main.c 28.4 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8
/*
 *  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
9
 *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
Linus Torvalds's avatar
Linus Torvalds committed
10 11
 */

12 13
#define DEBUG		/* Enable initcall_debug */

Linus Torvalds's avatar
Linus Torvalds committed
14
#include <linux/types.h>
15
#include <linux/extable.h>
Linus Torvalds's avatar
Linus Torvalds committed
16 17
#include <linux/module.h>
#include <linux/proc_fs.h>
18
#include <linux/binfmts.h>
Linus Torvalds's avatar
Linus Torvalds committed
19 20
#include <linux/kernel.h>
#include <linux/syscalls.h>
Ingo Molnar's avatar
Ingo Molnar committed
21
#include <linux/stackprotector.h>
Linus Torvalds's avatar
Linus Torvalds committed
22 23 24 25 26 27 28
#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>
29
#include <linux/acpi.h>
30
#include <linux/console.h>
31
#include <linux/nmi.h>
Linus Torvalds's avatar
Linus Torvalds committed
32 33
#include <linux/percpu.h>
#include <linux/kmod.h>
Nick Piggin's avatar
Nick Piggin committed
34
#include <linux/vmalloc.h>
Linus Torvalds's avatar
Linus Torvalds committed
35
#include <linux/kernel_stat.h>
36
#include <linux/start_kernel.h>
Linus Torvalds's avatar
Linus Torvalds committed
37
#include <linux/security.h>
38
#include <linux/smp.h>
Linus Torvalds's avatar
Linus Torvalds committed
39 40 41 42 43 44 45
#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>
46
#include <linux/cgroup.h>
Linus Torvalds's avatar
Linus Torvalds committed
47
#include <linux/efi.h>
48
#include <linux/tick.h>
49
#include <linux/sched/isolation.h>
50
#include <linux/interrupt.h>
51
#include <linux/taskstats_kern.h>
52
#include <linux/delayacct.h>
Linus Torvalds's avatar
Linus Torvalds committed
53
#include <linux/unistd.h>
54
#include <linux/utsname.h>
Linus Torvalds's avatar
Linus Torvalds committed
55 56 57
#include <linux/rmap.h>
#include <linux/mempolicy.h>
#include <linux/key.h>
Adrian Bunk's avatar
Adrian Bunk committed
58
#include <linux/buffer_head.h>
59
#include <linux/page_ext.h>
60
#include <linux/debug_locks.h>
61
#include <linux/debugobjects.h>
Ingo Molnar's avatar
Ingo Molnar committed
62
#include <linux/lockdep.h>
63
#include <linux/kmemleak.h>
64
#include <linux/pid_namespace.h>
65
#include <linux/device.h>
66
#include <linux/kthread.h>
67
#include <linux/sched.h>
68
#include <linux/sched/init.h>
69
#include <linux/signal.h>
70
#include <linux/idr.h>
71
#include <linux/kgdb.h>
72
#include <linux/ftrace.h>
73
#include <linux/async.h>
74
#include <linux/sfi.h>
75
#include <linux/shmem_fs.h>
76
#include <linux/slab.h>
77
#include <linux/perf_event.h>
78
#include <linux/ptrace.h>
79
#include <linux/pti.h>
80 81
#include <linux/blkdev.h>
#include <linux/elevator.h>
82
#include <linux/sched_clock.h>
83
#include <linux/sched/task.h>
84
#include <linux/sched/task_stack.h>
85
#include <linux/context_tracking.h>
86
#include <linux/random.h>
87
#include <linux/list.h>
88
#include <linux/integrity.h>
89
#include <linux/proc_ns.h>
90
#include <linux/io.h>
91
#include <linux/cache.h>
92
#include <linux/rodata_test.h>
93
#include <linux/jump_label.h>
Linus Torvalds's avatar
Linus Torvalds committed
94 95 96 97

#include <asm/io.h>
#include <asm/bugs.h>
#include <asm/setup.h>
98
#include <asm/sections.h>
99
#include <asm/cacheflush.h>
Linus Torvalds's avatar
Linus Torvalds committed
100

101 102 103
#define CREATE_TRACE_POINTS
#include <trace/events/initcall.h>

104
static int kernel_init(void *);
Linus Torvalds's avatar
Linus Torvalds committed
105 106

extern void init_IRQ(void);
107
extern void fork_init(void);
Linus Torvalds's avatar
Linus Torvalds committed
108 109
extern void radix_tree_init(void);

110 111 112 113 114 115 116 117 118
/*
 * 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;

119
enum system_states system_state __read_mostly;
Linus Torvalds's avatar
Linus Torvalds committed
120 121 122 123 124 125 126 127 128 129
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. */
130
void (*__initdata late_time_init)(void);
Linus Torvalds's avatar
Linus Torvalds committed
131

132 133 134 135 136 137
/* 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;
138 139
/* Command line for per-initcall parameter parsing */
static char *initcall_command_line;
Linus Torvalds's avatar
Linus Torvalds committed
140 141

static char *execute_command;
142
static char *ramdisk_execute_command;
Linus Torvalds's avatar
Linus Torvalds committed
143

144 145 146 147
/*
 * Used to generate warnings if static_key manipulation functions are used
 * before jump_label_init is called.
 */
148
bool static_key_initialized __read_mostly;
149 150
EXPORT_SYMBOL_GPL(static_key_initialized);

Jan Beulich's avatar
Jan Beulich committed
151 152 153 154 155 156
/*
 * 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.
157
 * For ex. kdump situation where previous kernel has crashed, BIOS has been
Jan Beulich's avatar
Jan Beulich committed
158 159 160 161
 * skipped and devices will be in unknown state.
 */
unsigned int reset_devices;
EXPORT_SYMBOL(reset_devices);
Linus Torvalds's avatar
Linus Torvalds committed
162

163 164 165 166 167 168 169 170
static int __init set_reset_devices(char *str)
{
	reset_devices = 1;
	return 1;
}

__setup("reset_devices", set_reset_devices);

171 172
static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
Linus Torvalds's avatar
Linus Torvalds committed
173 174
static const char *panic_later, *panic_param;

175
extern const struct obs_kernel_param __setup_start[], __setup_end[];
Linus Torvalds's avatar
Linus Torvalds committed
176

177
static bool __init obsolete_checksetup(char *line)
Linus Torvalds's avatar
Linus Torvalds committed
178
{
179
	const struct obs_kernel_param *p;
180
	bool had_early_param = false;
Linus Torvalds's avatar
Linus Torvalds committed
181 182 183 184

	p = __setup_start;
	do {
		int n = strlen(p->str);
185
		if (parameqn(line, p->str, n)) {
Linus Torvalds's avatar
Linus Torvalds committed
186
			if (p->early) {
187 188 189 190
				/* 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( */
Linus Torvalds's avatar
Linus Torvalds committed
191
				if (line[n] == '\0' || line[n] == '=')
192
					had_early_param = true;
Linus Torvalds's avatar
Linus Torvalds committed
193
			} else if (!p->setup_func) {
194 195
				pr_warn("Parameter %s is obsolete, ignored\n",
					p->str);
196
				return true;
Linus Torvalds's avatar
Linus Torvalds committed
197
			} else if (p->setup_func(line + n))
198
				return true;
Linus Torvalds's avatar
Linus Torvalds committed
199 200 201
		}
		p++;
	} while (p < __setup_end);
202 203

	return had_early_param;
Linus Torvalds's avatar
Linus Torvalds committed
204 205 206 207 208 209 210 211 212 213 214
}

/*
 * 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)
{
215
	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
216
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
217 218 219 220
}

static int __init quiet_kernel(char *str)
{
221
	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
222
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
223 224
}

225 226
early_param("debug", debug_kernel);
early_param("quiet", quiet_kernel);
Linus Torvalds's avatar
Linus Torvalds committed
227 228 229

static int __init loglevel(char *str)
{
230 231 232 233 234 235 236 237 238 239 240 241 242
	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;
Linus Torvalds's avatar
Linus Torvalds committed
243 244
}

245
early_param("loglevel", loglevel);
Linus Torvalds's avatar
Linus Torvalds committed
246

247
/* Change NUL term back to "=", to make "param" the whole string. */
248 249
static int __init repair_env_string(char *param, char *val,
				    const char *unused, void *arg)
Linus Torvalds's avatar
Linus Torvalds committed
250 251 252 253 254 255 256 257 258 259 260 261
{
	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();
	}
262 263 264
	return 0;
}

265
/* Anything after -- gets handed straight to init. */
266 267
static int __init set_init_arg(char *param, char *val,
			       const char *unused, void *arg)
268 269 270 271 272 273
{
	unsigned int i;

	if (panic_later)
		return 0;

274
	repair_env_string(param, val, unused, NULL);
275 276 277 278 279 280 281 282 283 284 285 286

	for (i = 0; argv_init[i]; i++) {
		if (i == MAX_INIT_ARGS) {
			panic_later = "init";
			panic_param = param;
			return 0;
		}
	}
	argv_init[i] = param;
	return 0;
}

287 288 289 290
/*
 * Unknown boot options get handed to init, unless they look like
 * unused parameters (modprobe will find them in /proc/cmdline).
 */
291 292
static int __init unknown_bootoption(char *param, char *val,
				     const char *unused, void *arg)
293
{
294
	repair_env_string(param, val, unused, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
295 296 297 298 299

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

300 301
	/* Unused module parameter. */
	if (strchr(param, '.') && (!val || strchr(param, '.') < val))
Linus Torvalds's avatar
Linus Torvalds committed
302 303 304 305 306 307 308 309 310 311
		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) {
312
				panic_later = "env";
Linus Torvalds's avatar
Linus Torvalds committed
313 314 315 316 317 318 319 320 321 322 323
				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) {
324
				panic_later = "init";
Linus Torvalds's avatar
Linus Torvalds committed
325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349
				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);

350 351 352 353 354 355 356 357 358 359 360 361
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);

Linus Torvalds's avatar
Linus Torvalds committed
362
#ifndef CONFIG_SMP
363
static const unsigned int setup_max_cpus = NR_CPUS;
364
static inline void setup_nr_cpu_ids(void) { }
Linus Torvalds's avatar
Linus Torvalds committed
365 366 367
static inline void smp_prepare_cpus(unsigned int maxcpus) { }
#endif

368 369 370 371 372 373 374 375
/*
 * 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)
{
376 377 378 379 380
	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);
381 382
	strcpy(saved_command_line, boot_command_line);
	strcpy(static_command_line, command_line);
383 384
}

Linus Torvalds's avatar
Linus Torvalds committed
385 386 387 388 389 390 391 392 393
/*
 * 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.
 */

394 395
static __initdata DECLARE_COMPLETION(kthreadd_done);

396
static noinline void __ref rest_init(void)
Linus Torvalds's avatar
Linus Torvalds committed
397
{
398
	struct task_struct *tsk;
399 400
	int pid;

401
	rcu_scheduler_starting();
402
	/*
Peter Zijlstra's avatar
Peter Zijlstra committed
403
	 * We need to spawn init first so that it obtains pid 1, however
404 405 406
	 * the init task will end up wanting to create kthreads, which, if
	 * we schedule it before we create kthreadd, will OOPS.
	 */
407 408 409 410 411 412 413 414 415 416 417
	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
	/*
	 * Pin init on the boot CPU. Task migration is not properly working
	 * until sched_init_smp() has been run. It will set the allowed
	 * CPUs for init to the non isolated CPUs.
	 */
	rcu_read_lock();
	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
	rcu_read_unlock();

Linus Torvalds's avatar
Linus Torvalds committed
418
	numa_default_policy();
419
	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
420
	rcu_read_lock();
421
	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
422
	rcu_read_unlock();
423 424 425

	/*
	 * Enable might_sleep() and smp_processor_id() checks.
Florian La Roche's avatar
Florian La Roche committed
426
	 * They cannot be enabled earlier because with CONFIG_PREEMPT=y
427 428 429 430 431 432
	 * kernel_thread() would trigger might_sleep() splats. With
	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
	 * already, but it's stuck on the kthreadd_done completion.
	 */
	system_state = SYSTEM_SCHEDULING;

433
	complete(&kthreadd_done);
434 435 436

	/*
	 * The boot idle thread must execute schedule()
437
	 * at least once to get things moving:
438
	 */
439
	schedule_preempt_disabled();
440
	/* Call into cpu_idle with preempt disabled */
441
	cpu_startup_entry(CPUHP_ONLINE);
442
}
Linus Torvalds's avatar
Linus Torvalds committed
443 444

/* Check for early params. */
445 446
static int __init do_early_param(char *param, char *val,
				 const char *unused, void *arg)
Linus Torvalds's avatar
Linus Torvalds committed
447
{
448
	const struct obs_kernel_param *p;
Linus Torvalds's avatar
Linus Torvalds committed
449 450

	for (p = __setup_start; p < __setup_end; p++) {
451
		if ((p->early && parameq(param, p->str)) ||
452 453 454
		    (strcmp(param, "console") == 0 &&
		     strcmp(p->str, "earlycon") == 0)
		) {
Linus Torvalds's avatar
Linus Torvalds committed
455
			if (p->setup_func(val) != 0)
456
				pr_warn("Malformed early option '%s'\n", param);
Linus Torvalds's avatar
Linus Torvalds committed
457 458 459 460 461 462
		}
	}
	/* We accept everything at this stage. */
	return 0;
}

463 464
void __init parse_early_options(char *cmdline)
{
465 466
	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
		   do_early_param);
467 468
}

Linus Torvalds's avatar
Linus Torvalds committed
469 470 471
/* Arch code calls this early on, or if not, just before other parsing. */
void __init parse_early_param(void)
{
472 473
	static int done __initdata;
	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
Linus Torvalds's avatar
Linus Torvalds committed
474 475 476 477 478

	if (done)
		return;

	/* All fall through to do_early_param. */
479
	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
480
	parse_early_options(tmp_cmdline);
Linus Torvalds's avatar
Linus Torvalds committed
481 482 483
	done = 1;
}

484 485
void __init __weak arch_post_acpi_subsys_init(void) { }

486
void __init __weak smp_setup_processor_id(void)
487 488 489
{
}

490
# if THREAD_SIZE >= PAGE_SIZE
491
void __init __weak thread_stack_cache_init(void)
492 493
{
}
494
#endif
495

496 497
void __init __weak mem_encrypt_init(void) { }

498 499
bool initcall_debug;
core_param(initcall_debug, initcall_debug, bool, 0644);
500 501

#ifdef TRACEPOINTS_ENABLED
502
static void __init initcall_debug_enable(void);
503 504 505 506 507
#else
static inline void initcall_debug_enable(void)
{
}
#endif
508

Pekka Enberg's avatar
Pekka Enberg committed
509 510 511 512 513
/*
 * Set up kernel memory allocators
 */
static void __init mm_init(void)
{
514 515 516 517 518
	/*
	 * page_ext requires contiguous pages,
	 * bigger than MAX_ORDER unless SPARSEMEM.
	 */
	page_ext_init_flatmem();
Pekka Enberg's avatar
Pekka Enberg committed
519 520
	mem_init();
	kmem_cache_init();
521
	pgtable_init();
Pekka Enberg's avatar
Pekka Enberg committed
522
	vmalloc_init();
523
	ioremap_huge_init();
524 525
	/* Should be run before the first non-init thread is created */
	init_espfix_bsp();
526 527
	/* Should be run after espfix64 is set up. */
	pti_init();
Pekka Enberg's avatar
Pekka Enberg committed
528 529
}

530
asmlinkage __visible void __init start_kernel(void)
Linus Torvalds's avatar
Linus Torvalds committed
531
{
532 533
	char *command_line;
	char *after_dashes;
534

535
	set_task_stack_end_magic(&init_task);
536
	smp_setup_processor_id();
537
	debug_objects_early_init();
538

539
	cgroup_init_early();
Ingo Molnar's avatar
Ingo Molnar committed
540 541

	local_irq_disable();
542
	early_boot_irqs_disabled = true;
Ingo Molnar's avatar
Ingo Molnar committed
543

544 545 546 547
	/*
	 * Interrupts are still disabled. Do necessary setups, then
	 * enable them.
	 */
548
	boot_cpu_init();
Linus Torvalds's avatar
Linus Torvalds committed
549
	page_address_init();
550
	pr_notice("%s", linux_banner);
Linus Torvalds's avatar
Linus Torvalds committed
551
	setup_arch(&command_line);
552 553
	/*
	 * Set up the the initial canary and entropy after arch
554
	 * and after adding latent and command line entropy.
555 556
	 */
	add_latent_entropy();
557
	add_device_randomness(command_line, strlen(command_line));
558
	boot_init_stack_canary();
559
	mm_init_cpumask(&init_mm);
560
	setup_command_line(command_line);
561
	setup_nr_cpu_ids();
562
	setup_per_cpu_areas();
563
	boot_cpu_state_init();
564
	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
Linus Torvalds's avatar
Linus Torvalds committed
565

566
	build_all_zonelists(NULL);
567 568
	page_alloc_init();

569
	pr_notice("Kernel command line: %s\n", boot_command_line);
570
	parse_early_param();
571 572 573
	after_dashes = parse_args("Booting kernel",
				  static_command_line, __start___param,
				  __stop___param - __start___param,
574
				  -1, -1, NULL, &unknown_bootoption);
575
	if (!IS_ERR_OR_NULL(after_dashes))
576
		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
577
			   NULL, set_init_arg);
578 579 580

	jump_label_init();

581 582 583 584
	/*
	 * These use large bootmem allocations and must precede
	 * kmem_cache_init()
	 */
585
	setup_log_buf(0);
586 587 588
	vfs_caches_init_early();
	sort_main_extable();
	trap_init();
Pekka Enberg's avatar
Pekka Enberg committed
589
	mm_init();
590

591 592
	ftrace_init();

593 594 595
	/* trace_printk can be enabled here */
	early_trace_init();

Linus Torvalds's avatar
Linus Torvalds committed
596 597 598 599 600 601 602 603 604 605 606
	/*
	 * 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();
607 608
	if (WARN(!irqs_disabled(),
		 "Interrupts were enabled *very* early, fixing it\n"))
609
		local_irq_disable();
610
	radix_tree_init();
611

612 613 614 615 616 617
	/*
	 * Set up housekeeping before setting up workqueues to allow the unbound
	 * workqueue to take non-housekeeping into account.
	 */
	housekeeping_init();

618 619 620 621 622 623 624
	/*
	 * Allow workqueue creation and work item queueing/cancelling
	 * early.  Work item execution depends on kthreads and starts after
	 * workqueue_init().
	 */
	workqueue_init_early();

Linus Torvalds's avatar
Linus Torvalds committed
625
	rcu_init();
626

627
	/* Trace events are available after this */
628 629
	trace_init();

630 631 632
	if (initcall_debug)
		initcall_debug_enable();

633
	context_tracking_init();
634 635
	/* init some links before init_ISA_irqs() */
	early_irq_init();
Linus Torvalds's avatar
Linus Torvalds committed
636
	init_IRQ();
Thomas Gleixner's avatar
Thomas Gleixner committed
637
	tick_init();
638
	rcu_init_nohz();
Linus Torvalds's avatar
Linus Torvalds committed
639
	init_timers();
640
	hrtimers_init();
Linus Torvalds's avatar
Linus Torvalds committed
641
	softirq_init();
642
	timekeeping_init();
643
	time_init();
644
	sched_clock_postinit();
645
	printk_safe_init();
646
	perf_event_init();
647
	profile_init();
648
	call_function_init();
649
	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
650
	early_boot_irqs_disabled = false;
651
	local_irq_enable();
652

653
	kmem_cache_init_late();
Linus Torvalds's avatar
Linus Torvalds committed
654 655 656 657 658 659 660 661

	/*
	 * 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)
662 663
		panic("Too many boot %s vars at `%s'", panic_later,
		      panic_param);
Ingo Molnar's avatar
Ingo Molnar committed
664 665 666

	lockdep_info();

667 668 669 670 671 672 673
	/*
	 * 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();

674 675 676 677 678 679 680 681
	/*
	 * This needs to be called before any devices perform DMA
	 * operations that might use the SWIOTLB bounce buffers. It will
	 * mark the bounce buffers as decrypted so that their usage will
	 * not cause "plain-text" data to be decrypted when accessed.
	 */
	mem_encrypt_init();

Linus Torvalds's avatar
Linus Torvalds committed
682 683
#ifdef CONFIG_BLK_DEV_INITRD
	if (initrd_start && !initrd_below_start_ok &&
684
	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
685
		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
686 687
		    page_to_pfn(virt_to_page((void *)initrd_start)),
		    min_low_pfn);
Linus Torvalds's avatar
Linus Torvalds committed
688 689 690
		initrd_start = 0;
	}
#endif
691
	page_ext_init();
692
	kmemleak_init();
693
	debug_objects_mem_init();
694
	setup_per_cpu_pageset();
Linus Torvalds's avatar
Linus Torvalds committed
695
	numa_policy_init();
696
	acpi_early_init();
Linus Torvalds's avatar
Linus Torvalds committed
697 698 699
	if (late_time_init)
		late_time_init();
	calibrate_delay();
700
	pid_idr_init();
Linus Torvalds's avatar
Linus Torvalds committed
701
	anon_vma_init();
702
#ifdef CONFIG_X86
703
	if (efi_enabled(EFI_RUNTIME_SERVICES))
704 705
		efi_enter_virtual_mode();
#endif
706
	thread_stack_cache_init();
707
	cred_init();
708
	fork_init();
Linus Torvalds's avatar
Linus Torvalds committed
709
	proc_caches_init();
710
	uts_ns_init();
Linus Torvalds's avatar
Linus Torvalds committed
711 712 713
	buffer_init();
	key_init();
	security_init();
714
	dbg_late_init();
715
	vfs_caches_init();
716
	pagecache_init();
Linus Torvalds's avatar
Linus Torvalds committed
717
	signals_init();
718
	seq_file_init();
Linus Torvalds's avatar
Linus Torvalds committed
719
	proc_root_init();
720
	nsfs_init();
Linus Torvalds's avatar
Linus Torvalds committed
721
	cpuset_init();
722
	cgroup_init();
723
	taskstats_init_early();
724
	delayacct_init();
Linus Torvalds's avatar
Linus Torvalds committed
725 726 727

	check_bugs();

728
	acpi_subsystem_init();
729
	arch_post_acpi_subsys_init();
730
	sfi_init_late();
Linus Torvalds's avatar
Linus Torvalds committed
731

732
	if (efi_enabled(EFI_RUNTIME_SERVICES)) {
733
		efi_free_boot_services();
734
	}
735

Linus Torvalds's avatar
Linus Torvalds committed
736 737 738 739
	/* Do the rest non-__init'ed, we're now alive */
	rest_init();
}

740 741 742 743
/* Call all constructor functions linked into the kernel. */
static void __init do_ctors(void)
{
#ifdef CONFIG_CONSTRUCTORS
744
	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
745

746 747
	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
		(*fn)();
748 749 750
#endif
}

751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781
#ifdef CONFIG_KALLSYMS
struct blacklist_entry {
	struct list_head next;
	char *buf;
};

static __initdata_or_module LIST_HEAD(blacklisted_initcalls);

static int __init initcall_blacklist(char *str)
{
	char *str_entry;
	struct blacklist_entry *entry;

	/* str argument is a comma-separated list of functions */
	do {
		str_entry = strsep(&str, ",");
		if (str_entry) {
			pr_debug("blacklisting initcall %s\n", str_entry);
			entry = alloc_bootmem(sizeof(*entry));
			entry->buf = alloc_bootmem(strlen(str_entry) + 1);
			strcpy(entry->buf, str_entry);
			list_add(&entry->next, &blacklisted_initcalls);
		}
	} while (str_entry);

	return 0;
}

static bool __init_or_module initcall_blacklisted(initcall_t fn)
{
	struct blacklist_entry *entry;
782
	char fn_name[KSYM_SYMBOL_LEN];
783
	unsigned long addr;
784

785
	if (list_empty(&blacklisted_initcalls))
786 787
		return false;

788 789
	addr = (unsigned long) dereference_function_descriptor(fn);
	sprint_symbol_no_offset(fn_name, addr);
790

791 792 793 794 795 796
	/*
	 * fn will be "function_name [module_name]" where [module_name] is not
	 * displayed for built-in init functions.  Strip off the [module_name].
	 */
	strreplace(fn_name, ' ', '\0');

797
	list_for_each_entry(entry, &blacklisted_initcalls, next) {
798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819
		if (!strcmp(fn_name, entry->buf)) {
			pr_debug("initcall %s blacklisted\n", fn_name);
			return true;
		}
	}

	return false;
}
#else
static int __init initcall_blacklist(char *str)
{
	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
	return 0;
}

static bool __init_or_module initcall_blacklisted(initcall_t fn)
{
	return false;
}
#endif
__setup("initcall_blacklist=", initcall_blacklist);

820 821
static __init_or_module void
trace_initcall_start_cb(void *data, initcall_t fn)
Linus Torvalds's avatar
Linus Torvalds committed
822
{
823
	ktime_t *calltime = (ktime_t *)data;
Linus Torvalds's avatar
Linus Torvalds committed
824

825
	printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
826 827 828 829 830 831 832 833 834 835
	*calltime = ktime_get();
}

static __init_or_module void
trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
{
	ktime_t *calltime = (ktime_t *)data;
	ktime_t delta, rettime;
	unsigned long long duration;

836
	rettime = ktime_get();
837
	delta = ktime_sub(rettime, *calltime);
838
	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
839
	printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
840
		 fn, ret, duration);
841
}
Linus Torvalds's avatar
Linus Torvalds committed
842

843 844
static ktime_t initcall_calltime;

845
#ifdef TRACEPOINTS_ENABLED
846 847 848 849 850 851 852 853 854
static void __init initcall_debug_enable(void)
{
	int ret;

	ret = register_trace_initcall_start(trace_initcall_start_cb,
					    &initcall_calltime);
	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
					      &initcall_calltime);
	WARN(ret, "Failed to register initcall tracepoints\n");
855
}
856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871
# define do_trace_initcall_start	trace_initcall_start
# define do_trace_initcall_finish	trace_initcall_finish
#else
static inline void do_trace_initcall_start(initcall_t fn)
{
	if (!initcall_debug)
		return;
	trace_initcall_start_cb(&initcall_calltime, fn);
}
static inline void do_trace_initcall_finish(initcall_t fn, int ret)
{
	if (!initcall_debug)
		return;
	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
}
#endif /* !TRACEPOINTS_ENABLED */
872

873
int __init_or_module do_one_initcall(initcall_t fn)
874 875
{
	int count = preempt_count();
876
	char msgbuf[64];
877
	int ret;
878

879 880 881
	if (initcall_blacklisted(fn))
		return -EPERM;

882
	do_trace_initcall_start(fn);
883
	ret = fn();
884
	do_trace_initcall_finish(fn, ret);
885

886
	msgbuf[0] = 0;
887

888
	if (preempt_count() != count) {
889
		sprintf(msgbuf, "preemption imbalance ");
890
		preempt_count_set(count);
Linus Torvalds's avatar
Linus Torvalds committed
891
	}
892
	if (irqs_disabled()) {
893
		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
894 895
		local_irq_enable();
	}
896
	WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
897

898
	add_latent_entropy();
Américo Wang's avatar
Américo Wang committed
899
	return ret;
900 901 902
}


903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925
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,
};

926
/* Keep these in sync with initcalls in include/linux/init.h */
927
static char *initcall_level_names[] __initdata = {
928
	"pure",
929 930 931 932 933 934 935
	"core",
	"postcore",
	"arch",
	"subsys",
	"fs",
	"device",
	"late",
936 937 938
};

static void __init do_initcall_level(int level)
939
{
940
	initcall_t *fn;
941

942
	strcpy(initcall_command_line, saved_command_line);
943
	parse_args(initcall_level_names[level],
944
		   initcall_command_line, __start___param,
945 946
		   __stop___param - __start___param,
		   level, level,
947
		   NULL, &repair_env_string);
948

949
	trace_initcall_level(initcall_level_names[level]);
950
	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
951
		do_one_initcall(*fn);
Linus Torvalds's avatar
Linus Torvalds committed
952 953
}

954 955 956 957
static void __init do_initcalls(void)
{
	int level;

958
	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
959 960 961
		do_initcall_level(level);
}

Linus Torvalds's avatar
Linus Torvalds committed
962 963 964 965 966 967 968 969 970
/*
 * 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)
{
971
	cpuset_init_smp();
972
	shmem_init();
Linus Torvalds's avatar
Linus Torvalds committed
973
	driver_init();
974
	init_irq_proc();
975
	do_ctors();
976
	usermodehelper_enable();
977
	do_initcalls();
Linus Torvalds's avatar
Linus Torvalds committed
978 979
}

980
static void __init do_pre_smp_initcalls(void)
981
{
982
	initcall_t *fn;
983

984
	trace_initcall_level("early");
985
	for (fn = __initcall_start; fn < __initcall0_start; fn++)
986
		do_one_initcall(*fn);
987 988
}

989 990 991 992 993 994 995 996 997 998 999
/*
 * 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();
}

1000
static int run_init_process(const char *init_filename)
Linus Torvalds's avatar
Linus Torvalds committed
1001 1002
{
	argv_init[0] = init_filename;
1003
	return do_execve(getname_kernel(init_filename),
1004 1005
		(const char __user *const __user *)argv_init,
		(const char __user *const __user *)envp_init);
Linus Torvalds's avatar
Linus Torvalds committed
1006 1007
}

1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
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;
}

1022
static noinline void __init kernel_init_freeable(void);
1023

1024
#if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1025
bool rodata_enabled __ro_after_init = true;
1026 1027 1028 1029 1030
static int __init set_debug_rodata(char *str)
{
	return strtobool(str, &rodata_enabled);
}
__setup("rodata=", set_debug_rodata);
1031
#endif
1032

1033
#ifdef CONFIG_STRICT_KERNEL_RWX
1034 1035
static void mark_readonly(void)
{
1036
	if (rodata_enabled) {
1037 1038 1039 1040 1041 1042 1043
		/*
		 * load_module() results in W+X mappings, which are cleaned up
		 * with call_rcu_sched().  Let's make sure that queued work is
		 * flushed so that we don't hit false positives looking for
		 * insecure pages which are W+X.
		 */
		rcu_barrier_sched();
1044
		mark_rodata_ro();
1045 1046
		rodata_test();
	} else
1047 1048 1049 1050 1051 1052 1053 1054 1055
		pr_info("Kernel memory protection disabled.\n");
}
#else
static inline void mark_readonly(void)
{
	pr_warn("This architecture does not have kernel memory protection.\n");
}
#endif

1056
static int __ref kernel_init(void *unused)
1057
{
1058 1059
	int ret;

1060
	kernel_init_freeable();
1061 1062
	/* need to finish all async __init code before freeing the memory */
	async_synchronize_full();
1063
	ftrace_free_init_mem();
1064
	jump_label_invalidate_initmem();
1065
	free_initmem();
1066
	mark_readonly();
1067 1068 1069
	system_state = SYSTEM_RUNNING;
	numa_default_policy();

1070 1071
	rcu_end_inkernel_boot();

1072
	if (ramdisk_execute_command) {
1073 1074
		ret = run_init_process(ramdisk_execute_command);
		if (!ret)
1075
			return 0;
1076 1077
		pr_err("Failed to execute %s (error %d)\n",
		       ramdisk_execute_command, ret);
1078 1079 1080 1081 1082 1083 1084 1085 1086
	}

	/*
	 * 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) {
1087 1088
		ret = run_init_process(execute_command);
		if (!ret)
1089
			return 0;
1090 1091
		panic("Requested init %s failed (error %d).",
		      execute_command, ret);
1092
	}
1093 1094 1095 1096
	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"))
1097
		return 0;
1098

1099
	panic("No working init found.  Try passing init= option to kernel. "
1100
	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1101 1102
}

1103
static noinline void __init kernel_init_freeable(void)
Linus Torvalds's avatar
Linus Torvalds committed
1104
{
1105 1106 1107 1108
	/*
	 * Wait until kthreadd is all set-up.
	 */
	wait_for_completion(&kthreadd_done);
1109 1110 1111 1112

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

1113 1114 1115
	/*
	 * init can allocate pages on any node
	 */
1116
	set_mems_allowed(node_states[N_MEMORY]);