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/initrd.h>
24 #include <linux/bootmem.h>
25 #include <linux/acpi.h>
26 #include <linux/tty.h>
27 #include <linux/percpu.h>
28 #include <linux/kmod.h>
29 #include <linux/vmalloc.h>
30 #include <linux/kernel_stat.h>
31 #include <linux/start_kernel.h>
32 #include <linux/security.h>
33 #include <linux/smp.h>
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>
41 #include <linux/cgroup.h>
42 #include <linux/efi.h>
43 #include <linux/tick.h>
44 #include <linux/interrupt.h>
45 #include <linux/taskstats_kern.h>
46 #include <linux/delayacct.h>
47 #include <linux/unistd.h>
48 #include <linux/rmap.h>
49 #include <linux/mempolicy.h>
50 #include <linux/key.h>
51 #include <linux/buffer_head.h>
52 #include <linux/page_cgroup.h>
53 #include <linux/debug_locks.h>
54 #include <linux/debugobjects.h>
55 #include <linux/lockdep.h>
56 #include <linux/kmemleak.h>
57 #include <linux/pid_namespace.h>
58 #include <linux/device.h>
59 #include <linux/kthread.h>
60 #include <linux/sched.h>
61 #include <linux/signal.h>
62 #include <linux/idr.h>
63 #include <linux/kgdb.h>
64 #include <linux/ftrace.h>
65 #include <linux/async.h>
66 #include <linux/kmemcheck.h>
67 #include <linux/sfi.h>
68 #include <linux/shmem_fs.h>
69 #include <linux/slab.h>
70 #include <linux/perf_event.h>
71 #include <linux/random.h>
72 
73 #include <asm/io.h>
74 #include <asm/bugs.h>
75 #include <asm/setup.h>
76 #include <asm/sections.h>
77 #include <asm/cacheflush.h>
78 
79 #ifdef CONFIG_X86_LOCAL_APIC
80 #include <asm/smp.h>
81 #endif
82 
83 static int kernel_init(void *);
84 
85 extern void init_IRQ(void);
86 extern void fork_init(unsigned long);
87 extern void mca_init(void);
88 extern void sbus_init(void);
89 extern void prio_tree_init(void);
90 extern void radix_tree_init(void);
91 #ifndef CONFIG_DEBUG_RODATA
mark_rodata_ro(void)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 /*
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 
108 enum system_states system_state __read_mostly;
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. */
119 void (*__initdata late_time_init)(void);
120 extern void softirq_init(void);
121 
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;
128 
129 static char *execute_command;
130 static char *ramdisk_execute_command;
131 
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);
143 
set_reset_devices(char * str)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 
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, };
154 static const char *panic_later, *panic_param;
155 
156 extern const struct obs_kernel_param __setup_start[], __setup_end[];
157 
obsolete_checksetup(char * line)158 static int __init obsolete_checksetup(char *line)
159 {
160 	const struct obs_kernel_param *p;
161 	int had_early_param = 0;
162 
163 	p = __setup_start;
164 	do {
165 		int n = strlen(p->str);
166 		if (parameqn(line, p->str, n)) {
167 			if (p->early) {
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( */
172 				if (line[n] == '\0' || line[n] == '=')
173 					had_early_param = 1;
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);
183 
184 	return had_early_param;
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 
debug_kernel(char * str)195 static int __init debug_kernel(char *str)
196 {
197 	console_loglevel = 10;
198 	return 0;
199 }
200 
quiet_kernel(char * str)201 static int __init quiet_kernel(char *str)
202 {
203 	console_loglevel = 4;
204 	return 0;
205 }
206 
207 early_param("debug", debug_kernel);
208 early_param("quiet", quiet_kernel);
209 
loglevel(char * str)210 static int __init loglevel(char *str)
211 {
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;
225 }
226 
227 early_param("loglevel", loglevel);
228 
229 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)230 static int __init repair_env_string(char *param, char *val)
231 {
232 	if (val) {
233 		/* param=val or param="val"? */
234 		if (val == param+strlen(param)+1)
235 			val[-1] = '=';
236 		else if (val == param+strlen(param)+2) {
237 			val[-2] = '=';
238 			memmove(val-1, val, strlen(val)+1);
239 			val--;
240 		} else
241 			BUG();
242 	}
243 	return 0;
244 }
245 
246 /*
247  * Unknown boot options get handed to init, unless they look like
248  * unused parameters (modprobe will find them in /proc/cmdline).
249  */
unknown_bootoption(char * param,char * val)250 static int __init unknown_bootoption(char *param, char *val)
251 {
252 	repair_env_string(param, val);
253 
254 	/* Handle obsolete-style parameters */
255 	if (obsolete_checksetup(param))
256 		return 0;
257 
258 	/* Unused module parameter. */
259 	if (strchr(param, '.') && (!val || strchr(param, '.') < val))
260 		return 0;
261 
262 	if (panic_later)
263 		return 0;
264 
265 	if (val) {
266 		/* Environment option */
267 		unsigned int i;
268 		for (i = 0; envp_init[i]; i++) {
269 			if (i == MAX_INIT_ENVS) {
270 				panic_later = "Too many boot env vars at `%s'";
271 				panic_param = param;
272 			}
273 			if (!strncmp(param, envp_init[i], val - param))
274 				break;
275 		}
276 		envp_init[i] = param;
277 	} else {
278 		/* Command line option */
279 		unsigned int i;
280 		for (i = 0; argv_init[i]; i++) {
281 			if (i == MAX_INIT_ARGS) {
282 				panic_later = "Too many boot init vars at `%s'";
283 				panic_param = param;
284 			}
285 		}
286 		argv_init[i] = param;
287 	}
288 	return 0;
289 }
290 
init_setup(char * str)291 static int __init init_setup(char *str)
292 {
293 	unsigned int i;
294 
295 	execute_command = str;
296 	/*
297 	 * In case LILO is going to boot us with default command line,
298 	 * it prepends "auto" before the whole cmdline which makes
299 	 * the shell think it should execute a script with such name.
300 	 * So we ignore all arguments entered _before_ init=... [MJ]
301 	 */
302 	for (i = 1; i < MAX_INIT_ARGS; i++)
303 		argv_init[i] = NULL;
304 	return 1;
305 }
306 __setup("init=", init_setup);
307 
rdinit_setup(char * str)308 static int __init rdinit_setup(char *str)
309 {
310 	unsigned int i;
311 
312 	ramdisk_execute_command = str;
313 	/* See "auto" comment in init_setup */
314 	for (i = 1; i < MAX_INIT_ARGS; i++)
315 		argv_init[i] = NULL;
316 	return 1;
317 }
318 __setup("rdinit=", rdinit_setup);
319 
320 #ifndef CONFIG_SMP
321 static const unsigned int setup_max_cpus = NR_CPUS;
322 #ifdef CONFIG_X86_LOCAL_APIC
smp_init(void)323 static void __init smp_init(void)
324 {
325 	APIC_init_uniprocessor();
326 }
327 #else
328 #define smp_init()	do { } while (0)
329 #endif
330 
setup_nr_cpu_ids(void)331 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)332 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
333 #endif
334 
335 /*
336  * We need to store the untouched command line for future reference.
337  * We also need to store the touched command line since the parameter
338  * parsing is performed in place, and we should allow a component to
339  * store reference of name/value for future reference.
340  */
setup_command_line(char * command_line)341 static void __init setup_command_line(char *command_line)
342 {
343 	saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
344 	static_command_line = alloc_bootmem(strlen (command_line)+1);
345 	strcpy (saved_command_line, boot_command_line);
346 	strcpy (static_command_line, command_line);
347 }
348 
349 /*
350  * We need to finalize in a non-__init function or else race conditions
351  * between the root thread and the init thread may cause start_kernel to
352  * be reaped by free_initmem before the root thread has proceeded to
353  * cpu_idle.
354  *
355  * gcc-3.4 accidentally inlines this function, so use noinline.
356  */
357 
358 static __initdata DECLARE_COMPLETION(kthreadd_done);
359 
rest_init(void)360 static noinline void __init_refok rest_init(void)
361 {
362 	int pid;
363 
364 	rcu_scheduler_starting();
365 	/*
366 	 * We need to spawn init first so that it obtains pid 1, however
367 	 * the init task will end up wanting to create kthreads, which, if
368 	 * we schedule it before we create kthreadd, will OOPS.
369 	 */
370 	kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
371 	numa_default_policy();
372 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
373 	rcu_read_lock();
374 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
375 	rcu_read_unlock();
376 	complete(&kthreadd_done);
377 
378 	/*
379 	 * The boot idle thread must execute schedule()
380 	 * at least once to get things moving:
381 	 */
382 	init_idle_bootup_task(current);
383 	schedule_preempt_disabled();
384 	/* Call into cpu_idle with preempt disabled */
385 	cpu_idle();
386 }
387 
388 /* Check for early params. */
do_early_param(char * param,char * val)389 static int __init do_early_param(char *param, char *val)
390 {
391 	const struct obs_kernel_param *p;
392 
393 	for (p = __setup_start; p < __setup_end; p++) {
394 		if ((p->early && parameq(param, p->str)) ||
395 		    (strcmp(param, "console") == 0 &&
396 		     strcmp(p->str, "earlycon") == 0)
397 		) {
398 			if (p->setup_func(val) != 0)
399 				printk(KERN_WARNING
400 				       "Malformed early option '%s'\n", param);
401 		}
402 	}
403 	/* We accept everything at this stage. */
404 	return 0;
405 }
406 
parse_early_options(char * cmdline)407 void __init parse_early_options(char *cmdline)
408 {
409 	parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
410 }
411 
412 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)413 void __init parse_early_param(void)
414 {
415 	static __initdata int done = 0;
416 	static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
417 
418 	if (done)
419 		return;
420 
421 	/* All fall through to do_early_param. */
422 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
423 	parse_early_options(tmp_cmdline);
424 	done = 1;
425 }
426 
427 /*
428  *	Activate the first processor.
429  */
430 
boot_cpu_init(void)431 static void __init boot_cpu_init(void)
432 {
433 	int cpu = smp_processor_id();
434 	/* Mark the boot cpu "present", "online" etc for SMP and UP case */
435 	set_cpu_online(cpu, true);
436 	set_cpu_active(cpu, true);
437 	set_cpu_present(cpu, true);
438 	set_cpu_possible(cpu, true);
439 }
440 
smp_setup_processor_id(void)441 void __init __weak smp_setup_processor_id(void)
442 {
443 }
444 
thread_info_cache_init(void)445 void __init __weak thread_info_cache_init(void)
446 {
447 }
448 
449 /*
450  * Set up kernel memory allocators
451  */
mm_init(void)452 static void __init mm_init(void)
453 {
454 	/*
455 	 * page_cgroup requires contiguous pages,
456 	 * bigger than MAX_ORDER unless SPARSEMEM.
457 	 */
458 	page_cgroup_init_flatmem();
459 	mem_init();
460 	kmem_cache_init();
461 	percpu_init_late();
462 	pgtable_cache_init();
463 	vmalloc_init();
464 }
465 
start_kernel(void)466 asmlinkage void __init start_kernel(void)
467 {
468 	char * command_line;
469 	extern const struct kernel_param __start___param[], __stop___param[];
470 
471 	/*
472 	 * Need to run as early as possible, to initialize the
473 	 * lockdep hash:
474 	 */
475 	lockdep_init();
476 	smp_setup_processor_id();
477 	debug_objects_early_init();
478 
479 	/*
480 	 * Set up the the initial canary ASAP:
481 	 */
482 	boot_init_stack_canary();
483 
484 	cgroup_init_early();
485 
486 	local_irq_disable();
487 	early_boot_irqs_disabled = true;
488 
489 /*
490  * Interrupts are still disabled. Do necessary setups, then
491  * enable them
492  */
493 	tick_init();
494 	boot_cpu_init();
495 	page_address_init();
496 	printk(KERN_NOTICE "%s", linux_banner);
497 	setup_arch(&command_line);
498 	mm_init_owner(&init_mm, &init_task);
499 	mm_init_cpumask(&init_mm);
500 	setup_command_line(command_line);
501 	setup_nr_cpu_ids();
502 	setup_per_cpu_areas();
503 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
504 
505 	build_all_zonelists(NULL);
506 	page_alloc_init();
507 
508 	printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
509 	parse_early_param();
510 	parse_args("Booting kernel", static_command_line, __start___param,
511 		   __stop___param - __start___param,
512 		   -1, -1, &unknown_bootoption);
513 
514 	jump_label_init();
515 
516 	/*
517 	 * These use large bootmem allocations and must precede
518 	 * kmem_cache_init()
519 	 */
520 	setup_log_buf(0);
521 	pidhash_init();
522 	vfs_caches_init_early();
523 	sort_main_extable();
524 	trap_init();
525 	mm_init();
526 
527 	/*
528 	 * Set up the scheduler prior starting any interrupts (such as the
529 	 * timer interrupt). Full topology setup happens at smp_init()
530 	 * time - but meanwhile we still have a functioning scheduler.
531 	 */
532 	sched_init();
533 	/*
534 	 * Disable preemption - early bootup scheduling is extremely
535 	 * fragile until we cpu_idle() for the first time.
536 	 */
537 	preempt_disable();
538 	if (!irqs_disabled()) {
539 		printk(KERN_WARNING "start_kernel(): bug: interrupts were "
540 				"enabled *very* early, fixing it\n");
541 		local_irq_disable();
542 	}
543 	idr_init_cache();
544 	perf_event_init();
545 	rcu_init();
546 	radix_tree_init();
547 	/* init some links before init_ISA_irqs() */
548 	early_irq_init();
549 	init_IRQ();
550 	prio_tree_init();
551 	init_timers();
552 	hrtimers_init();
553 	softirq_init();
554 	timekeeping_init();
555 	time_init();
556 	profile_init();
557 	call_function_init();
558 	if (!irqs_disabled())
559 		printk(KERN_CRIT "start_kernel(): bug: interrupts were "
560 				 "enabled early\n");
561 	early_boot_irqs_disabled = false;
562 	local_irq_enable();
563 
564 	kmem_cache_init_late();
565 
566 	/*
567 	 * HACK ALERT! This is early. We're enabling the console before
568 	 * we've done PCI setups etc, and console_init() must be aware of
569 	 * this. But we do want output early, in case something goes wrong.
570 	 */
571 	console_init();
572 	if (panic_later)
573 		panic(panic_later, panic_param);
574 
575 	lockdep_info();
576 
577 	/*
578 	 * Need to run this when irqs are enabled, because it wants
579 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
580 	 * too:
581 	 */
582 	locking_selftest();
583 
584 #ifdef CONFIG_BLK_DEV_INITRD
585 	if (initrd_start && !initrd_below_start_ok &&
586 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
587 		printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
588 		    "disabling it.\n",
589 		    page_to_pfn(virt_to_page((void *)initrd_start)),
590 		    min_low_pfn);
591 		initrd_start = 0;
592 	}
593 #endif
594 	page_cgroup_init();
595 	debug_objects_mem_init();
596 	kmemleak_init();
597 	setup_per_cpu_pageset();
598 	numa_policy_init();
599 	if (late_time_init)
600 		late_time_init();
601 	sched_clock_init();
602 	calibrate_delay();
603 	pidmap_init();
604 	anon_vma_init();
605 #ifdef CONFIG_X86
606 	if (efi_enabled(EFI_RUNTIME_SERVICES))
607 		efi_enter_virtual_mode();
608 #endif
609 	thread_info_cache_init();
610 	cred_init();
611 	fork_init(totalram_pages);
612 	proc_caches_init();
613 	buffer_init();
614 	key_init();
615 	security_init();
616 	dbg_late_init();
617 	vfs_caches_init(totalram_pages);
618 	signals_init();
619 	/* rootfs populating might need page-writeback */
620 	page_writeback_init();
621 #ifdef CONFIG_PROC_FS
622 	proc_root_init();
623 #endif
624 	cgroup_init();
625 	cpuset_init();
626 	taskstats_init_early();
627 	delayacct_init();
628 
629 	check_bugs();
630 
631 	acpi_early_init(); /* before LAPIC and SMP init */
632 	sfi_init_late();
633 
634 	if (efi_enabled(EFI_RUNTIME_SERVICES))
635 		efi_free_boot_services();
636 
637 	ftrace_init();
638 
639 	/* Do the rest non-__init'ed, we're now alive */
640 	rest_init();
641 }
642 
643 /* Call all constructor functions linked into the kernel. */
do_ctors(void)644 static void __init do_ctors(void)
645 {
646 #ifdef CONFIG_CONSTRUCTORS
647 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
648 
649 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
650 		(*fn)();
651 #endif
652 }
653 
654 bool initcall_debug;
655 core_param(initcall_debug, initcall_debug, bool, 0644);
656 
657 static char msgbuf[64];
658 
do_one_initcall_debug(initcall_t fn)659 static int __init_or_module do_one_initcall_debug(initcall_t fn)
660 {
661 	ktime_t calltime, delta, rettime;
662 	unsigned long long duration;
663 	int ret;
664 
665 	printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
666 	calltime = ktime_get();
667 	ret = fn();
668 	rettime = ktime_get();
669 	delta = ktime_sub(rettime, calltime);
670 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
671 	printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
672 		ret, duration);
673 
674 	return ret;
675 }
676 
do_one_initcall(initcall_t fn)677 int __init_or_module do_one_initcall(initcall_t fn)
678 {
679 	int count = preempt_count();
680 	int ret;
681 
682 	if (initcall_debug)
683 		ret = do_one_initcall_debug(fn);
684 	else
685 		ret = fn();
686 
687 	msgbuf[0] = 0;
688 
689 	if (ret && ret != -ENODEV && initcall_debug)
690 		sprintf(msgbuf, "error code %d ", ret);
691 
692 	if (preempt_count() != count) {
693 		strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
694 		preempt_count() = count;
695 	}
696 	if (irqs_disabled()) {
697 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
698 		local_irq_enable();
699 	}
700 	if (msgbuf[0]) {
701 		printk("initcall %pF returned with %s\n", fn, msgbuf);
702 	}
703 
704 	return ret;
705 }
706 
707 
708 extern initcall_t __initcall_start[];
709 extern initcall_t __initcall0_start[];
710 extern initcall_t __initcall1_start[];
711 extern initcall_t __initcall2_start[];
712 extern initcall_t __initcall3_start[];
713 extern initcall_t __initcall4_start[];
714 extern initcall_t __initcall5_start[];
715 extern initcall_t __initcall6_start[];
716 extern initcall_t __initcall7_start[];
717 extern initcall_t __initcall_end[];
718 
719 static initcall_t *initcall_levels[] __initdata = {
720 	__initcall0_start,
721 	__initcall1_start,
722 	__initcall2_start,
723 	__initcall3_start,
724 	__initcall4_start,
725 	__initcall5_start,
726 	__initcall6_start,
727 	__initcall7_start,
728 	__initcall_end,
729 };
730 
731 static char *initcall_level_names[] __initdata = {
732 	"early parameters",
733 	"core parameters",
734 	"postcore parameters",
735 	"arch parameters",
736 	"subsys parameters",
737 	"fs parameters",
738 	"device parameters",
739 	"late parameters",
740 };
741 
do_initcall_level(int level)742 static void __init do_initcall_level(int level)
743 {
744 	extern const struct kernel_param __start___param[], __stop___param[];
745 	initcall_t *fn;
746 
747 	strcpy(static_command_line, saved_command_line);
748 	parse_args(initcall_level_names[level],
749 		   static_command_line, __start___param,
750 		   __stop___param - __start___param,
751 		   level, level,
752 		   repair_env_string);
753 
754 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
755 		do_one_initcall(*fn);
756 }
757 
do_initcalls(void)758 static void __init do_initcalls(void)
759 {
760 	int level;
761 
762 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
763 		do_initcall_level(level);
764 }
765 
766 /*
767  * Ok, the machine is now initialized. None of the devices
768  * have been touched yet, but the CPU subsystem is up and
769  * running, and memory and process management works.
770  *
771  * Now we can finally start doing some real work..
772  */
do_basic_setup(void)773 static void __init do_basic_setup(void)
774 {
775 	cpuset_init_smp();
776 	usermodehelper_init();
777 	shmem_init();
778 	driver_init();
779 	init_irq_proc();
780 	do_ctors();
781 	usermodehelper_enable();
782 	do_initcalls();
783 	random_int_secret_init();
784 }
785 
do_pre_smp_initcalls(void)786 static void __init do_pre_smp_initcalls(void)
787 {
788 	initcall_t *fn;
789 
790 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
791 		do_one_initcall(*fn);
792 }
793 
run_init_process(const char * init_filename)794 static void run_init_process(const char *init_filename)
795 {
796 	argv_init[0] = init_filename;
797 	kernel_execve(init_filename, argv_init, envp_init);
798 }
799 
800 /* This is a non __init function. Force it to be noinline otherwise gcc
801  * makes it inline to init() and it becomes part of init.text section
802  */
init_post(void)803 static noinline int init_post(void)
804 {
805 	/* need to finish all async __init code before freeing the memory */
806 	async_synchronize_full();
807 	free_initmem();
808 	mark_rodata_ro();
809 	system_state = SYSTEM_RUNNING;
810 	numa_default_policy();
811 
812 
813 	current->signal->flags |= SIGNAL_UNKILLABLE;
814 
815 	if (ramdisk_execute_command) {
816 		run_init_process(ramdisk_execute_command);
817 		printk(KERN_WARNING "Failed to execute %s\n",
818 				ramdisk_execute_command);
819 	}
820 
821 	/*
822 	 * We try each of these until one succeeds.
823 	 *
824 	 * The Bourne shell can be used instead of init if we are
825 	 * trying to recover a really broken machine.
826 	 */
827 	if (execute_command) {
828 		run_init_process(execute_command);
829 		printk(KERN_WARNING "Failed to execute %s.  Attempting "
830 					"defaults...\n", execute_command);
831 	}
832 	run_init_process("/sbin/init");
833 	run_init_process("/etc/init");
834 	run_init_process("/bin/init");
835 	run_init_process("/bin/sh");
836 
837 	panic("No init found.  Try passing init= option to kernel. "
838 	      "See Linux Documentation/init.txt for guidance.");
839 }
840 
kernel_init(void * unused)841 static int __init kernel_init(void * unused)
842 {
843 	/*
844 	 * Wait until kthreadd is all set-up.
845 	 */
846 	wait_for_completion(&kthreadd_done);
847 
848 	/* Now the scheduler is fully set up and can do blocking allocations */
849 	gfp_allowed_mask = __GFP_BITS_MASK;
850 
851 	/*
852 	 * init can allocate pages on any node
853 	 */
854 	set_mems_allowed(node_states[N_HIGH_MEMORY]);
855 	/*
856 	 * init can run on any cpu.
857 	 */
858 	set_cpus_allowed_ptr(current, cpu_all_mask);
859 
860 	cad_pid = task_pid(current);
861 
862 	smp_prepare_cpus(setup_max_cpus);
863 
864 	do_pre_smp_initcalls();
865 	lockup_detector_init();
866 
867 	smp_init();
868 	sched_init_smp();
869 
870 	do_basic_setup();
871 
872 	/* Open the /dev/console on the rootfs, this should never fail */
873 	if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
874 		printk(KERN_WARNING "Warning: unable to open an initial console.\n");
875 
876 	(void) sys_dup(0);
877 	(void) sys_dup(0);
878 	/*
879 	 * check if there is an early userspace init.  If yes, let it do all
880 	 * the work
881 	 */
882 
883 	if (!ramdisk_execute_command)
884 		ramdisk_execute_command = "/init";
885 
886 	if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
887 		ramdisk_execute_command = NULL;
888 		prepare_namespace();
889 	}
890 
891 	/*
892 	 * Ok, we have completed the initial bootup, and
893 	 * we're essentially up and running. Get rid of the
894 	 * initmem segments and start the user-mode stuff..
895 	 */
896 
897 	init_post();
898 	return 0;
899 }
900