proc) */ | |
116 | char *saved_command_line; | |
117 | /* Command line for parameter parsing */ | |
118 | static char *static_command_line; | |
119 | ||
120 | static char *execute_command; | |
121 | static char *ramdisk_execute_command; | |
122 | ||
123 | #ifdef CONFIG_SMP | |
124 | /* Setup configured maximum number of CPUs to activate */ | |
125 | unsigned int setup_max_cpus = NR_CPUS; | |
126 | EXPORT_SYMBOL(setup_max_cpus); | |
127 | ||
128 | ||
129 | /* | |
130 | * Setup routine for controlling SMP activation | |
131 | * | |
132 | * Command-line option of "nosmp" or "maxcpus=0" will disable SMP | |
133 | * activation entirely (the MPS table probe still happens, though). | |
134 | * | |
135 | * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer | |
136 | * greater than 0, limits the maximum number of CPUs activated in | |
137 | * SMP mode to <NUM>. | |
138 | */ | |
139 | ||
140 | void __weak arch_disable_smp_support(void) { } | |
141 | ||
142 | static int __init nosmp(char *str) | |
143 | { | |
144 | setup_max_cpus = 0; | |
145 | arch_disable_smp_support(); | |
146 | ||
147 | return 0; | |
148 | } | |
149 | ||
150 | early_param("nosmp", nosmp); | |
151 | ||
152 | /* this is hard limit */ | |
153 | static int __init nrcpus(char *str) | |
154 | { | |
155 | int nr_cpus; | |
156 | ||
157 | get_option(&str, &nr_cpus); | |
158 | if (nr_cpus > 0 && nr_cpus < nr_cpu_ids) | |
159 | nr_cpu_ids = nr_cpus; | |
160 | ||
161 | return 0; | |
162 | } | |
163 | ||
164 | early_param("nr_cpus", nrcpus); | |
165 | ||
166 | static int __init maxcpus(char *str) | |
167 | { | |
168 | get_option(&str, &setup_max_cpus); | |
169 | if (setup_max_cpus == 0) | |
170 | arch_disable_smp_support(); | |
171 | ||
172 | return 0; | |
173 | } | |
174 | ||
175 | early_param("maxcpus", maxcpus); | |
176 | #else | |
177 | static const unsigned int setup_max_cpus = NR_CPUS; | |
178 | #endif | |
179 | ||
180 | /* | |
181 | * If set, this is an indication to the drivers that reset the underlying | |
182 | * device before going ahead with the initialization otherwise driver might | |
183 | * rely on the BIOS and skip the reset operation. | |
184 | * | |
185 | * This is useful if kernel is booting in an unreliable environment. | |
186 | * For ex. kdump situaiton where previous kernel has crashed, BIOS has been | |
187 | * skipped and devices will be in unknown state. | |
188 | */ | |
189 | unsigned int reset_devices; | |
190 | EXPORT_SYMBOL(reset_devices); | |
191 | ||
192 | static int __init set_reset_devices(char *str) | |
193 | { | |
194 | reset_devices = 1; | |
195 | return 1; | |
196 | } | |
197 | ||
198 | __setup("reset_devices", set_reset_devices); | |
199 | ||
200 | static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, }; | |
201 | char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, }; | |
202 | static const char *panic_later, *panic_param; | |
203 | ||
204 | extern struct obs_kernel_param __setup_start[], __setup_end[]; | |
205 | ||
206 | static int __init obsolete_checksetup(char *line) | |
207 | { | |
208 | struct obs_kernel_param *p; | |
209 | int had_early_param = 0; | |
210 | ||
211 | p = __setup_start; | |
212 | do { | |
213 | int n = strlen(p->str); | |
214 | if (!strncmp(line, p->str, n)) { | |
215 | if (p->early) { | |
216 | /* Already done in parse_early_param? | |
217 | * (Needs exact match on param part). | |
218 | * Keep iterating, as we can have early | |
219 | * params and __setups of same names 8( */ | |
220 | if (line[n] == '\0' || line[n] == '=') | |
221 | had_early_param = 1; | |
222 | } else if (!p->setup_func) { | |
223 | printk(KERN_WARNING "Parameter %s is obsolete," | |
224 | " ignored\n", p->str); | |
225 | return 1; | |
226 | } else if (p->setup_func(line + n)) | |
227 | return 1; | |
228 | } | |
229 | p++; | |
230 | } while (p < __setup_end); | |
231 | ||
232 | return had_early_param; | |
233 | } | |
234 | ||
235 | /* | |
236 | * This should be approx 2 Bo*oMips to start (note initial shift), and will | |
237 | * still work even if initially too large, it will just take slightly longer | |
238 | */ | |
239 | unsigned long loops_per_jiffy = (1<<12); | |
240 | ||
241 | EXPORT_SYMBOL(loops_per_jiffy); | |
242 | ||
243 | static int __init debug_kernel(char *str) | |
244 | { | |
245 | console_loglevel = 10; | |
246 | return 0; | |
247 | } | |
248 | ||
249 | static int __init quiet_kernel(char *str) | |
250 | { | |
251 | console_loglevel = 4; | |
252 | return 0; | |
253 | } | |
254 | ||
255 | early_param("debug", debug_kernel); | |
256 | early_param("quiet", quiet_kernel); | |
257 | ||
258 | static int __init loglevel(char *str) | |
259 | { | |
260 | get_option(&str, &console_loglevel); | |
261 | return 0; | |
262 | } | |
263 | ||
264 | early_param("loglevel", loglevel); | |
265 | ||
266 | /* | |
267 | * Unknown boot options get handed to init, unless they look like | |
268 | * unused parameters (modprobe will find them in /proc/cmdline). | |
269 | */ | |
270 | static int __init unknown_bootoption(char *param, char *val) | |
271 | { | |
272 | /* Change NUL term back to "=", to make "param" the whole string. */ | |
273 | if (val) { | |
274 | /* param=val or param="val"? */ | |
275 | if (val == param+strlen(param)+1) | |
276 | val[-1] = '='; | |
277 | else if (val == param+strlen(param)+2) { | |
278 | val[-2] = '='; | |
279 | memmove(val-1, val, strlen(val)+1); | |
280 | val--; | |
281 | } else | |
282 | BUG(); | |
283 | } | |
284 | ||
285 | /* Handle obsolete-style parameters */ | |
286 | if (obsolete_checksetup(param)) | |
287 | return 0; | |
288 | ||
289 | /* Unused module parameter. */ | |
290 | if (strchr(param, '.') && (!val || strchr(param, '.') < val)) | |
291 | return 0; | |
292 | ||
293 | if (panic_later) | |
294 | return 0; | |
295 | ||
296 | if (val) { | |
297 | /* Environment option */ | |
298 | unsigned int i; | |
299 | for (i = 0; envp_init[i]; i++) { | |
300 | if (i == MAX_INIT_ENVS) { | |
301 | panic_later = "Too many boot env vars at `%s'"; | |
302 | panic_param = param; | |
303 | } | |
304 | if (!strncmp(param, envp_init[i], val - param)) | |
305 | break; | |
306 | } | |
307 | envp_init[i] = param; | |
308 | } else { | |
309 | /* Command line option */ | |
310 | unsigned int i; | |
311 | for (i = 0; argv_init[i]; i++) { | |
312 | if (i == MAX_INIT_ARGS) { | |
313 | panic_later = "Too many boot init vars at `%s'"; | |
314 | panic_param = param; | |
315 | } | |
316 | } | |
317 | argv_init[i] = param; | |
318 | } | |
319 | return 0; | |
320 | } | |
321 | ||
322 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
323 | int __read_mostly debug_pagealloc_enabled = 0; | |
324 | #endif | |
325 | ||
326 | static int __init init_setup(char *str) | |
327 | { | |
328 | unsigned int i; | |
329 | ||
330 | execute_command = str; | |
331 | /* | |
332 | * In case LILO is going to boot us with default command line, | |
333 | * it prepends "auto" before the whole cmdline which makes | |
334 | * the shell think it should execute a script with such name. | |
335 | * So we ignore all arguments entered _before_ init=... [MJ] | |
336 | */ | |
337 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
338 | argv_init[i] = NULL; | |
339 | return 1; | |
340 | } | |
341 | __setup("init=", init_setup); | |
342 | ||
343 | static int __init rdinit_setup(char *str) | |
344 | { | |
345 | unsigned int i; | |
346 | ||
347 | ramdisk_execute_command = str; | |
348 | /* See "auto" comment in init_setup */ | |
349 | for (i = 1; i < MAX_INIT_ARGS; i++) | |
350 | argv_init[i] = NULL; | |
351 | return 1; | |
352 | } | |
353 | __setup("rdinit=", rdinit_setup); | |
354 | ||
355 | #ifndef CONFIG_SMP | |
356 | ||
357 | #ifdef CONFIG_X86_LOCAL_APIC | |
358 | static void __init smp_init(void) | |
359 | { | |
360 | APIC_init_uniprocessor(); | |
361 | } | |
362 | #else | |
363 | #define smp_init() do { } while (0) | |
364 | #endif | |
365 | ||
366 | static inline void setup_nr_cpu_ids(void) { } | |
367 | static inline void smp_prepare_cpus(unsigned int maxcpus) { } | |
368 | ||
369 | #else | |
370 | ||
371 | /* Setup number of possible processor ids */ | |
372 | int nr_cpu_ids __read_mostly = NR_CPUS; | |
373 | EXPORT_SYMBOL(nr_cpu_ids); | |
374 | ||
375 | /* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */ | |
376 | static void __init setup_nr_cpu_ids(void) | |
377 | { | |
378 | nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1; | |
379 | } | |
380 | ||
381 | /* Called by boot processor to activate the rest. */ | |
382 | static void __init smp_init(void) | |
383 | { | |
384 | unsigned int cpu; | |
385 | ||
386 | /* FIXME: This should be done in userspace --RR */ | |
387 | for_each_present_cpu(cpu) { | |
388 | if (num_online_cpus() >= setup_max_cpus) | |
389 | break; | |
390 | if (!cpu_online(cpu)) | |
391 | cpu_up(cpu); | |
392 | } | |
393 | ||
394 | /* Any cleanup work */ | |
395 | printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus()); | |
396 | smp_cpus_done(setup_max_cpus); | |
397 | } | |
398 | ||
399 | #endif | |
400 | ||
401 | /* | |
402 | * We need to store the untouched command line for future reference. | |
403 | * We also need to store the touched command line since the parameter | |
404 | * parsing is performed in place, and we should allow a component to | |
405 | * store reference of name/value for future reference. | |
406 | */ | |
407 | static void __init setup_command_line(char *command_line) | |
408 | { | |
409 | saved_command_line = alloc_bootmem(strlen (boot_command_line)+1); | |
410 | static_command_line = alloc_bootmem(strlen (command_line)+1); | |
411 | strcpy (saved_command_line, boot_command_line); | |
412 | strcpy (static_command_line, command_line); | |
413 | } | |
414 | ||
415 | /* | |
416 | * We need to finalize in a non-__init function or else race conditions | |
417 | * between the root thread and the init thread may cause start_kernel to | |
418 | * be reaped by free_initmem before the root thread has proceeded to | |
419 | * cpu_idle. | |
420 | * | |
421 | * gcc-3.4 accidentally inlines this function, so use noinline. | |
422 | */ | |
423 | ||
424 | static __initdata DECLARE_COMPLETION(kthreadd_done); | |
425 | ||
426 | static noinline void __init_refok rest_init(void) | |
427 | __releases(kernel_lock) | |
428 | { | |
429 | int pid; | |
430 | ||
431 | rcu_scheduler_starting(); | |
432 | /* | |
433 | * We need to spawn init first so that it obtains pid 1, however | |
434 | * the init task will end up wanting to create kthreads, which, if | |
435 | * we schedule it before we create kthreadd, will OOPS. | |
436 | */ | |
437 | kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND); | |
438 | numa_default_policy(); | |
439 | pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES); | |
440 | rcu_read_lock(); | |
441 | kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns); | |
442 | rcu_read_unlock(); | |
443 | complete(&kthreadd_done); | |
444 | ||
445 | /* | |
446 | * The boot idle thread must execute schedule() | |
447 | * at least once to get things moving: | |
448 | */ | |
449 | init_idle_bootup_task(current); | |
450 | preempt_enable_no_resched(); | |
451 | schedule(); | |
452 | preempt_disable(); | |
453 | ||
454 | /* Call into cpu_idle with preempt disabled */ | |
455 | cpu_idle(); | |
456 | } | |
457 | ||
458 | /* Check for early params. */ | |
459 | static int __init do_early_param(char *param, char *val) | |
460 | { | |
461 | struct obs_kernel_param *p; | |
462 | ||
463 | for (p = __setup_start; p < __setup_end; p++) { | |
464 | if ((p->early && strcmp(param, p->str) == 0) || | |
465 | (strcmp(param, "console") == 0 && | |
466 | strcmp(p->str, "earlycon") == 0) | |
467 | ) { | |
468 | if (p->setup_func(val) != 0) | |
469 | printk(KERN_WARNING | |
470 | "Malformed early option '%s'\n", param); | |
471 | } | |
472 | } | |
473 | /* We accept everything at this stage. */ | |
474 | return 0; | |
475 | } | |
476 | ||
477 | void __init parse_early_options(char *cmdline) | |
478 | { | |
479 | parse_args("early options", cmdline, NULL, 0, do_early_param); | |
480 | } | |
481 | ||
482 | /* Arch code calls this early on, or if not, just before other parsing. */ | |
483 | void __init parse_early_param(void) | |
484 | { | |
485 | static __initdata int done = 0; | |
486 | static __initdata char tmp_cmdline[COMMAND_LINE_SIZE]; | |
487 | ||
488 | if (done) | |
489 | return; | |
490 | ||
491 | /* All fall through to do_early_param. */ | |
492 | strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE); | |
493 | parse_early_options(tmp_cmdline); | |
494 | done = 1; | |
495 | } | |
496 | ||
497 | /* | |
498 | * Activate the first processor. | |
499 | */ | |
500 | ||
501 | static void __init boot_cpu_init(void) | |
502 | { | |
503 | int cpu = smp_processor_id(); | |
504 | /* Mark the boot cpu "present", "online" etc for SMP and UP case */ | |
505 | set_cpu_online(cpu, true); | |
506 | set_cpu_active(cpu, true); | |
507 | set_cpu_present(cpu, true); | |
508 | set_cpu_possible(cpu, true); | |
509 | } | |
510 | ||
511 | void __init __weak smp_setup_processor_id(void) | |
512 | { | |
513 | } | |
514 | ||
515 | void __init __weak thread_info_cache_init(void) | |
516 | { | |
517 | } | |
518 | ||
519 | /* | |
520 | * Set up kernel memory allocators | |
521 | */ | |
522 | static void __init mm_init(void) | |
523 | { | |
524 | /* | |
525 | * page_cgroup requires countinous pages as memmap | |
526 | * and it's bigger than MAX_ORDER unless SPARSEMEM. | |
527 | */ | |
528 | page_cgroup_init_flatmem(); | |
529 | mem_init(); | |
530 | kmem_cache_init(); | |
531 | percpu_init_late(); | |
532 | pgtable_cache_init(); | |
533 | vmalloc_init(); | |
534 | } | |
535 | ||
536 | asmlinkage void __init start_kernel(void) | |
537 | { | |
538 | char * command_line; | |
539 | extern struct kernel_param __start___param[], __stop___param[]; | |
540 | ||
541 | smp_setup_processor_id(); | |
542 | ||
543 | /* | |
544 | * Need to run as early as possible, to initialize the | |
545 | * lockdep hash: | |
546 | */ | |
547 | lockdep_init(); | |
548 | debug_objects_early_init(); | |
549 | ||
550 | /* | |
551 | * Set up the the initial canary ASAP: | |
552 | */ | |
553 | boot_init_stack_canary(); | |
554 | ||
555 | cgroup_init_early(); | |
556 | ||
557 | local_irq_disable(); | |
558 | early_boot_irqs_off(); | |
559 | early_init_irq_lock_class(); | |
560 | ||
561 | /* | |
562 | * Interrupts are still disabled. Do necessary setups, then | |
563 | * enable them | |
564 | */ | |
565 | tick_init(); | |
566 | boot_cpu_init(); | |
567 | page_address_init(); | |
568 | printk(KERN_NOTICE "%s", linux_banner); | |
569 | setup_arch(&command_line); | |
570 | mm_init_owner(&init_mm, &init_task); | |
571 | setup_command_line(command_line); | |
572 | setup_nr_cpu_ids(); | |
573 | setup_per_cpu_areas(); | |
574 | smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */ | |
575 | ||
576 | build_all_zonelists(NULL); | |
577 | page_alloc_init(); | |
578 | ||
579 | printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line); | |
580 | parse_early_param(); | |
581 | parse_args("Booting kernel", static_command_line, __start___param, | |
582 | __stop___param - __start___param, | |
583 | &unknown_bootoption); | |
584 | /* | |
585 | * These use large bootmem allocations and must precede | |
586 | * kmem_cache_init() | |
587 | */ | |
588 | pidhash_init(); | |
589 | vfs_caches_init_early(); | |
590 | sort_main_extable(); | |
591 | trap_init(); | |
592 | mm_init(); | |
593 | /* | |
594 | * Set up the scheduler prior starting any interrupts (such as the | |
595 | * timer interrupt). Full topology setup happens at smp_init() | |
596 | * time - but meanwhile we still have a functioning scheduler. | |
597 | */ | |
598 | sched_init(); | |
599 | /* | |
600 | * Disable preemption - early bootup scheduling is extremely | |
601 | * fragile until we cpu_idle() for the first time. | |
602 | */ | |
603 | preempt_disable(); | |
604 | if (!irqs_disabled()) { | |
605 | printk(KERN_WARNING "start_kernel(): bug: interrupts were " | |
606 | "enabled *very* early, fixing it\n"); | |
607 | local_irq_disable(); | |
608 | } | |
609 | rcu_init(); | |
610 | radix_tree_init(); | |
611 | /* init some links before init_ISA_irqs() */ | |
612 | early_irq_init(); | |
613 | init_IRQ(); | |
614 | prio_tree_init(); | |
615 | init_timers(); | |
616 | hrtimers_init(); | |
617 | softirq_init(); | |
618 | timekeeping_init(); | |
619 | time_init(); | |
620 | profile_init(); | |
621 | if (!irqs_disabled()) | |
622 | printk(KERN_CRIT "start_kernel(): bug: interrupts were " | |
623 | "enabled early\n"); | |
624 | early_boot_irqs_on(); | |
625 | local_irq_enable(); | |
626 | ||
627 | /* Interrupts are enabled now so all GFP allocations are safe. */ | |
628 | gfp_allowed_mask = __GFP_BITS_MASK; | |
629 | ||
630 | kmem_cache_init_late(); | |
631 | ||
632 | /* | |
633 | * HACK ALERT! This is early. We're enabling the console before | |
634 | * we've done PCI setups etc, and console_init() must be aware of | |
635 | * this. But we do want output early, in case something goes wrong. | |
636 | */ | |
637 | console_init(); | |
638 | if (panic_later) | |
639 | panic(panic_later, panic_param); | |
640 | ||
641 | lockdep_info(); | |
642 | ||
643 | /* | |
644 | * Need to run this when irqs are enabled, because it wants | |
645 | * to self-test [hard/soft]-irqs on/off lock inversion bugs | |
646 | * too: | |
647 | */ | |
648 | locking_selftest(); | |
649 | ||
650 | #ifdef CONFIG_BLK_DEV_INITRD | |
651 | if (initrd_start && !initrd_below_start_ok && | |
652 | page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) { | |
653 | printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - " | |
654 | "disabling it.\n", | |
655 | page_to_pfn(virt_to_page((void *)initrd_start)), | |
656 | min_low_pfn); | |
657 | initrd_start = 0; | |
658 | } | |
659 | #endif | |
660 | page_cgroup_init(); | |
661 | enable_debug_pagealloc(); | |
662 | kmemleak_init(); | |
663 | debug_objects_mem_init(); | |
664 | idr_init_cache(); | |
665 | setup_per_cpu_pageset(); | |
666 | numa_policy_init(); | |
667 | if (late_time_init) | |
668 | late_time_init(); | |
669 | sched_clock_init(); | |
670 | calibrate_delay(); | |
671 | pidmap_init(); | |
672 | anon_vma_init(); | |
673 | #ifdef CONFIG_X86 | |
674 | if (efi_enabled) | |
675 | efi_enter_virtual_mode(); | |
676 | #endif | |
677 | thread_info_cache_init(); | |
678 | cred_init(); | |
679 | fork_init(totalram_pages); | |
680 | proc_caches_init(); | |
681 | buffer_init(); | |
682 | key_init(); | |
683 | security_init(); | |
684 | dbg_late_init(); | |
685 | vfs_caches_init(totalram_pages); | |
686 | signals_init(); | |
687 | /* rootfs populating might need page-writeback */ | |
688 | page_writeback_init(); | |
689 | #ifdef CONFIG_PROC_FS | |
690 | proc_root_init(); | |
691 | #endif | |
692 | cgroup_init(); | |
693 | cpuset_init(); | |
694 | taskstats_init_early(); | |
695 | delayacct_init(); | |
696 | ||
697 | check_bugs(); | |
698 | ||
699 | acpi_early_init(); /* before LAPIC and SMP init */ | |
700 | sfi_init_late(); | |
701 | ||
702 | ftrace_init(); | |
703 | ||
704 | /* Do the rest non-__init'ed, we're now alive */ | |
705 | rest_init(); | |
706 | } | |
707 | ||
708 | /* Call all constructor functions linked into the kernel. */ | |
709 | static void __init do_ctors(void) | |
710 | { | |
711 | #ifdef CONFIG_CONSTRUCTORS | |
712 | ctor_fn_t *fn = (ctor_fn_t *) __ctors_start; | |
713 | ||
714 | for (; fn < (ctor_fn_t *) __ctors_end; fn++) | |
715 | (*fn)(); | |
716 | #endif | |
717 | } | |
718 | ||
719 | int initcall_debug; | |
720 | core_param(initcall_debug, initcall_debug, bool, 0644); | |
721 | ||
722 | static char msgbuf[64]; | |
723 | ||
724 | static int do_one_initcall_debug(initcall_t fn) | |
725 | { | |
726 | ktime_t calltime, delta, rettime; | |
727 | unsigned long long duration; | |
728 | int ret; | |
729 | ||
730 | printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current)); | |
731 | calltime = ktime_get(); | |
732 | ret = fn(); | |
733 | rettime = ktime_get(); | |
734 | delta = ktime_sub(rettime, calltime); | |
735 | duration = (unsigned long long) ktime_to_ns(delta) >> 10; | |
736 | printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn, | |
737 | ret, duration); | |
738 | ||
739 | return ret; | |
740 | } | |
741 | ||
742 | int do_one_initcall(initcall_t fn) | |
743 | { | |
744 | int count = preempt_count(); | |
745 | int ret; | |
746 | ||
747 | if (initcall_debug) | |
748 | ret = do_one_initcall_debug(fn); | |
749 | else | |
750 | ret = fn(); | |
751 | ||
752 | msgbuf[0] = 0; | |
753 | ||
754 | if (ret && ret != -ENODEV && initcall_debug) | |
755 | sprintf(msgbuf, "error code %d ", ret); | |
756 | ||
757 | if (preempt_count() != count) { | |
758 | strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf)); | |
759 | preempt_count() = count; | |
760 | } | |
761 | if (irqs_disabled()) { | |
762 | strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf)); | |
763 | local_irq_enable(); | |
764 | } | |
765 | if (msgbuf[0]) { | |
766 | printk("initcall %pF returned with %s\n", fn, msgbuf); | |
767 | } | |
768 | ||
769 | return ret; | |
770 | } | |
771 | ||
772 | ||
773 | extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[]; | |
774 | ||
775 | static void __init do_initcalls(void) | |
776 | { | |
777 | initcall_t *fn; | |
778 | ||
779 | for (fn = __early_initcall_end; fn < __initcall_end; fn++) | |
780 | do_one_initcall(*fn); | |
781 | ||
782 | /* Make sure there is no pending stuff from the initcall sequence */ | |
783 | flush_scheduled_work(); | |
784 | } | |
785 | ||
786 | /* | |
787 | * Ok, the machine is now initialized. None of the devices | |
788 | * have been touched yet, but the CPU subsystem is up and | |
789 | * running, and memory and process management works. | |
790 | * | |
791 | * Now we can finally start doing some real work.. | |
792 | */ | |
793 | static void __init do_basic_setup(void) | |
794 | { | |
795 | cpuset_init_smp(); | |
796 | usermodehelper_init(); | |
797 | init_tmpfs(); | |
798 | driver_init(); | |
799 | init_irq_proc(); | |
800 | do_ctors(); | |
801 | do_initcalls(); | |
802 | } | |
803 | ||
804 | static void __init do_pre_smp_initcalls(void) | |
805 | { | |
806 | initcall_t *fn; | |
807 | ||
808 | for (fn = __initcall_start; fn < __early_initcall_end; fn++) | |
809 | do_one_initcall(*fn); | |
810 | } | |
811 | ||
812 | static void run_init_process(char *init_filename) | |
813 | { | |
814 | argv_init[0] = init_filename; | |
815 | kernel_execve(init_filename, argv_init, envp_init); | |
816 | } | |
817 | ||
818 | /* This is a non __init function. Force it to be noinline otherwise gcc | |
819 | * makes it inline to init() and it becomes part of init.text section | |
820 | */ | |
821 | static noinline int init_post(void) | |
822 | __releases(kernel_lock) | |
823 | { | |
824 | /* need to finish all async __init code before freeing the memory */ | |
825 | async_synchronize_full(); | |
826 | free_initmem(); | |
827 | mark_rodata_ro(); | |
828 | system_state = SYSTEM_RUNNING; | |
829 | numa_default_policy(); | |
830 | ||
831 | ||
832 | current->signal->flags |= SIGNAL_UNKILLABLE; | |
833 | ||
834 | if (ramdisk_execute_command) { | |
835 | run_init_process(ramdisk_execute_command); | |
836 | printk(KERN_WARNING "Failed to execute %s\n", | |
837 | ramdisk_execute_command); | |
838 | } | |
839 | ||
840 | /* | |
841 | * We try each of these until one succeeds. | |
842 | * | |
843 | * The Bourne shell can be used instead of init if we are | |
844 | * trying to recover a really broken machine. | |
845 | */ | |
846 | if (execute_command) { | |
847 | run_init_process(execute_command); | |
848 | printk(KERN_WARNING "Failed to execute %s. Attempting " | |
849 | "defaults...\n", execute_command); | |
850 | } | |
851 | run_init_process("/sbin/init"); | |
852 | run_init_process("/etc/init"); | |
853 | run_init_process("/bin/init"); | |
854 | run_init_process("/bin/sh"); | |
855 | ||
856 | panic("No init found. Try passing init= option to kernel. " | |
857 | "See Linux Documentation/init.txt for guidance."); | |
858 | } | |
859 | ||
860 | static int __init kernel_init(void * unused) | |
861 | { | |
862 | /* | |
863 | * Wait until kthreadd is all set-up. | |
864 | */ | |
865 | wait_for_completion(&kthreadd_done); | |
866 | /* | |
867 | * init can allocate pages on any node | |
868 | */ | |
869 | set_mems_allowed(node_states[N_HIGH_MEMORY]); | |
870 | /* | |
871 | * init can run on any cpu. | |
872 | */ | |
873 | set_cpus_allowed_ptr(current, cpu_all_mask); | |
874 | /* | |
875 | * Tell the world that we're going to be the grim | |
876 | * reaper of innocent orphaned children. | |
877 | * | |
878 | * We don't want people to have to make incorrect | |
879 | * assumptions about where in the task array this | |
880 | * can be found. | |
881 | */ | |
882 | init_pid_ns.child_reaper = current; | |
883 | ||
884 | cad_pid = task_pid(current); | |
885 | ||
886 | smp_prepare_cpus(setup_max_cpus); | |
887 | ||
888 | do_pre_smp_initcalls(); | |
889 | ||
890 | smp_init(); | |
891 | sched_init_smp(); | |
892 | ||
893 | do_basic_setup(); | |
894 | ||
895 | /* Open the /dev/console on the rootfs, this should never fail */ | |
896 | if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) | |
897 | printk(KERN_WARNING "Warning: unable to open an initial console.\n"); | |
898 | ||
899 | (void) sys_dup(0); | |
900 | (void) sys_dup(0); | |
901 | /* | |
902 | * check if there is an early userspace init. If yes, let it do all | |
903 | * the work | |
904 | */ | |
905 | ||
906 | if (!ramdisk_execute_command) | |
907 | ramdisk_execute_command = "/init"; | |
908 | ||
909 | if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { | |
910 | ramdisk_execute_command = NULL; | |
911 | prepare_namespace(); | |
912 | } | |
913 | ||
914 | /* | |
915 | * Ok, we have completed the initial bootup, and | |
916 | * we're essentially up and running. Get rid of the | |
917 | * initmem segments and start the user-mode stuff.. | |
918 | */ | |
919 | ||
920 | init_post(); | |
921 | return 0; | |
922 | } |