xref: /linux/init/main.c (revision 45dcbb6f5ef78b0a9c1b91bea2f6f227642a65aa)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/init/main.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
8  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
11  */
12 
13 #define DEBUG		/* Enable initcall_debug */
14 
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/vmalloc.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/start_kernel.h>
39 #include <linux/security.h>
40 #include <linux/smp.h>
41 #include <linux/profile.h>
42 #include <linux/rcupdate.h>
43 #include <linux/moduleparam.h>
44 #include <linux/kallsyms.h>
45 #include <linux/writeback.h>
46 #include <linux/cpu.h>
47 #include <linux/cpuset.h>
48 #include <linux/cgroup.h>
49 #include <linux/efi.h>
50 #include <linux/tick.h>
51 #include <linux/sched/isolation.h>
52 #include <linux/interrupt.h>
53 #include <linux/taskstats_kern.h>
54 #include <linux/delayacct.h>
55 #include <linux/unistd.h>
56 #include <linux/utsname.h>
57 #include <linux/rmap.h>
58 #include <linux/mempolicy.h>
59 #include <linux/key.h>
60 #include <linux/buffer_head.h>
61 #include <linux/page_ext.h>
62 #include <linux/debug_locks.h>
63 #include <linux/debugobjects.h>
64 #include <linux/lockdep.h>
65 #include <linux/kmemleak.h>
66 #include <linux/padata.h>
67 #include <linux/pid_namespace.h>
68 #include <linux/device/driver.h>
69 #include <linux/kthread.h>
70 #include <linux/sched.h>
71 #include <linux/sched/init.h>
72 #include <linux/signal.h>
73 #include <linux/idr.h>
74 #include <linux/kgdb.h>
75 #include <linux/ftrace.h>
76 #include <linux/async.h>
77 #include <linux/sfi.h>
78 #include <linux/shmem_fs.h>
79 #include <linux/slab.h>
80 #include <linux/perf_event.h>
81 #include <linux/ptrace.h>
82 #include <linux/pti.h>
83 #include <linux/blkdev.h>
84 #include <linux/elevator.h>
85 #include <linux/sched/clock.h>
86 #include <linux/sched/task.h>
87 #include <linux/sched/task_stack.h>
88 #include <linux/context_tracking.h>
89 #include <linux/random.h>
90 #include <linux/list.h>
91 #include <linux/integrity.h>
92 #include <linux/proc_ns.h>
93 #include <linux/io.h>
94 #include <linux/cache.h>
95 #include <linux/rodata_test.h>
96 #include <linux/jump_label.h>
97 #include <linux/mem_encrypt.h>
98 #include <linux/kcsan.h>
99 #include <linux/init_syscalls.h>
100 
101 #include <asm/io.h>
102 #include <asm/bugs.h>
103 #include <asm/setup.h>
104 #include <asm/sections.h>
105 #include <asm/cacheflush.h>
106 
107 #define CREATE_TRACE_POINTS
108 #include <trace/events/initcall.h>
109 
110 #include <kunit/test.h>
111 
112 static int kernel_init(void *);
113 
114 extern void init_IRQ(void);
115 extern void radix_tree_init(void);
116 
117 /*
118  * Debug helper: via this flag we know that we are in 'early bootup code'
119  * where only the boot processor is running with IRQ disabled.  This means
120  * two things - IRQ must not be enabled before the flag is cleared and some
121  * operations which are not allowed with IRQ disabled are allowed while the
122  * flag is set.
123  */
124 bool early_boot_irqs_disabled __read_mostly;
125 
126 enum system_states system_state __read_mostly;
127 EXPORT_SYMBOL(system_state);
128 
129 /*
130  * Boot command-line arguments
131  */
132 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
133 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
134 
135 extern void time_init(void);
136 /* Default late time init is NULL. archs can override this later. */
137 void (*__initdata late_time_init)(void);
138 
139 /* Untouched command line saved by arch-specific code. */
140 char __initdata boot_command_line[COMMAND_LINE_SIZE];
141 /* Untouched saved command line (eg. for /proc) */
142 char *saved_command_line;
143 /* Command line for parameter parsing */
144 static char *static_command_line;
145 /* Untouched extra command line */
146 static char *extra_command_line;
147 /* Extra init arguments */
148 static char *extra_init_args;
149 
150 #ifdef CONFIG_BOOT_CONFIG
151 /* Is bootconfig on command line? */
152 static bool bootconfig_found;
153 static bool initargs_found;
154 #else
155 # define bootconfig_found false
156 # define initargs_found false
157 #endif
158 
159 static char *execute_command;
160 static char *ramdisk_execute_command = "/init";
161 
162 /*
163  * Used to generate warnings if static_key manipulation functions are used
164  * before jump_label_init is called.
165  */
166 bool static_key_initialized __read_mostly;
167 EXPORT_SYMBOL_GPL(static_key_initialized);
168 
169 /*
170  * If set, this is an indication to the drivers that reset the underlying
171  * device before going ahead with the initialization otherwise driver might
172  * rely on the BIOS and skip the reset operation.
173  *
174  * This is useful if kernel is booting in an unreliable environment.
175  * For ex. kdump situation where previous kernel has crashed, BIOS has been
176  * skipped and devices will be in unknown state.
177  */
178 unsigned int reset_devices;
179 EXPORT_SYMBOL(reset_devices);
180 
181 static int __init set_reset_devices(char *str)
182 {
183 	reset_devices = 1;
184 	return 1;
185 }
186 
187 __setup("reset_devices", set_reset_devices);
188 
189 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
190 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
191 static const char *panic_later, *panic_param;
192 
193 extern const struct obs_kernel_param __setup_start[], __setup_end[];
194 
195 static bool __init obsolete_checksetup(char *line)
196 {
197 	const struct obs_kernel_param *p;
198 	bool had_early_param = false;
199 
200 	p = __setup_start;
201 	do {
202 		int n = strlen(p->str);
203 		if (parameqn(line, p->str, n)) {
204 			if (p->early) {
205 				/* Already done in parse_early_param?
206 				 * (Needs exact match on param part).
207 				 * Keep iterating, as we can have early
208 				 * params and __setups of same names 8( */
209 				if (line[n] == '\0' || line[n] == '=')
210 					had_early_param = true;
211 			} else if (!p->setup_func) {
212 				pr_warn("Parameter %s is obsolete, ignored\n",
213 					p->str);
214 				return true;
215 			} else if (p->setup_func(line + n))
216 				return true;
217 		}
218 		p++;
219 	} while (p < __setup_end);
220 
221 	return had_early_param;
222 }
223 
224 /*
225  * This should be approx 2 Bo*oMips to start (note initial shift), and will
226  * still work even if initially too large, it will just take slightly longer
227  */
228 unsigned long loops_per_jiffy = (1<<12);
229 EXPORT_SYMBOL(loops_per_jiffy);
230 
231 static int __init debug_kernel(char *str)
232 {
233 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
234 	return 0;
235 }
236 
237 static int __init quiet_kernel(char *str)
238 {
239 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
240 	return 0;
241 }
242 
243 early_param("debug", debug_kernel);
244 early_param("quiet", quiet_kernel);
245 
246 static int __init loglevel(char *str)
247 {
248 	int newlevel;
249 
250 	/*
251 	 * Only update loglevel value when a correct setting was passed,
252 	 * to prevent blind crashes (when loglevel being set to 0) that
253 	 * are quite hard to debug
254 	 */
255 	if (get_option(&str, &newlevel)) {
256 		console_loglevel = newlevel;
257 		return 0;
258 	}
259 
260 	return -EINVAL;
261 }
262 
263 early_param("loglevel", loglevel);
264 
265 #ifdef CONFIG_BLK_DEV_INITRD
266 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
267 {
268 	u32 size, csum;
269 	char *data;
270 	u32 *hdr;
271 
272 	if (!initrd_end)
273 		return NULL;
274 
275 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
276 	if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
277 		return NULL;
278 
279 	hdr = (u32 *)(data - 8);
280 	size = hdr[0];
281 	csum = hdr[1];
282 
283 	data = ((void *)hdr) - size;
284 	if ((unsigned long)data < initrd_start) {
285 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
286 			size, initrd_end - initrd_start);
287 		return NULL;
288 	}
289 
290 	/* Remove bootconfig from initramfs/initrd */
291 	initrd_end = (unsigned long)data;
292 	if (_size)
293 		*_size = size;
294 	if (_csum)
295 		*_csum = csum;
296 
297 	return data;
298 }
299 #else
300 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
301 {
302 	return NULL;
303 }
304 #endif
305 
306 #ifdef CONFIG_BOOT_CONFIG
307 
308 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
309 
310 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
311 
312 static int __init xbc_snprint_cmdline(char *buf, size_t size,
313 				      struct xbc_node *root)
314 {
315 	struct xbc_node *knode, *vnode;
316 	char *end = buf + size;
317 	const char *val;
318 	int ret;
319 
320 	xbc_node_for_each_key_value(root, knode, val) {
321 		ret = xbc_node_compose_key_after(root, knode,
322 					xbc_namebuf, XBC_KEYLEN_MAX);
323 		if (ret < 0)
324 			return ret;
325 
326 		vnode = xbc_node_get_child(knode);
327 		if (!vnode) {
328 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
329 			if (ret < 0)
330 				return ret;
331 			buf += ret;
332 			continue;
333 		}
334 		xbc_array_for_each_value(vnode, val) {
335 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
336 				       xbc_namebuf, val);
337 			if (ret < 0)
338 				return ret;
339 			buf += ret;
340 		}
341 	}
342 
343 	return buf - (end - size);
344 }
345 #undef rest
346 
347 /* Make an extra command line under given key word */
348 static char * __init xbc_make_cmdline(const char *key)
349 {
350 	struct xbc_node *root;
351 	char *new_cmdline;
352 	int ret, len = 0;
353 
354 	root = xbc_find_node(key);
355 	if (!root)
356 		return NULL;
357 
358 	/* Count required buffer size */
359 	len = xbc_snprint_cmdline(NULL, 0, root);
360 	if (len <= 0)
361 		return NULL;
362 
363 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
364 	if (!new_cmdline) {
365 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
366 		return NULL;
367 	}
368 
369 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
370 	if (ret < 0 || ret > len) {
371 		pr_err("Failed to print extra kernel cmdline.\n");
372 		return NULL;
373 	}
374 
375 	return new_cmdline;
376 }
377 
378 static u32 boot_config_checksum(unsigned char *p, u32 size)
379 {
380 	u32 ret = 0;
381 
382 	while (size--)
383 		ret += *p++;
384 
385 	return ret;
386 }
387 
388 static int __init bootconfig_params(char *param, char *val,
389 				    const char *unused, void *arg)
390 {
391 	if (strcmp(param, "bootconfig") == 0) {
392 		bootconfig_found = true;
393 	}
394 	return 0;
395 }
396 
397 static void __init setup_boot_config(const char *cmdline)
398 {
399 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
400 	const char *msg;
401 	int pos;
402 	u32 size, csum;
403 	char *data, *copy, *err;
404 	int ret;
405 
406 	/* Cut out the bootconfig data even if we have no bootconfig option */
407 	data = get_boot_config_from_initrd(&size, &csum);
408 
409 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
410 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
411 			 bootconfig_params);
412 
413 	if (IS_ERR(err) || !bootconfig_found)
414 		return;
415 
416 	/* parse_args() stops at '--' and returns an address */
417 	if (err)
418 		initargs_found = true;
419 
420 	if (!data) {
421 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
422 		return;
423 	}
424 
425 	if (size >= XBC_DATA_MAX) {
426 		pr_err("bootconfig size %d greater than max size %d\n",
427 			size, XBC_DATA_MAX);
428 		return;
429 	}
430 
431 	if (boot_config_checksum((unsigned char *)data, size) != csum) {
432 		pr_err("bootconfig checksum failed\n");
433 		return;
434 	}
435 
436 	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
437 	if (!copy) {
438 		pr_err("Failed to allocate memory for bootconfig\n");
439 		return;
440 	}
441 
442 	memcpy(copy, data, size);
443 	copy[size] = '\0';
444 
445 	ret = xbc_init(copy, &msg, &pos);
446 	if (ret < 0) {
447 		if (pos < 0)
448 			pr_err("Failed to init bootconfig: %s.\n", msg);
449 		else
450 			pr_err("Failed to parse bootconfig: %s at %d.\n",
451 				msg, pos);
452 	} else {
453 		pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
454 		/* keys starting with "kernel." are passed via cmdline */
455 		extra_command_line = xbc_make_cmdline("kernel");
456 		/* Also, "init." keys are init arguments */
457 		extra_init_args = xbc_make_cmdline("init");
458 	}
459 	return;
460 }
461 
462 #else
463 
464 static void __init setup_boot_config(const char *cmdline)
465 {
466 	/* Remove bootconfig data from initrd */
467 	get_boot_config_from_initrd(NULL, NULL);
468 }
469 
470 static int __init warn_bootconfig(char *str)
471 {
472 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
473 	return 0;
474 }
475 early_param("bootconfig", warn_bootconfig);
476 
477 #endif
478 
479 /* Change NUL term back to "=", to make "param" the whole string. */
480 static void __init repair_env_string(char *param, char *val)
481 {
482 	if (val) {
483 		/* param=val or param="val"? */
484 		if (val == param+strlen(param)+1)
485 			val[-1] = '=';
486 		else if (val == param+strlen(param)+2) {
487 			val[-2] = '=';
488 			memmove(val-1, val, strlen(val)+1);
489 		} else
490 			BUG();
491 	}
492 }
493 
494 /* Anything after -- gets handed straight to init. */
495 static int __init set_init_arg(char *param, char *val,
496 			       const char *unused, void *arg)
497 {
498 	unsigned int i;
499 
500 	if (panic_later)
501 		return 0;
502 
503 	repair_env_string(param, val);
504 
505 	for (i = 0; argv_init[i]; i++) {
506 		if (i == MAX_INIT_ARGS) {
507 			panic_later = "init";
508 			panic_param = param;
509 			return 0;
510 		}
511 	}
512 	argv_init[i] = param;
513 	return 0;
514 }
515 
516 /*
517  * Unknown boot options get handed to init, unless they look like
518  * unused parameters (modprobe will find them in /proc/cmdline).
519  */
520 static int __init unknown_bootoption(char *param, char *val,
521 				     const char *unused, void *arg)
522 {
523 	size_t len = strlen(param);
524 
525 	repair_env_string(param, val);
526 
527 	/* Handle obsolete-style parameters */
528 	if (obsolete_checksetup(param))
529 		return 0;
530 
531 	/* Unused module parameter. */
532 	if (strnchr(param, len, '.'))
533 		return 0;
534 
535 	if (panic_later)
536 		return 0;
537 
538 	if (val) {
539 		/* Environment option */
540 		unsigned int i;
541 		for (i = 0; envp_init[i]; i++) {
542 			if (i == MAX_INIT_ENVS) {
543 				panic_later = "env";
544 				panic_param = param;
545 			}
546 			if (!strncmp(param, envp_init[i], len+1))
547 				break;
548 		}
549 		envp_init[i] = param;
550 	} else {
551 		/* Command line option */
552 		unsigned int i;
553 		for (i = 0; argv_init[i]; i++) {
554 			if (i == MAX_INIT_ARGS) {
555 				panic_later = "init";
556 				panic_param = param;
557 			}
558 		}
559 		argv_init[i] = param;
560 	}
561 	return 0;
562 }
563 
564 static int __init init_setup(char *str)
565 {
566 	unsigned int i;
567 
568 	execute_command = str;
569 	/*
570 	 * In case LILO is going to boot us with default command line,
571 	 * it prepends "auto" before the whole cmdline which makes
572 	 * the shell think it should execute a script with such name.
573 	 * So we ignore all arguments entered _before_ init=... [MJ]
574 	 */
575 	for (i = 1; i < MAX_INIT_ARGS; i++)
576 		argv_init[i] = NULL;
577 	return 1;
578 }
579 __setup("init=", init_setup);
580 
581 static int __init rdinit_setup(char *str)
582 {
583 	unsigned int i;
584 
585 	ramdisk_execute_command = str;
586 	/* See "auto" comment in init_setup */
587 	for (i = 1; i < MAX_INIT_ARGS; i++)
588 		argv_init[i] = NULL;
589 	return 1;
590 }
591 __setup("rdinit=", rdinit_setup);
592 
593 #ifndef CONFIG_SMP
594 static const unsigned int setup_max_cpus = NR_CPUS;
595 static inline void setup_nr_cpu_ids(void) { }
596 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
597 #endif
598 
599 /*
600  * We need to store the untouched command line for future reference.
601  * We also need to store the touched command line since the parameter
602  * parsing is performed in place, and we should allow a component to
603  * store reference of name/value for future reference.
604  */
605 static void __init setup_command_line(char *command_line)
606 {
607 	size_t len, xlen = 0, ilen = 0;
608 
609 	if (extra_command_line)
610 		xlen = strlen(extra_command_line);
611 	if (extra_init_args)
612 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
613 
614 	len = xlen + strlen(boot_command_line) + 1;
615 
616 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
617 	if (!saved_command_line)
618 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
619 
620 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
621 	if (!static_command_line)
622 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
623 
624 	if (xlen) {
625 		/*
626 		 * We have to put extra_command_line before boot command
627 		 * lines because there could be dashes (separator of init
628 		 * command line) in the command lines.
629 		 */
630 		strcpy(saved_command_line, extra_command_line);
631 		strcpy(static_command_line, extra_command_line);
632 	}
633 	strcpy(saved_command_line + xlen, boot_command_line);
634 	strcpy(static_command_line + xlen, command_line);
635 
636 	if (ilen) {
637 		/*
638 		 * Append supplemental init boot args to saved_command_line
639 		 * so that user can check what command line options passed
640 		 * to init.
641 		 */
642 		len = strlen(saved_command_line);
643 		if (initargs_found) {
644 			saved_command_line[len++] = ' ';
645 		} else {
646 			strcpy(saved_command_line + len, " -- ");
647 			len += 4;
648 		}
649 
650 		strcpy(saved_command_line + len, extra_init_args);
651 	}
652 }
653 
654 /*
655  * We need to finalize in a non-__init function or else race conditions
656  * between the root thread and the init thread may cause start_kernel to
657  * be reaped by free_initmem before the root thread has proceeded to
658  * cpu_idle.
659  *
660  * gcc-3.4 accidentally inlines this function, so use noinline.
661  */
662 
663 static __initdata DECLARE_COMPLETION(kthreadd_done);
664 
665 noinline void __ref rest_init(void)
666 {
667 	struct task_struct *tsk;
668 	int pid;
669 
670 	rcu_scheduler_starting();
671 	/*
672 	 * We need to spawn init first so that it obtains pid 1, however
673 	 * the init task will end up wanting to create kthreads, which, if
674 	 * we schedule it before we create kthreadd, will OOPS.
675 	 */
676 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
677 	/*
678 	 * Pin init on the boot CPU. Task migration is not properly working
679 	 * until sched_init_smp() has been run. It will set the allowed
680 	 * CPUs for init to the non isolated CPUs.
681 	 */
682 	rcu_read_lock();
683 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
684 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
685 	rcu_read_unlock();
686 
687 	numa_default_policy();
688 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
689 	rcu_read_lock();
690 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
691 	rcu_read_unlock();
692 
693 	/*
694 	 * Enable might_sleep() and smp_processor_id() checks.
695 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
696 	 * kernel_thread() would trigger might_sleep() splats. With
697 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
698 	 * already, but it's stuck on the kthreadd_done completion.
699 	 */
700 	system_state = SYSTEM_SCHEDULING;
701 
702 	complete(&kthreadd_done);
703 
704 	/*
705 	 * The boot idle thread must execute schedule()
706 	 * at least once to get things moving:
707 	 */
708 	schedule_preempt_disabled();
709 	/* Call into cpu_idle with preempt disabled */
710 	cpu_startup_entry(CPUHP_ONLINE);
711 }
712 
713 /* Check for early params. */
714 static int __init do_early_param(char *param, char *val,
715 				 const char *unused, void *arg)
716 {
717 	const struct obs_kernel_param *p;
718 
719 	for (p = __setup_start; p < __setup_end; p++) {
720 		if ((p->early && parameq(param, p->str)) ||
721 		    (strcmp(param, "console") == 0 &&
722 		     strcmp(p->str, "earlycon") == 0)
723 		) {
724 			if (p->setup_func(val) != 0)
725 				pr_warn("Malformed early option '%s'\n", param);
726 		}
727 	}
728 	/* We accept everything at this stage. */
729 	return 0;
730 }
731 
732 void __init parse_early_options(char *cmdline)
733 {
734 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
735 		   do_early_param);
736 }
737 
738 /* Arch code calls this early on, or if not, just before other parsing. */
739 void __init parse_early_param(void)
740 {
741 	static int done __initdata;
742 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
743 
744 	if (done)
745 		return;
746 
747 	/* All fall through to do_early_param. */
748 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
749 	parse_early_options(tmp_cmdline);
750 	done = 1;
751 }
752 
753 void __init __weak arch_post_acpi_subsys_init(void) { }
754 
755 void __init __weak smp_setup_processor_id(void)
756 {
757 }
758 
759 # if THREAD_SIZE >= PAGE_SIZE
760 void __init __weak thread_stack_cache_init(void)
761 {
762 }
763 #endif
764 
765 void __init __weak mem_encrypt_init(void) { }
766 
767 void __init __weak poking_init(void) { }
768 
769 void __init __weak pgtable_cache_init(void) { }
770 
771 bool initcall_debug;
772 core_param(initcall_debug, initcall_debug, bool, 0644);
773 
774 #ifdef TRACEPOINTS_ENABLED
775 static void __init initcall_debug_enable(void);
776 #else
777 static inline void initcall_debug_enable(void)
778 {
779 }
780 #endif
781 
782 /* Report memory auto-initialization states for this boot. */
783 static void __init report_meminit(void)
784 {
785 	const char *stack;
786 
787 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
788 		stack = "all(pattern)";
789 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
790 		stack = "all(zero)";
791 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
792 		stack = "byref_all(zero)";
793 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
794 		stack = "byref(zero)";
795 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
796 		stack = "__user(zero)";
797 	else
798 		stack = "off";
799 
800 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
801 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
802 		want_init_on_free() ? "on" : "off");
803 	if (want_init_on_free())
804 		pr_info("mem auto-init: clearing system memory may take some time...\n");
805 }
806 
807 /*
808  * Set up kernel memory allocators
809  */
810 static void __init mm_init(void)
811 {
812 	/*
813 	 * page_ext requires contiguous pages,
814 	 * bigger than MAX_ORDER unless SPARSEMEM.
815 	 */
816 	page_ext_init_flatmem();
817 	init_debug_pagealloc();
818 	report_meminit();
819 	mem_init();
820 	kmem_cache_init();
821 	kmemleak_init();
822 	pgtable_init();
823 	debug_objects_mem_init();
824 	vmalloc_init();
825 	ioremap_huge_init();
826 	/* Should be run before the first non-init thread is created */
827 	init_espfix_bsp();
828 	/* Should be run after espfix64 is set up. */
829 	pti_init();
830 }
831 
832 void __init __weak arch_call_rest_init(void)
833 {
834 	rest_init();
835 }
836 
837 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
838 {
839 	char *command_line;
840 	char *after_dashes;
841 
842 	set_task_stack_end_magic(&init_task);
843 	smp_setup_processor_id();
844 	debug_objects_early_init();
845 
846 	cgroup_init_early();
847 
848 	local_irq_disable();
849 	early_boot_irqs_disabled = true;
850 
851 	/*
852 	 * Interrupts are still disabled. Do necessary setups, then
853 	 * enable them.
854 	 */
855 	boot_cpu_init();
856 	page_address_init();
857 	pr_notice("%s", linux_banner);
858 	early_security_init();
859 	setup_arch(&command_line);
860 	setup_boot_config(command_line);
861 	setup_command_line(command_line);
862 	setup_nr_cpu_ids();
863 	setup_per_cpu_areas();
864 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
865 	boot_cpu_hotplug_init();
866 
867 	build_all_zonelists(NULL);
868 	page_alloc_init();
869 
870 	pr_notice("Kernel command line: %s\n", saved_command_line);
871 	/* parameters may set static keys */
872 	jump_label_init();
873 	parse_early_param();
874 	after_dashes = parse_args("Booting kernel",
875 				  static_command_line, __start___param,
876 				  __stop___param - __start___param,
877 				  -1, -1, NULL, &unknown_bootoption);
878 	if (!IS_ERR_OR_NULL(after_dashes))
879 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
880 			   NULL, set_init_arg);
881 	if (extra_init_args)
882 		parse_args("Setting extra init args", extra_init_args,
883 			   NULL, 0, -1, -1, NULL, set_init_arg);
884 
885 	/*
886 	 * These use large bootmem allocations and must precede
887 	 * kmem_cache_init()
888 	 */
889 	setup_log_buf(0);
890 	vfs_caches_init_early();
891 	sort_main_extable();
892 	trap_init();
893 	mm_init();
894 
895 	ftrace_init();
896 
897 	/* trace_printk can be enabled here */
898 	early_trace_init();
899 
900 	/*
901 	 * Set up the scheduler prior starting any interrupts (such as the
902 	 * timer interrupt). Full topology setup happens at smp_init()
903 	 * time - but meanwhile we still have a functioning scheduler.
904 	 */
905 	sched_init();
906 	/*
907 	 * Disable preemption - early bootup scheduling is extremely
908 	 * fragile until we cpu_idle() for the first time.
909 	 */
910 	preempt_disable();
911 	if (WARN(!irqs_disabled(),
912 		 "Interrupts were enabled *very* early, fixing it\n"))
913 		local_irq_disable();
914 	radix_tree_init();
915 
916 	/*
917 	 * Set up housekeeping before setting up workqueues to allow the unbound
918 	 * workqueue to take non-housekeeping into account.
919 	 */
920 	housekeeping_init();
921 
922 	/*
923 	 * Allow workqueue creation and work item queueing/cancelling
924 	 * early.  Work item execution depends on kthreads and starts after
925 	 * workqueue_init().
926 	 */
927 	workqueue_init_early();
928 
929 	rcu_init();
930 
931 	/* Trace events are available after this */
932 	trace_init();
933 
934 	if (initcall_debug)
935 		initcall_debug_enable();
936 
937 	context_tracking_init();
938 	/* init some links before init_ISA_irqs() */
939 	early_irq_init();
940 	init_IRQ();
941 	tick_init();
942 	rcu_init_nohz();
943 	init_timers();
944 	hrtimers_init();
945 	softirq_init();
946 	timekeeping_init();
947 
948 	/*
949 	 * For best initial stack canary entropy, prepare it after:
950 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
951 	 * - timekeeping_init() for ktime entropy used in rand_initialize()
952 	 * - rand_initialize() to get any arch-specific entropy like RDRAND
953 	 * - add_latent_entropy() to get any latent entropy
954 	 * - adding command line entropy
955 	 */
956 	rand_initialize();
957 	add_latent_entropy();
958 	add_device_randomness(command_line, strlen(command_line));
959 	boot_init_stack_canary();
960 
961 	time_init();
962 	perf_event_init();
963 	profile_init();
964 	call_function_init();
965 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
966 
967 	early_boot_irqs_disabled = false;
968 	local_irq_enable();
969 
970 	kmem_cache_init_late();
971 
972 	/*
973 	 * HACK ALERT! This is early. We're enabling the console before
974 	 * we've done PCI setups etc, and console_init() must be aware of
975 	 * this. But we do want output early, in case something goes wrong.
976 	 */
977 	console_init();
978 	if (panic_later)
979 		panic("Too many boot %s vars at `%s'", panic_later,
980 		      panic_param);
981 
982 	lockdep_init();
983 
984 	/*
985 	 * Need to run this when irqs are enabled, because it wants
986 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
987 	 * too:
988 	 */
989 	locking_selftest();
990 
991 	/*
992 	 * This needs to be called before any devices perform DMA
993 	 * operations that might use the SWIOTLB bounce buffers. It will
994 	 * mark the bounce buffers as decrypted so that their usage will
995 	 * not cause "plain-text" data to be decrypted when accessed.
996 	 */
997 	mem_encrypt_init();
998 
999 #ifdef CONFIG_BLK_DEV_INITRD
1000 	if (initrd_start && !initrd_below_start_ok &&
1001 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1002 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1003 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1004 		    min_low_pfn);
1005 		initrd_start = 0;
1006 	}
1007 #endif
1008 	setup_per_cpu_pageset();
1009 	numa_policy_init();
1010 	acpi_early_init();
1011 	if (late_time_init)
1012 		late_time_init();
1013 	sched_clock_init();
1014 	calibrate_delay();
1015 	pid_idr_init();
1016 	anon_vma_init();
1017 #ifdef CONFIG_X86
1018 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1019 		efi_enter_virtual_mode();
1020 #endif
1021 	thread_stack_cache_init();
1022 	cred_init();
1023 	fork_init();
1024 	proc_caches_init();
1025 	uts_ns_init();
1026 	buffer_init();
1027 	key_init();
1028 	security_init();
1029 	dbg_late_init();
1030 	vfs_caches_init();
1031 	pagecache_init();
1032 	signals_init();
1033 	seq_file_init();
1034 	proc_root_init();
1035 	nsfs_init();
1036 	cpuset_init();
1037 	cgroup_init();
1038 	taskstats_init_early();
1039 	delayacct_init();
1040 
1041 	poking_init();
1042 	check_bugs();
1043 
1044 	acpi_subsystem_init();
1045 	arch_post_acpi_subsys_init();
1046 	sfi_init_late();
1047 	kcsan_init();
1048 
1049 	/* Do the rest non-__init'ed, we're now alive */
1050 	arch_call_rest_init();
1051 
1052 	prevent_tail_call_optimization();
1053 }
1054 
1055 /* Call all constructor functions linked into the kernel. */
1056 static void __init do_ctors(void)
1057 {
1058 #ifdef CONFIG_CONSTRUCTORS
1059 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1060 
1061 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1062 		(*fn)();
1063 #endif
1064 }
1065 
1066 #ifdef CONFIG_KALLSYMS
1067 struct blacklist_entry {
1068 	struct list_head next;
1069 	char *buf;
1070 };
1071 
1072 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1073 
1074 static int __init initcall_blacklist(char *str)
1075 {
1076 	char *str_entry;
1077 	struct blacklist_entry *entry;
1078 
1079 	/* str argument is a comma-separated list of functions */
1080 	do {
1081 		str_entry = strsep(&str, ",");
1082 		if (str_entry) {
1083 			pr_debug("blacklisting initcall %s\n", str_entry);
1084 			entry = memblock_alloc(sizeof(*entry),
1085 					       SMP_CACHE_BYTES);
1086 			if (!entry)
1087 				panic("%s: Failed to allocate %zu bytes\n",
1088 				      __func__, sizeof(*entry));
1089 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1090 						    SMP_CACHE_BYTES);
1091 			if (!entry->buf)
1092 				panic("%s: Failed to allocate %zu bytes\n",
1093 				      __func__, strlen(str_entry) + 1);
1094 			strcpy(entry->buf, str_entry);
1095 			list_add(&entry->next, &blacklisted_initcalls);
1096 		}
1097 	} while (str_entry);
1098 
1099 	return 0;
1100 }
1101 
1102 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1103 {
1104 	struct blacklist_entry *entry;
1105 	char fn_name[KSYM_SYMBOL_LEN];
1106 	unsigned long addr;
1107 
1108 	if (list_empty(&blacklisted_initcalls))
1109 		return false;
1110 
1111 	addr = (unsigned long) dereference_function_descriptor(fn);
1112 	sprint_symbol_no_offset(fn_name, addr);
1113 
1114 	/*
1115 	 * fn will be "function_name [module_name]" where [module_name] is not
1116 	 * displayed for built-in init functions.  Strip off the [module_name].
1117 	 */
1118 	strreplace(fn_name, ' ', '\0');
1119 
1120 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1121 		if (!strcmp(fn_name, entry->buf)) {
1122 			pr_debug("initcall %s blacklisted\n", fn_name);
1123 			return true;
1124 		}
1125 	}
1126 
1127 	return false;
1128 }
1129 #else
1130 static int __init initcall_blacklist(char *str)
1131 {
1132 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1133 	return 0;
1134 }
1135 
1136 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1137 {
1138 	return false;
1139 }
1140 #endif
1141 __setup("initcall_blacklist=", initcall_blacklist);
1142 
1143 static __init_or_module void
1144 trace_initcall_start_cb(void *data, initcall_t fn)
1145 {
1146 	ktime_t *calltime = (ktime_t *)data;
1147 
1148 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1149 	*calltime = ktime_get();
1150 }
1151 
1152 static __init_or_module void
1153 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1154 {
1155 	ktime_t *calltime = (ktime_t *)data;
1156 	ktime_t delta, rettime;
1157 	unsigned long long duration;
1158 
1159 	rettime = ktime_get();
1160 	delta = ktime_sub(rettime, *calltime);
1161 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1162 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1163 		 fn, ret, duration);
1164 }
1165 
1166 static ktime_t initcall_calltime;
1167 
1168 #ifdef TRACEPOINTS_ENABLED
1169 static void __init initcall_debug_enable(void)
1170 {
1171 	int ret;
1172 
1173 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1174 					    &initcall_calltime);
1175 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1176 					      &initcall_calltime);
1177 	WARN(ret, "Failed to register initcall tracepoints\n");
1178 }
1179 # define do_trace_initcall_start	trace_initcall_start
1180 # define do_trace_initcall_finish	trace_initcall_finish
1181 #else
1182 static inline void do_trace_initcall_start(initcall_t fn)
1183 {
1184 	if (!initcall_debug)
1185 		return;
1186 	trace_initcall_start_cb(&initcall_calltime, fn);
1187 }
1188 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1189 {
1190 	if (!initcall_debug)
1191 		return;
1192 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1193 }
1194 #endif /* !TRACEPOINTS_ENABLED */
1195 
1196 int __init_or_module do_one_initcall(initcall_t fn)
1197 {
1198 	int count = preempt_count();
1199 	char msgbuf[64];
1200 	int ret;
1201 
1202 	if (initcall_blacklisted(fn))
1203 		return -EPERM;
1204 
1205 	do_trace_initcall_start(fn);
1206 	ret = fn();
1207 	do_trace_initcall_finish(fn, ret);
1208 
1209 	msgbuf[0] = 0;
1210 
1211 	if (preempt_count() != count) {
1212 		sprintf(msgbuf, "preemption imbalance ");
1213 		preempt_count_set(count);
1214 	}
1215 	if (irqs_disabled()) {
1216 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1217 		local_irq_enable();
1218 	}
1219 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1220 
1221 	add_latent_entropy();
1222 	return ret;
1223 }
1224 
1225 
1226 extern initcall_entry_t __initcall_start[];
1227 extern initcall_entry_t __initcall0_start[];
1228 extern initcall_entry_t __initcall1_start[];
1229 extern initcall_entry_t __initcall2_start[];
1230 extern initcall_entry_t __initcall3_start[];
1231 extern initcall_entry_t __initcall4_start[];
1232 extern initcall_entry_t __initcall5_start[];
1233 extern initcall_entry_t __initcall6_start[];
1234 extern initcall_entry_t __initcall7_start[];
1235 extern initcall_entry_t __initcall_end[];
1236 
1237 static initcall_entry_t *initcall_levels[] __initdata = {
1238 	__initcall0_start,
1239 	__initcall1_start,
1240 	__initcall2_start,
1241 	__initcall3_start,
1242 	__initcall4_start,
1243 	__initcall5_start,
1244 	__initcall6_start,
1245 	__initcall7_start,
1246 	__initcall_end,
1247 };
1248 
1249 /* Keep these in sync with initcalls in include/linux/init.h */
1250 static const char *initcall_level_names[] __initdata = {
1251 	"pure",
1252 	"core",
1253 	"postcore",
1254 	"arch",
1255 	"subsys",
1256 	"fs",
1257 	"device",
1258 	"late",
1259 };
1260 
1261 static int __init ignore_unknown_bootoption(char *param, char *val,
1262 			       const char *unused, void *arg)
1263 {
1264 	return 0;
1265 }
1266 
1267 static void __init do_initcall_level(int level, char *command_line)
1268 {
1269 	initcall_entry_t *fn;
1270 
1271 	parse_args(initcall_level_names[level],
1272 		   command_line, __start___param,
1273 		   __stop___param - __start___param,
1274 		   level, level,
1275 		   NULL, ignore_unknown_bootoption);
1276 
1277 	trace_initcall_level(initcall_level_names[level]);
1278 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1279 		do_one_initcall(initcall_from_entry(fn));
1280 }
1281 
1282 static void __init do_initcalls(void)
1283 {
1284 	int level;
1285 	size_t len = strlen(saved_command_line) + 1;
1286 	char *command_line;
1287 
1288 	command_line = kzalloc(len, GFP_KERNEL);
1289 	if (!command_line)
1290 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1291 
1292 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1293 		/* Parser modifies command_line, restore it each time */
1294 		strcpy(command_line, saved_command_line);
1295 		do_initcall_level(level, command_line);
1296 	}
1297 
1298 	kfree(command_line);
1299 }
1300 
1301 /*
1302  * Ok, the machine is now initialized. None of the devices
1303  * have been touched yet, but the CPU subsystem is up and
1304  * running, and memory and process management works.
1305  *
1306  * Now we can finally start doing some real work..
1307  */
1308 static void __init do_basic_setup(void)
1309 {
1310 	cpuset_init_smp();
1311 	driver_init();
1312 	init_irq_proc();
1313 	do_ctors();
1314 	usermodehelper_enable();
1315 	do_initcalls();
1316 }
1317 
1318 static void __init do_pre_smp_initcalls(void)
1319 {
1320 	initcall_entry_t *fn;
1321 
1322 	trace_initcall_level("early");
1323 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1324 		do_one_initcall(initcall_from_entry(fn));
1325 }
1326 
1327 static int run_init_process(const char *init_filename)
1328 {
1329 	const char *const *p;
1330 
1331 	argv_init[0] = init_filename;
1332 	pr_info("Run %s as init process\n", init_filename);
1333 	pr_debug("  with arguments:\n");
1334 	for (p = argv_init; *p; p++)
1335 		pr_debug("    %s\n", *p);
1336 	pr_debug("  with environment:\n");
1337 	for (p = envp_init; *p; p++)
1338 		pr_debug("    %s\n", *p);
1339 	return kernel_execve(init_filename, argv_init, envp_init);
1340 }
1341 
1342 static int try_to_run_init_process(const char *init_filename)
1343 {
1344 	int ret;
1345 
1346 	ret = run_init_process(init_filename);
1347 
1348 	if (ret && ret != -ENOENT) {
1349 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1350 		       init_filename, ret);
1351 	}
1352 
1353 	return ret;
1354 }
1355 
1356 static noinline void __init kernel_init_freeable(void);
1357 
1358 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1359 bool rodata_enabled __ro_after_init = true;
1360 static int __init set_debug_rodata(char *str)
1361 {
1362 	return strtobool(str, &rodata_enabled);
1363 }
1364 __setup("rodata=", set_debug_rodata);
1365 #endif
1366 
1367 #ifdef CONFIG_STRICT_KERNEL_RWX
1368 static void mark_readonly(void)
1369 {
1370 	if (rodata_enabled) {
1371 		/*
1372 		 * load_module() results in W+X mappings, which are cleaned
1373 		 * up with call_rcu().  Let's make sure that queued work is
1374 		 * flushed so that we don't hit false positives looking for
1375 		 * insecure pages which are W+X.
1376 		 */
1377 		rcu_barrier();
1378 		mark_rodata_ro();
1379 		rodata_test();
1380 	} else
1381 		pr_info("Kernel memory protection disabled.\n");
1382 }
1383 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1384 static inline void mark_readonly(void)
1385 {
1386 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1387 }
1388 #else
1389 static inline void mark_readonly(void)
1390 {
1391 	pr_warn("This architecture does not have kernel memory protection.\n");
1392 }
1393 #endif
1394 
1395 void __weak free_initmem(void)
1396 {
1397 	free_initmem_default(POISON_FREE_INITMEM);
1398 }
1399 
1400 static int __ref kernel_init(void *unused)
1401 {
1402 	int ret;
1403 
1404 	kernel_init_freeable();
1405 	/* need to finish all async __init code before freeing the memory */
1406 	async_synchronize_full();
1407 	ftrace_free_init_mem();
1408 	free_initmem();
1409 	mark_readonly();
1410 
1411 	/*
1412 	 * Kernel mappings are now finalized - update the userspace page-table
1413 	 * to finalize PTI.
1414 	 */
1415 	pti_finalize();
1416 
1417 	system_state = SYSTEM_RUNNING;
1418 	numa_default_policy();
1419 
1420 	rcu_end_inkernel_boot();
1421 
1422 	do_sysctl_args();
1423 
1424 	if (ramdisk_execute_command) {
1425 		ret = run_init_process(ramdisk_execute_command);
1426 		if (!ret)
1427 			return 0;
1428 		pr_err("Failed to execute %s (error %d)\n",
1429 		       ramdisk_execute_command, ret);
1430 	}
1431 
1432 	/*
1433 	 * We try each of these until one succeeds.
1434 	 *
1435 	 * The Bourne shell can be used instead of init if we are
1436 	 * trying to recover a really broken machine.
1437 	 */
1438 	if (execute_command) {
1439 		ret = run_init_process(execute_command);
1440 		if (!ret)
1441 			return 0;
1442 		panic("Requested init %s failed (error %d).",
1443 		      execute_command, ret);
1444 	}
1445 
1446 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1447 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1448 		if (ret)
1449 			pr_err("Default init %s failed (error %d)\n",
1450 			       CONFIG_DEFAULT_INIT, ret);
1451 		else
1452 			return 0;
1453 	}
1454 
1455 	if (!try_to_run_init_process("/sbin/init") ||
1456 	    !try_to_run_init_process("/etc/init") ||
1457 	    !try_to_run_init_process("/bin/init") ||
1458 	    !try_to_run_init_process("/bin/sh"))
1459 		return 0;
1460 
1461 	panic("No working init found.  Try passing init= option to kernel. "
1462 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1463 }
1464 
1465 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1466 void __init console_on_rootfs(void)
1467 {
1468 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1469 
1470 	if (IS_ERR(file)) {
1471 		pr_err("Warning: unable to open an initial console.\n");
1472 		return;
1473 	}
1474 	init_dup(file);
1475 	init_dup(file);
1476 	init_dup(file);
1477 	fput(file);
1478 }
1479 
1480 static noinline void __init kernel_init_freeable(void)
1481 {
1482 	/*
1483 	 * Wait until kthreadd is all set-up.
1484 	 */
1485 	wait_for_completion(&kthreadd_done);
1486 
1487 	/* Now the scheduler is fully set up and can do blocking allocations */
1488 	gfp_allowed_mask = __GFP_BITS_MASK;
1489 
1490 	/*
1491 	 * init can allocate pages on any node
1492 	 */
1493 	set_mems_allowed(node_states[N_MEMORY]);
1494 
1495 	cad_pid = task_pid(current);
1496 
1497 	smp_prepare_cpus(setup_max_cpus);
1498 
1499 	workqueue_init();
1500 
1501 	init_mm_internals();
1502 
1503 	do_pre_smp_initcalls();
1504 	lockup_detector_init();
1505 
1506 	smp_init();
1507 	sched_init_smp();
1508 
1509 	padata_init();
1510 	page_alloc_init_late();
1511 	/* Initialize page ext after all struct pages are initialized. */
1512 	page_ext_init();
1513 
1514 	do_basic_setup();
1515 
1516 	kunit_run_all_tests();
1517 
1518 	console_on_rootfs();
1519 
1520 	/*
1521 	 * check if there is an early userspace init.  If yes, let it do all
1522 	 * the work
1523 	 */
1524 	if (init_eaccess(ramdisk_execute_command) != 0) {
1525 		ramdisk_execute_command = NULL;
1526 		prepare_namespace();
1527 	}
1528 
1529 	/*
1530 	 * Ok, we have completed the initial bootup, and
1531 	 * we're essentially up and running. Get rid of the
1532 	 * initmem segments and start the user-mode stuff..
1533 	 *
1534 	 * rootfs is available now, try loading the public keys
1535 	 * and default modules
1536 	 */
1537 
1538 	integrity_load_keys();
1539 }
1540