main.c 22.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11
/*
 *  linux/init/main.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 *
 *  GK 2/5/95  -  Changed to support mounting root fs via NFS
 *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
 *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
 *  Simplified starting of init:  Michael A. Griffith <grif@acm.org> 
 */

A
Andrew Morton 已提交
12 13
#define DEBUG		/* Enable initcall_debug */

L
Linus Torvalds 已提交
14 15 16 17 18
#include <linux/types.h>
#include <linux/module.h>
#include <linux/proc_fs.h>
#include <linux/kernel.h>
#include <linux/syscalls.h>
I
Ingo Molnar 已提交
19
#include <linux/stackprotector.h>
L
Linus Torvalds 已提交
20 21 22 23 24 25 26
#include <linux/string.h>
#include <linux/ctype.h>
#include <linux/delay.h>
#include <linux/ioport.h>
#include <linux/init.h>
#include <linux/initrd.h>
#include <linux/bootmem.h>
27
#include <linux/acpi.h>
L
Linus Torvalds 已提交
28 29 30
#include <linux/tty.h>
#include <linux/percpu.h>
#include <linux/kmod.h>
N
Nick Piggin 已提交
31
#include <linux/vmalloc.h>
L
Linus Torvalds 已提交
32
#include <linux/kernel_stat.h>
33
#include <linux/start_kernel.h>
L
Linus Torvalds 已提交
34
#include <linux/security.h>
35
#include <linux/smp.h>
L
Linus Torvalds 已提交
36 37 38 39 40 41 42
#include <linux/profile.h>
#include <linux/rcupdate.h>
#include <linux/moduleparam.h>
#include <linux/kallsyms.h>
#include <linux/writeback.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
43
#include <linux/cgroup.h>
L
Linus Torvalds 已提交
44
#include <linux/efi.h>
45
#include <linux/tick.h>
46
#include <linux/interrupt.h>
47
#include <linux/taskstats_kern.h>
48
#include <linux/delayacct.h>
L
Linus Torvalds 已提交
49 50 51 52
#include <linux/unistd.h>
#include <linux/rmap.h>
#include <linux/mempolicy.h>
#include <linux/key.h>
A
Adrian Bunk 已提交
53
#include <linux/buffer_head.h>
54
#include <linux/page_cgroup.h>
55
#include <linux/debug_locks.h>
56
#include <linux/debugobjects.h>
I
Ingo Molnar 已提交
57
#include <linux/lockdep.h>
C
Catalin Marinas 已提交
58
#include <linux/kmemleak.h>
59
#include <linux/pid_namespace.h>
60
#include <linux/device.h>
61
#include <linux/kthread.h>
62
#include <linux/sched.h>
63
#include <linux/signal.h>
64
#include <linux/idr.h>
65
#include <linux/kgdb.h>
66
#include <linux/ftrace.h>
67
#include <linux/async.h>
68
#include <linux/kmemcheck.h>
69
#include <linux/sfi.h>
70
#include <linux/shmem_fs.h>
71
#include <linux/slab.h>
72
#include <linux/perf_event.h>
A
Al Viro 已提交
73
#include <linux/file.h>
74
#include <linux/ptrace.h>
75 76
#include <linux/blkdev.h>
#include <linux/elevator.h>
77
#include <linux/sched_clock.h>
78
#include <linux/context_tracking.h>
79
#include <linux/random.h>
L
Linus Torvalds 已提交
80 81 82 83

#include <asm/io.h>
#include <asm/bugs.h>
#include <asm/setup.h>
84
#include <asm/sections.h>
85
#include <asm/cacheflush.h>
L
Linus Torvalds 已提交
86 87 88 89 90

#ifdef CONFIG_X86_LOCAL_APIC
#include <asm/smp.h>
#endif

91
static int kernel_init(void *);
L
Linus Torvalds 已提交
92 93 94 95 96 97

extern void init_IRQ(void);
extern void fork_init(unsigned long);
extern void mca_init(void);
extern void sbus_init(void);
extern void radix_tree_init(void);
98 99 100
#ifndef CONFIG_DEBUG_RODATA
static inline void mark_rodata_ro(void) { }
#endif
L
Linus Torvalds 已提交
101

102 103 104 105 106 107 108 109 110
/*
 * Debug helper: via this flag we know that we are in 'early bootup code'
 * where only the boot processor is running with IRQ disabled.  This means
 * two things - IRQ must not be enabled before the flag is cleared and some
 * operations which are not allowed with IRQ disabled are allowed while the
 * flag is set.
 */
bool early_boot_irqs_disabled __read_mostly;

111
enum system_states system_state __read_mostly;
L
Linus Torvalds 已提交
112 113 114 115 116 117 118 119 120 121
EXPORT_SYMBOL(system_state);

/*
 * Boot command-line arguments
 */
#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT

extern void time_init(void);
/* Default late time init is NULL. archs can override this later. */
122
void (*__initdata late_time_init)(void);
L
Linus Torvalds 已提交
123

124 125 126 127 128 129
/* Untouched command line saved by arch-specific code. */
char __initdata boot_command_line[COMMAND_LINE_SIZE];
/* Untouched saved command line (eg. for /proc) */
char *saved_command_line;
/* Command line for parameter parsing */
static char *static_command_line;
130 131
/* Command line for per-initcall parameter parsing */
static char *initcall_command_line;
L
Linus Torvalds 已提交
132 133

static char *execute_command;
134
static char *ramdisk_execute_command;
L
Linus Torvalds 已提交
135

136 137 138 139 140 141 142
/*
 * Used to generate warnings if static_key manipulation functions are used
 * before jump_label_init is called.
 */
bool static_key_initialized __read_mostly = false;
EXPORT_SYMBOL_GPL(static_key_initialized);

J
Jan Beulich 已提交
143 144 145 146 147 148 149 150 151 152 153
/*
 * If set, this is an indication to the drivers that reset the underlying
 * device before going ahead with the initialization otherwise driver might
 * rely on the BIOS and skip the reset operation.
 *
 * This is useful if kernel is booting in an unreliable environment.
 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
 * skipped and devices will be in unknown state.
 */
unsigned int reset_devices;
EXPORT_SYMBOL(reset_devices);
L
Linus Torvalds 已提交
154

155 156 157 158 159 160 161 162
static int __init set_reset_devices(char *str)
{
	reset_devices = 1;
	return 1;
}

__setup("reset_devices", set_reset_devices);

163 164
static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
L
Linus Torvalds 已提交
165 166
static const char *panic_later, *panic_param;

167
extern const struct obs_kernel_param __setup_start[], __setup_end[];
L
Linus Torvalds 已提交
168 169 170

static int __init obsolete_checksetup(char *line)
{
171
	const struct obs_kernel_param *p;
172
	int had_early_param = 0;
L
Linus Torvalds 已提交
173 174 175 176

	p = __setup_start;
	do {
		int n = strlen(p->str);
177
		if (parameqn(line, p->str, n)) {
L
Linus Torvalds 已提交
178
			if (p->early) {
179 180 181 182
				/* Already done in parse_early_param?
				 * (Needs exact match on param part).
				 * Keep iterating, as we can have early
				 * params and __setups of same names 8( */
L
Linus Torvalds 已提交
183
				if (line[n] == '\0' || line[n] == '=')
184
					had_early_param = 1;
L
Linus Torvalds 已提交
185
			} else if (!p->setup_func) {
A
Andrew Morton 已提交
186 187
				pr_warn("Parameter %s is obsolete, ignored\n",
					p->str);
L
Linus Torvalds 已提交
188 189 190 191 192 193
				return 1;
			} else if (p->setup_func(line + n))
				return 1;
		}
		p++;
	} while (p < __setup_end);
194 195

	return had_early_param;
L
Linus Torvalds 已提交
196 197 198 199 200 201 202 203 204 205 206 207 208
}

/*
 * This should be approx 2 Bo*oMips to start (note initial shift), and will
 * still work even if initially too large, it will just take slightly longer
 */
unsigned long loops_per_jiffy = (1<<12);

EXPORT_SYMBOL(loops_per_jiffy);

static int __init debug_kernel(char *str)
{
	console_loglevel = 10;
209
	return 0;
L
Linus Torvalds 已提交
210 211 212 213 214
}

static int __init quiet_kernel(char *str)
{
	console_loglevel = 4;
215
	return 0;
L
Linus Torvalds 已提交
216 217
}

218 219
early_param("debug", debug_kernel);
early_param("quiet", quiet_kernel);
L
Linus Torvalds 已提交
220 221 222

static int __init loglevel(char *str)
{
223 224 225 226 227 228 229 230 231 232 233 234 235
	int newlevel;

	/*
	 * Only update loglevel value when a correct setting was passed,
	 * to prevent blind crashes (when loglevel being set to 0) that
	 * are quite hard to debug
	 */
	if (get_option(&str, &newlevel)) {
		console_loglevel = newlevel;
		return 0;
	}

	return -EINVAL;
L
Linus Torvalds 已提交
236 237
}

238
early_param("loglevel", loglevel);
L
Linus Torvalds 已提交
239

240
/* Change NUL term back to "=", to make "param" the whole string. */
241
static int __init repair_env_string(char *param, char *val, const char *unused)
L
Linus Torvalds 已提交
242 243 244 245 246 247 248 249 250 251 252 253
{
	if (val) {
		/* param=val or param="val"? */
		if (val == param+strlen(param)+1)
			val[-1] = '=';
		else if (val == param+strlen(param)+2) {
			val[-2] = '=';
			memmove(val-1, val, strlen(val)+1);
			val--;
		} else
			BUG();
	}
254 255 256 257 258 259 260
	return 0;
}

/*
 * Unknown boot options get handed to init, unless they look like
 * unused parameters (modprobe will find them in /proc/cmdline).
 */
261
static int __init unknown_bootoption(char *param, char *val, const char *unused)
262
{
263
	repair_env_string(param, val, unused);
L
Linus Torvalds 已提交
264 265 266 267 268

	/* Handle obsolete-style parameters */
	if (obsolete_checksetup(param))
		return 0;

269 270
	/* Unused module parameter. */
	if (strchr(param, '.') && (!val || strchr(param, '.') < val))
L
Linus Torvalds 已提交
271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318
		return 0;

	if (panic_later)
		return 0;

	if (val) {
		/* Environment option */
		unsigned int i;
		for (i = 0; envp_init[i]; i++) {
			if (i == MAX_INIT_ENVS) {
				panic_later = "Too many boot env vars at `%s'";
				panic_param = param;
			}
			if (!strncmp(param, envp_init[i], val - param))
				break;
		}
		envp_init[i] = param;
	} else {
		/* Command line option */
		unsigned int i;
		for (i = 0; argv_init[i]; i++) {
			if (i == MAX_INIT_ARGS) {
				panic_later = "Too many boot init vars at `%s'";
				panic_param = param;
			}
		}
		argv_init[i] = param;
	}
	return 0;
}

static int __init init_setup(char *str)
{
	unsigned int i;

	execute_command = str;
	/*
	 * In case LILO is going to boot us with default command line,
	 * it prepends "auto" before the whole cmdline which makes
	 * the shell think it should execute a script with such name.
	 * So we ignore all arguments entered _before_ init=... [MJ]
	 */
	for (i = 1; i < MAX_INIT_ARGS; i++)
		argv_init[i] = NULL;
	return 1;
}
__setup("init=", init_setup);

319 320 321 322 323 324 325 326 327 328 329 330
static int __init rdinit_setup(char *str)
{
	unsigned int i;

	ramdisk_execute_command = str;
	/* See "auto" comment in init_setup */
	for (i = 1; i < MAX_INIT_ARGS; i++)
		argv_init[i] = NULL;
	return 1;
}
__setup("rdinit=", rdinit_setup);

L
Linus Torvalds 已提交
331
#ifndef CONFIG_SMP
332
static const unsigned int setup_max_cpus = NR_CPUS;
L
Linus Torvalds 已提交
333 334 335 336 337 338 339 340 341
#ifdef CONFIG_X86_LOCAL_APIC
static void __init smp_init(void)
{
	APIC_init_uniprocessor();
}
#else
#define smp_init()	do { } while (0)
#endif

342
static inline void setup_nr_cpu_ids(void) { }
L
Linus Torvalds 已提交
343 344 345
static inline void smp_prepare_cpus(unsigned int maxcpus) { }
#endif

346 347 348 349 350 351 352 353
/*
 * We need to store the untouched command line for future reference.
 * We also need to store the touched command line since the parameter
 * parsing is performed in place, and we should allow a component to
 * store reference of name/value for future reference.
 */
static void __init setup_command_line(char *command_line)
{
354 355 356 357 358
	saved_command_line =
		memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
	initcall_command_line =
		memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
	static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
359 360 361 362
	strcpy (saved_command_line, boot_command_line);
	strcpy (static_command_line, command_line);
}

L
Linus Torvalds 已提交
363 364 365 366 367 368 369 370 371
/*
 * We need to finalize in a non-__init function or else race conditions
 * between the root thread and the init thread may cause start_kernel to
 * be reaped by free_initmem before the root thread has proceeded to
 * cpu_idle.
 *
 * gcc-3.4 accidentally inlines this function, so use noinline.
 */

372 373
static __initdata DECLARE_COMPLETION(kthreadd_done);

374
static noinline void __init_refok rest_init(void)
L
Linus Torvalds 已提交
375
{
376 377
	int pid;

378
	rcu_scheduler_starting();
379
	/*
P
Peter Zijlstra 已提交
380
	 * We need to spawn init first so that it obtains pid 1, however
381 382 383
	 * the init task will end up wanting to create kthreads, which, if
	 * we schedule it before we create kthreadd, will OOPS.
	 */
384
	kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
L
Linus Torvalds 已提交
385
	numa_default_policy();
386
	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
387
	rcu_read_lock();
P
Pavel Emelyanov 已提交
388
	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
389
	rcu_read_unlock();
390
	complete(&kthreadd_done);
391 392 393

	/*
	 * The boot idle thread must execute schedule()
I
Ingo Molnar 已提交
394
	 * at least once to get things moving:
395
	 */
I
Ingo Molnar 已提交
396
	init_idle_bootup_task(current);
397
	schedule_preempt_disabled();
398
	/* Call into cpu_idle with preempt disabled */
399
	cpu_startup_entry(CPUHP_ONLINE);
I
Ingo Molnar 已提交
400
}
L
Linus Torvalds 已提交
401 402

/* Check for early params. */
403
static int __init do_early_param(char *param, char *val, const char *unused)
L
Linus Torvalds 已提交
404
{
405
	const struct obs_kernel_param *p;
L
Linus Torvalds 已提交
406 407

	for (p = __setup_start; p < __setup_end; p++) {
408
		if ((p->early && parameq(param, p->str)) ||
409 410 411
		    (strcmp(param, "console") == 0 &&
		     strcmp(p->str, "earlycon") == 0)
		) {
L
Linus Torvalds 已提交
412
			if (p->setup_func(val) != 0)
A
Andrew Morton 已提交
413
				pr_warn("Malformed early option '%s'\n", param);
L
Linus Torvalds 已提交
414 415 416 417 418 419
		}
	}
	/* We accept everything at this stage. */
	return 0;
}

M
Magnus Damm 已提交
420 421
void __init parse_early_options(char *cmdline)
{
422
	parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
M
Magnus Damm 已提交
423 424
}

L
Linus Torvalds 已提交
425 426 427 428 429 430 431 432 433 434
/* Arch code calls this early on, or if not, just before other parsing. */
void __init parse_early_param(void)
{
	static __initdata int done = 0;
	static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];

	if (done)
		return;

	/* All fall through to do_early_param. */
435
	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
M
Magnus Damm 已提交
436
	parse_early_options(tmp_cmdline);
L
Linus Torvalds 已提交
437 438 439 440 441 442 443
	done = 1;
}

/*
 *	Activate the first processor.
 */

444 445 446 447
static void __init boot_cpu_init(void)
{
	int cpu = smp_processor_id();
	/* Mark the boot cpu "present", "online" etc for SMP and UP case */
448
	set_cpu_online(cpu, true);
449
	set_cpu_active(cpu, true);
450 451
	set_cpu_present(cpu, true);
	set_cpu_possible(cpu, true);
452 453
}

454
void __init __weak smp_setup_processor_id(void)
455 456 457
{
}

458
# if THREAD_SIZE >= PAGE_SIZE
459 460 461
void __init __weak thread_info_cache_init(void)
{
}
462
#endif
463

P
Pekka Enberg 已提交
464 465 466 467 468
/*
 * Set up kernel memory allocators
 */
static void __init mm_init(void)
{
469
	/*
470 471
	 * page_cgroup requires contiguous pages,
	 * bigger than MAX_ORDER unless SPARSEMEM.
472 473
	 */
	page_cgroup_init_flatmem();
P
Pekka Enberg 已提交
474 475
	mem_init();
	kmem_cache_init();
476
	percpu_init_late();
477
	pgtable_init();
P
Pekka Enberg 已提交
478 479 480
	vmalloc_init();
}

L
Linus Torvalds 已提交
481 482 483
asmlinkage void __init start_kernel(void)
{
	char * command_line;
484
	extern const struct kernel_param __start___param[], __stop___param[];
485

I
Ingo Molnar 已提交
486 487 488 489 490
	/*
	 * Need to run as early as possible, to initialize the
	 * lockdep hash:
	 */
	lockdep_init();
491
	smp_setup_processor_id();
492
	debug_objects_early_init();
493 494 495 496 497 498

	/*
	 * Set up the the initial canary ASAP:
	 */
	boot_init_stack_canary();

499
	cgroup_init_early();
I
Ingo Molnar 已提交
500 501

	local_irq_disable();
502
	early_boot_irqs_disabled = true;
I
Ingo Molnar 已提交
503

L
Linus Torvalds 已提交
504 505 506 507
/*
 * Interrupts are still disabled. Do necessary setups, then
 * enable them
 */
508
	boot_cpu_init();
L
Linus Torvalds 已提交
509
	page_address_init();
A
Andrew Morton 已提交
510
	pr_notice("%s", linux_banner);
L
Linus Torvalds 已提交
511
	setup_arch(&command_line);
512
	mm_init_owner(&init_mm, &init_task);
513
	mm_init_cpumask(&init_mm);
514
	setup_command_line(command_line);
515
	setup_nr_cpu_ids();
516
	setup_per_cpu_areas();
517
	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
L
Linus Torvalds 已提交
518

519
	build_all_zonelists(NULL, NULL);
520 521
	page_alloc_init();

A
Andrew Morton 已提交
522
	pr_notice("Kernel command line: %s\n", boot_command_line);
523 524 525
	parse_early_param();
	parse_args("Booting kernel", static_command_line, __start___param,
		   __stop___param - __start___param,
526
		   -1, -1, &unknown_bootoption);
527 528 529

	jump_label_init();

530 531 532 533
	/*
	 * These use large bootmem allocations and must precede
	 * kmem_cache_init()
	 */
534
	setup_log_buf(0);
535 536 537 538
	pidhash_init();
	vfs_caches_init_early();
	sort_main_extable();
	trap_init();
P
Pekka Enberg 已提交
539
	mm_init();
540

L
Linus Torvalds 已提交
541 542 543 544 545 546 547 548 549 550 551
	/*
	 * Set up the scheduler prior starting any interrupts (such as the
	 * timer interrupt). Full topology setup happens at smp_init()
	 * time - but meanwhile we still have a functioning scheduler.
	 */
	sched_init();
	/*
	 * Disable preemption - early bootup scheduling is extremely
	 * fragile until we cpu_idle() for the first time.
	 */
	preempt_disable();
552
	if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
553
		local_irq_disable();
P
Peter Zijlstra 已提交
554
	idr_init_cache();
L
Linus Torvalds 已提交
555
	rcu_init();
556
	tick_nohz_init();
557
	context_tracking_init();
Y
Yinghai Lu 已提交
558
	radix_tree_init();
559 560
	/* init some links before init_ISA_irqs() */
	early_irq_init();
L
Linus Torvalds 已提交
561
	init_IRQ();
T
Thomas Gleixner 已提交
562
	tick_init();
L
Linus Torvalds 已提交
563
	init_timers();
564
	hrtimers_init();
L
Linus Torvalds 已提交
565
	softirq_init();
566
	timekeeping_init();
J
john stultz 已提交
567
	time_init();
568
	sched_clock_postinit();
569
	perf_event_init();
570
	profile_init();
571
	call_function_init();
572
	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
573
	early_boot_irqs_disabled = false;
574
	local_irq_enable();
575

576
	kmem_cache_init_late();
L
Linus Torvalds 已提交
577 578 579 580 581 582 583 584 585

	/*
	 * HACK ALERT! This is early. We're enabling the console before
	 * we've done PCI setups etc, and console_init() must be aware of
	 * this. But we do want output early, in case something goes wrong.
	 */
	console_init();
	if (panic_later)
		panic(panic_later, panic_param);
I
Ingo Molnar 已提交
586 587 588

	lockdep_info();

589 590 591 592 593 594 595
	/*
	 * Need to run this when irqs are enabled, because it wants
	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
	 * too:
	 */
	locking_selftest();

L
Linus Torvalds 已提交
596 597
#ifdef CONFIG_BLK_DEV_INITRD
	if (initrd_start && !initrd_below_start_ok &&
598
	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
A
Andrew Morton 已提交
599
		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
600 601
		    page_to_pfn(virt_to_page((void *)initrd_start)),
		    min_low_pfn);
L
Linus Torvalds 已提交
602 603 604
		initrd_start = 0;
	}
#endif
605
	page_cgroup_init();
606
	debug_objects_mem_init();
607
	kmemleak_init();
608
	setup_per_cpu_pageset();
L
Linus Torvalds 已提交
609 610 611
	numa_policy_init();
	if (late_time_init)
		late_time_init();
612
	sched_clock_init();
L
Linus Torvalds 已提交
613 614 615
	calibrate_delay();
	pidmap_init();
	anon_vma_init();
616
#ifdef CONFIG_X86
617
	if (efi_enabled(EFI_RUNTIME_SERVICES))
618 619
		efi_enter_virtual_mode();
#endif
620
	thread_info_cache_init();
D
David Howells 已提交
621
	cred_init();
622
	fork_init(totalram_pages);
L
Linus Torvalds 已提交
623 624 625 626
	proc_caches_init();
	buffer_init();
	key_init();
	security_init();
627
	dbg_late_init();
628
	vfs_caches_init(totalram_pages);
L
Linus Torvalds 已提交
629 630 631 632 633 634
	signals_init();
	/* rootfs populating might need page-writeback */
	page_writeback_init();
#ifdef CONFIG_PROC_FS
	proc_root_init();
#endif
635
	cgroup_init();
L
Linus Torvalds 已提交
636
	cpuset_init();
637
	taskstats_init_early();
638
	delayacct_init();
L
Linus Torvalds 已提交
639 640 641 642

	check_bugs();

	acpi_early_init(); /* before LAPIC and SMP init */
643
	sfi_init_late();
L
Linus Torvalds 已提交
644

645
	if (efi_enabled(EFI_RUNTIME_SERVICES)) {
646
		efi_late_init();
647
		efi_free_boot_services();
648
	}
649

650 651
	ftrace_init();

L
Linus Torvalds 已提交
652 653 654 655
	/* Do the rest non-__init'ed, we're now alive */
	rest_init();
}

656 657 658 659
/* Call all constructor functions linked into the kernel. */
static void __init do_ctors(void)
{
#ifdef CONFIG_CONSTRUCTORS
660
	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
661

662 663
	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
		(*fn)();
664 665 666
#endif
}

667
bool initcall_debug;
R
Rusty Russell 已提交
668
core_param(initcall_debug, initcall_debug, bool, 0644);
L
Linus Torvalds 已提交
669

670
static int __init_or_module do_one_initcall_debug(initcall_t fn)
L
Linus Torvalds 已提交
671
{
672
	ktime_t calltime, delta, rettime;
A
Américo Wang 已提交
673 674
	unsigned long long duration;
	int ret;
L
Linus Torvalds 已提交
675

A
Andrew Morton 已提交
676
	pr_debug("calling  %pF @ %i\n", fn, task_pid_nr(current));
677
	calltime = ktime_get();
A
Américo Wang 已提交
678
	ret = fn();
679 680 681
	rettime = ktime_get();
	delta = ktime_sub(rettime, calltime);
	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
A
Andrew Morton 已提交
682 683
	pr_debug("initcall %pF returned %d after %lld usecs\n",
		 fn, ret, duration);
L
Linus Torvalds 已提交
684

685 686 687
	return ret;
}

688
int __init_or_module do_one_initcall(initcall_t fn)
689 690 691
{
	int count = preempt_count();
	int ret;
692
	char msgbuf[64];
693 694 695 696 697

	if (initcall_debug)
		ret = do_one_initcall_debug(fn);
	else
		ret = fn();
698

699
	msgbuf[0] = 0;
700

701
	if (preempt_count() != count) {
702
		sprintf(msgbuf, "preemption imbalance ");
703
		preempt_count_set(count);
L
Linus Torvalds 已提交
704
	}
705
	if (irqs_disabled()) {
706
		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
707 708
		local_irq_enable();
	}
709
	WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
710

A
Américo Wang 已提交
711
	return ret;
712 713 714
}


715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737
extern initcall_t __initcall_start[];
extern initcall_t __initcall0_start[];
extern initcall_t __initcall1_start[];
extern initcall_t __initcall2_start[];
extern initcall_t __initcall3_start[];
extern initcall_t __initcall4_start[];
extern initcall_t __initcall5_start[];
extern initcall_t __initcall6_start[];
extern initcall_t __initcall7_start[];
extern initcall_t __initcall_end[];

static initcall_t *initcall_levels[] __initdata = {
	__initcall0_start,
	__initcall1_start,
	__initcall2_start,
	__initcall3_start,
	__initcall4_start,
	__initcall5_start,
	__initcall6_start,
	__initcall7_start,
	__initcall_end,
};

738
/* Keep these in sync with initcalls in include/linux/init.h */
739
static char *initcall_level_names[] __initdata = {
740 741 742 743 744 745 746 747
	"early",
	"core",
	"postcore",
	"arch",
	"subsys",
	"fs",
	"device",
	"late",
748 749 750
};

static void __init do_initcall_level(int level)
751
{
752
	extern const struct kernel_param __start___param[], __stop___param[];
753
	initcall_t *fn;
754

755
	strcpy(initcall_command_line, saved_command_line);
756
	parse_args(initcall_level_names[level],
757
		   initcall_command_line, __start___param,
758 759
		   __stop___param - __start___param,
		   level, level,
760
		   &repair_env_string);
761 762

	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
763
		do_one_initcall(*fn);
L
Linus Torvalds 已提交
764 765
}

766 767 768 769
static void __init do_initcalls(void)
{
	int level;

770
	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
771 772 773
		do_initcall_level(level);
}

L
Linus Torvalds 已提交
774 775 776 777 778 779 780 781 782
/*
 * Ok, the machine is now initialized. None of the devices
 * have been touched yet, but the CPU subsystem is up and
 * running, and memory and process management works.
 *
 * Now we can finally start doing some real work..
 */
static void __init do_basic_setup(void)
{
783
	cpuset_init_smp();
L
Linus Torvalds 已提交
784
	usermodehelper_init();
785
	shmem_init();
L
Linus Torvalds 已提交
786
	driver_init();
787
	init_irq_proc();
788
	do_ctors();
789
	usermodehelper_enable();
790
	do_initcalls();
791
	random_int_secret_init();
L
Linus Torvalds 已提交
792 793
}

794
static void __init do_pre_smp_initcalls(void)
795
{
796
	initcall_t *fn;
797

798
	for (fn = __initcall_start; fn < __initcall0_start; fn++)
799
		do_one_initcall(*fn);
800 801
}

802 803 804 805 806 807 808 809 810 811 812
/*
 * This function requests modules which should be loaded by default and is
 * called twice right after initrd is mounted and right before init is
 * exec'd.  If such modules are on either initrd or rootfs, they will be
 * loaded before control is passed to userland.
 */
void __init load_default_modules(void)
{
	load_default_elevator_module();
}

813
static int run_init_process(const char *init_filename)
L
Linus Torvalds 已提交
814 815
{
	argv_init[0] = init_filename;
816 817 818
	return do_execve(init_filename,
		(const char __user *const __user *)argv_init,
		(const char __user *const __user *)envp_init);
L
Linus Torvalds 已提交
819 820
}

821 822 823 824 825 826 827 828 829 830 831 832 833 834
static int try_to_run_init_process(const char *init_filename)
{
	int ret;

	ret = run_init_process(init_filename);

	if (ret && ret != -ENOENT) {
		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
		       init_filename, ret);
	}

	return ret;
}

835
static noinline void __init kernel_init_freeable(void);
836 837

static int __ref kernel_init(void *unused)
838
{
839 840
	int ret;

841
	kernel_init_freeable();
842 843
	/* need to finish all async __init code before freeing the memory */
	async_synchronize_full();
844 845 846 847 848
	free_initmem();
	mark_rodata_ro();
	system_state = SYSTEM_RUNNING;
	numa_default_policy();

A
Al Viro 已提交
849
	flush_delayed_fput();
850

851
	if (ramdisk_execute_command) {
852 853
		ret = run_init_process(ramdisk_execute_command);
		if (!ret)
854
			return 0;
855 856
		pr_err("Failed to execute %s (error %d)\n",
		       ramdisk_execute_command, ret);
857 858 859 860 861 862 863 864 865
	}

	/*
	 * We try each of these until one succeeds.
	 *
	 * The Bourne shell can be used instead of init if we are
	 * trying to recover a really broken machine.
	 */
	if (execute_command) {
866 867
		ret = run_init_process(execute_command);
		if (!ret)
868
			return 0;
869 870
		pr_err("Failed to execute %s (error %d).  Attempting defaults...\n",
			execute_command, ret);
871
	}
872 873 874 875
	if (!try_to_run_init_process("/sbin/init") ||
	    !try_to_run_init_process("/etc/init") ||
	    !try_to_run_init_process("/bin/init") ||
	    !try_to_run_init_process("/bin/sh"))
876
		return 0;
877

878
	panic("No working init found.  Try passing init= option to kernel. "
879
	      "See Linux Documentation/init.txt for guidance.");
880 881
}

882
static noinline void __init kernel_init_freeable(void)
L
Linus Torvalds 已提交
883
{
884 885 886 887
	/*
	 * Wait until kthreadd is all set-up.
	 */
	wait_for_completion(&kthreadd_done);
888 889 890 891

	/* Now the scheduler is fully set up and can do blocking allocations */
	gfp_allowed_mask = __GFP_BITS_MASK;

892 893 894
	/*
	 * init can allocate pages on any node
	 */
895
	set_mems_allowed(node_states[N_MEMORY]);
L
Linus Torvalds 已提交
896 897 898
	/*
	 * init can run on any cpu.
	 */
899
	set_cpus_allowed_ptr(current, cpu_all_mask);
L
Linus Torvalds 已提交
900

901 902
	cad_pid = task_pid(current);

903
	smp_prepare_cpus(setup_max_cpus);
L
Linus Torvalds 已提交
904 905

	do_pre_smp_initcalls();
906
	lockup_detector_init();
L
Linus Torvalds 已提交
907 908 909 910 911 912

	smp_init();
	sched_init_smp();

	do_basic_setup();

913 914
	/* Open the /dev/console on the rootfs, this should never fail */
	if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
A
Andrew Morton 已提交
915
		pr_err("Warning: unable to open an initial console.\n");
916 917 918

	(void) sys_dup(0);
	(void) sys_dup(0);
L
Linus Torvalds 已提交
919 920 921 922
	/*
	 * check if there is an early userspace init.  If yes, let it do all
	 * the work
	 */
923 924 925 926 927 928

	if (!ramdisk_execute_command)
		ramdisk_execute_command = "/init";

	if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
		ramdisk_execute_command = NULL;
L
Linus Torvalds 已提交
929
		prepare_namespace();
930
	}
L
Linus Torvalds 已提交
931 932 933 934 935 936

	/*
	 * Ok, we have completed the initial bootup, and
	 * we're essentially up and running. Get rid of the
	 * initmem segments and start the user-mode stuff..
	 */
937 938 939

	/* rootfs is available now, try loading default modules */
	load_default_modules();
L
Linus Torvalds 已提交
940
}