xref: /openbmc/linux/init/main.c (revision 5c05bdd9)
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/kprobes.h>
37 #include <linux/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/cgroup.h>
54 #include <linux/efi.h>
55 #include <linux/tick.h>
56 #include <linux/sched/isolation.h>
57 #include <linux/interrupt.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/delayacct.h>
60 #include <linux/unistd.h>
61 #include <linux/utsname.h>
62 #include <linux/rmap.h>
63 #include <linux/mempolicy.h>
64 #include <linux/key.h>
65 #include <linux/debug_locks.h>
66 #include <linux/debugobjects.h>
67 #include <linux/lockdep.h>
68 #include <linux/kmemleak.h>
69 #include <linux/padata.h>
70 #include <linux/pid_namespace.h>
71 #include <linux/device/driver.h>
72 #include <linux/kthread.h>
73 #include <linux/sched.h>
74 #include <linux/sched/init.h>
75 #include <linux/signal.h>
76 #include <linux/idr.h>
77 #include <linux/kgdb.h>
78 #include <linux/ftrace.h>
79 #include <linux/async.h>
80 #include <linux/shmem_fs.h>
81 #include <linux/slab.h>
82 #include <linux/perf_event.h>
83 #include <linux/ptrace.h>
84 #include <linux/pti.h>
85 #include <linux/blkdev.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/moduleloader.h>
92 #include <linux/list.h>
93 #include <linux/integrity.h>
94 #include <linux/proc_ns.h>
95 #include <linux/io.h>
96 #include <linux/cache.h>
97 #include <linux/rodata_test.h>
98 #include <linux/jump_label.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #include <linux/stackdepot.h>
102 #include <linux/randomize_kstack.h>
103 #include <net/net_namespace.h>
104 
105 #include <asm/io.h>
106 #include <asm/setup.h>
107 #include <asm/sections.h>
108 #include <asm/cacheflush.h>
109 
110 #define CREATE_TRACE_POINTS
111 #include <trace/events/initcall.h>
112 
113 #include <kunit/test.h>
114 
115 static int kernel_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 /* Default late time init is NULL. archs can override this later. */
136 void (*__initdata late_time_init)(void);
137 
138 /* Untouched command line saved by arch-specific code. */
139 char __initdata boot_command_line[COMMAND_LINE_SIZE];
140 /* Untouched saved command line (eg. for /proc) */
141 char *saved_command_line __ro_after_init;
142 unsigned int saved_command_line_len __ro_after_init;
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 size_t initargs_offs;
154 #else
155 # define bootconfig_found false
156 # define initargs_offs 0
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 static bool __init obsolete_checksetup(char *line)
194 {
195 	const struct obs_kernel_param *p;
196 	bool had_early_param = false;
197 
198 	p = __setup_start;
199 	do {
200 		int n = strlen(p->str);
201 		if (parameqn(line, p->str, n)) {
202 			if (p->early) {
203 				/* Already done in parse_early_param?
204 				 * (Needs exact match on param part).
205 				 * Keep iterating, as we can have early
206 				 * params and __setups of same names 8( */
207 				if (line[n] == '\0' || line[n] == '=')
208 					had_early_param = true;
209 			} else if (!p->setup_func) {
210 				pr_warn("Parameter %s is obsolete, ignored\n",
211 					p->str);
212 				return true;
213 			} else if (p->setup_func(line + n))
214 				return true;
215 		}
216 		p++;
217 	} while (p < __setup_end);
218 
219 	return had_early_param;
220 }
221 
222 /*
223  * This should be approx 2 Bo*oMips to start (note initial shift), and will
224  * still work even if initially too large, it will just take slightly longer
225  */
226 unsigned long loops_per_jiffy = (1<<12);
227 EXPORT_SYMBOL(loops_per_jiffy);
228 
229 static int __init debug_kernel(char *str)
230 {
231 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
232 	return 0;
233 }
234 
235 static int __init quiet_kernel(char *str)
236 {
237 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
238 	return 0;
239 }
240 
241 early_param("debug", debug_kernel);
242 early_param("quiet", quiet_kernel);
243 
244 static int __init loglevel(char *str)
245 {
246 	int newlevel;
247 
248 	/*
249 	 * Only update loglevel value when a correct setting was passed,
250 	 * to prevent blind crashes (when loglevel being set to 0) that
251 	 * are quite hard to debug
252 	 */
253 	if (get_option(&str, &newlevel)) {
254 		console_loglevel = newlevel;
255 		return 0;
256 	}
257 
258 	return -EINVAL;
259 }
260 
261 early_param("loglevel", loglevel);
262 
263 #ifdef CONFIG_BLK_DEV_INITRD
264 static void * __init get_boot_config_from_initrd(size_t *_size)
265 {
266 	u32 size, csum;
267 	char *data;
268 	u32 *hdr;
269 	int i;
270 
271 	if (!initrd_end)
272 		return NULL;
273 
274 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
275 	/*
276 	 * Since Grub may align the size of initrd to 4, we must
277 	 * check the preceding 3 bytes as well.
278 	 */
279 	for (i = 0; i < 4; i++) {
280 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
281 			goto found;
282 		data--;
283 	}
284 	return NULL;
285 
286 found:
287 	hdr = (u32 *)(data - 8);
288 	size = le32_to_cpu(hdr[0]);
289 	csum = le32_to_cpu(hdr[1]);
290 
291 	data = ((void *)hdr) - size;
292 	if ((unsigned long)data < initrd_start) {
293 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
294 			size, initrd_end - initrd_start);
295 		return NULL;
296 	}
297 
298 	if (xbc_calc_checksum(data, size) != csum) {
299 		pr_err("bootconfig checksum failed\n");
300 		return NULL;
301 	}
302 
303 	/* Remove bootconfig from initramfs/initrd */
304 	initrd_end = (unsigned long)data;
305 	if (_size)
306 		*_size = size;
307 
308 	return data;
309 }
310 #else
311 static void * __init get_boot_config_from_initrd(size_t *_size)
312 {
313 	return NULL;
314 }
315 #endif
316 
317 #ifdef CONFIG_BOOT_CONFIG
318 
319 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
320 
321 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
322 
323 static int __init xbc_snprint_cmdline(char *buf, size_t size,
324 				      struct xbc_node *root)
325 {
326 	struct xbc_node *knode, *vnode;
327 	char *end = buf + size;
328 	const char *val;
329 	int ret;
330 
331 	xbc_node_for_each_key_value(root, knode, val) {
332 		ret = xbc_node_compose_key_after(root, knode,
333 					xbc_namebuf, XBC_KEYLEN_MAX);
334 		if (ret < 0)
335 			return ret;
336 
337 		vnode = xbc_node_get_child(knode);
338 		if (!vnode) {
339 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
340 			if (ret < 0)
341 				return ret;
342 			buf += ret;
343 			continue;
344 		}
345 		xbc_array_for_each_value(vnode, val) {
346 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
347 				       xbc_namebuf, val);
348 			if (ret < 0)
349 				return ret;
350 			buf += ret;
351 		}
352 	}
353 
354 	return buf - (end - size);
355 }
356 #undef rest
357 
358 /* Make an extra command line under given key word */
359 static char * __init xbc_make_cmdline(const char *key)
360 {
361 	struct xbc_node *root;
362 	char *new_cmdline;
363 	int ret, len = 0;
364 
365 	root = xbc_find_node(key);
366 	if (!root)
367 		return NULL;
368 
369 	/* Count required buffer size */
370 	len = xbc_snprint_cmdline(NULL, 0, root);
371 	if (len <= 0)
372 		return NULL;
373 
374 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
375 	if (!new_cmdline) {
376 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
377 		return NULL;
378 	}
379 
380 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
381 	if (ret < 0 || ret > len) {
382 		pr_err("Failed to print extra kernel cmdline.\n");
383 		memblock_free(new_cmdline, len + 1);
384 		return NULL;
385 	}
386 
387 	return new_cmdline;
388 }
389 
390 static int __init bootconfig_params(char *param, char *val,
391 				    const char *unused, void *arg)
392 {
393 	if (strcmp(param, "bootconfig") == 0) {
394 		bootconfig_found = true;
395 	}
396 	return 0;
397 }
398 
399 static int __init warn_bootconfig(char *str)
400 {
401 	/* The 'bootconfig' has been handled by bootconfig_params(). */
402 	return 0;
403 }
404 
405 static void __init setup_boot_config(void)
406 {
407 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
408 	const char *msg, *data;
409 	int pos, ret;
410 	size_t size;
411 	char *err;
412 
413 	/* Cut out the bootconfig data even if we have no bootconfig option */
414 	data = get_boot_config_from_initrd(&size);
415 	/* If there is no bootconfig in initrd, try embedded one. */
416 	if (!data)
417 		data = xbc_get_embedded_bootconfig(&size);
418 
419 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
420 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
421 			 bootconfig_params);
422 
423 	if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
424 		return;
425 
426 	/* parse_args() stops at the next param of '--' and returns an address */
427 	if (err)
428 		initargs_offs = err - tmp_cmdline;
429 
430 	if (!data) {
431 		/* If user intended to use bootconfig, show an error level message */
432 		if (bootconfig_found)
433 			pr_err("'bootconfig' found on command line, but no bootconfig found\n");
434 		else
435 			pr_info("No bootconfig data provided, so skipping bootconfig");
436 		return;
437 	}
438 
439 	if (size >= XBC_DATA_MAX) {
440 		pr_err("bootconfig size %ld greater than max size %d\n",
441 			(long)size, XBC_DATA_MAX);
442 		return;
443 	}
444 
445 	ret = xbc_init(data, size, &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 		xbc_get_info(&ret, NULL);
454 		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
455 		/* keys starting with "kernel." are passed via cmdline */
456 		extra_command_line = xbc_make_cmdline("kernel");
457 		/* Also, "init." keys are init arguments */
458 		extra_init_args = xbc_make_cmdline("init");
459 	}
460 	return;
461 }
462 
463 static void __init exit_boot_config(void)
464 {
465 	xbc_exit();
466 }
467 
468 #else	/* !CONFIG_BOOT_CONFIG */
469 
470 static void __init setup_boot_config(void)
471 {
472 	/* Remove bootconfig data from initrd */
473 	get_boot_config_from_initrd(NULL);
474 }
475 
476 static int __init warn_bootconfig(char *str)
477 {
478 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
479 	return 0;
480 }
481 
482 #define exit_boot_config()	do {} while (0)
483 
484 #endif	/* CONFIG_BOOT_CONFIG */
485 
486 early_param("bootconfig", warn_bootconfig);
487 
488 /* Change NUL term back to "=", to make "param" the whole string. */
489 static void __init repair_env_string(char *param, char *val)
490 {
491 	if (val) {
492 		/* param=val or param="val"? */
493 		if (val == param+strlen(param)+1)
494 			val[-1] = '=';
495 		else if (val == param+strlen(param)+2) {
496 			val[-2] = '=';
497 			memmove(val-1, val, strlen(val)+1);
498 		} else
499 			BUG();
500 	}
501 }
502 
503 /* Anything after -- gets handed straight to init. */
504 static int __init set_init_arg(char *param, char *val,
505 			       const char *unused, void *arg)
506 {
507 	unsigned int i;
508 
509 	if (panic_later)
510 		return 0;
511 
512 	repair_env_string(param, val);
513 
514 	for (i = 0; argv_init[i]; i++) {
515 		if (i == MAX_INIT_ARGS) {
516 			panic_later = "init";
517 			panic_param = param;
518 			return 0;
519 		}
520 	}
521 	argv_init[i] = param;
522 	return 0;
523 }
524 
525 /*
526  * Unknown boot options get handed to init, unless they look like
527  * unused parameters (modprobe will find them in /proc/cmdline).
528  */
529 static int __init unknown_bootoption(char *param, char *val,
530 				     const char *unused, void *arg)
531 {
532 	size_t len = strlen(param);
533 
534 	/* Handle params aliased to sysctls */
535 	if (sysctl_is_alias(param))
536 		return 0;
537 
538 	repair_env_string(param, val);
539 
540 	/* Handle obsolete-style parameters */
541 	if (obsolete_checksetup(param))
542 		return 0;
543 
544 	/* Unused module parameter. */
545 	if (strnchr(param, len, '.'))
546 		return 0;
547 
548 	if (panic_later)
549 		return 0;
550 
551 	if (val) {
552 		/* Environment option */
553 		unsigned int i;
554 		for (i = 0; envp_init[i]; i++) {
555 			if (i == MAX_INIT_ENVS) {
556 				panic_later = "env";
557 				panic_param = param;
558 			}
559 			if (!strncmp(param, envp_init[i], len+1))
560 				break;
561 		}
562 		envp_init[i] = param;
563 	} else {
564 		/* Command line option */
565 		unsigned int i;
566 		for (i = 0; argv_init[i]; i++) {
567 			if (i == MAX_INIT_ARGS) {
568 				panic_later = "init";
569 				panic_param = param;
570 			}
571 		}
572 		argv_init[i] = param;
573 	}
574 	return 0;
575 }
576 
577 static int __init init_setup(char *str)
578 {
579 	unsigned int i;
580 
581 	execute_command = str;
582 	/*
583 	 * In case LILO is going to boot us with default command line,
584 	 * it prepends "auto" before the whole cmdline which makes
585 	 * the shell think it should execute a script with such name.
586 	 * So we ignore all arguments entered _before_ init=... [MJ]
587 	 */
588 	for (i = 1; i < MAX_INIT_ARGS; i++)
589 		argv_init[i] = NULL;
590 	return 1;
591 }
592 __setup("init=", init_setup);
593 
594 static int __init rdinit_setup(char *str)
595 {
596 	unsigned int i;
597 
598 	ramdisk_execute_command = str;
599 	/* See "auto" comment in init_setup */
600 	for (i = 1; i < MAX_INIT_ARGS; i++)
601 		argv_init[i] = NULL;
602 	return 1;
603 }
604 __setup("rdinit=", rdinit_setup);
605 
606 #ifndef CONFIG_SMP
607 static const unsigned int setup_max_cpus = NR_CPUS;
608 static inline void setup_nr_cpu_ids(void) { }
609 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
610 #endif
611 
612 /*
613  * We need to store the untouched command line for future reference.
614  * We also need to store the touched command line since the parameter
615  * parsing is performed in place, and we should allow a component to
616  * store reference of name/value for future reference.
617  */
618 static void __init setup_command_line(char *command_line)
619 {
620 	size_t len, xlen = 0, ilen = 0;
621 
622 	if (extra_command_line)
623 		xlen = strlen(extra_command_line);
624 	if (extra_init_args)
625 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
626 
627 	len = xlen + strlen(boot_command_line) + 1;
628 
629 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
630 	if (!saved_command_line)
631 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
632 
633 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
634 	if (!static_command_line)
635 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
636 
637 	if (xlen) {
638 		/*
639 		 * We have to put extra_command_line before boot command
640 		 * lines because there could be dashes (separator of init
641 		 * command line) in the command lines.
642 		 */
643 		strcpy(saved_command_line, extra_command_line);
644 		strcpy(static_command_line, extra_command_line);
645 	}
646 	strcpy(saved_command_line + xlen, boot_command_line);
647 	strcpy(static_command_line + xlen, command_line);
648 
649 	if (ilen) {
650 		/*
651 		 * Append supplemental init boot args to saved_command_line
652 		 * so that user can check what command line options passed
653 		 * to init.
654 		 * The order should always be
655 		 * " -- "[bootconfig init-param][cmdline init-param]
656 		 */
657 		if (initargs_offs) {
658 			len = xlen + initargs_offs;
659 			strcpy(saved_command_line + len, extra_init_args);
660 			len += ilen - 4;	/* strlen(extra_init_args) */
661 			strcpy(saved_command_line + len,
662 				boot_command_line + initargs_offs - 1);
663 		} else {
664 			len = strlen(saved_command_line);
665 			strcpy(saved_command_line + len, " -- ");
666 			len += 4;
667 			strcpy(saved_command_line + len, extra_init_args);
668 		}
669 	}
670 
671 	saved_command_line_len = strlen(saved_command_line);
672 }
673 
674 /*
675  * We need to finalize in a non-__init function or else race conditions
676  * between the root thread and the init thread may cause start_kernel to
677  * be reaped by free_initmem before the root thread has proceeded to
678  * cpu_idle.
679  *
680  * gcc-3.4 accidentally inlines this function, so use noinline.
681  */
682 
683 static __initdata DECLARE_COMPLETION(kthreadd_done);
684 
685 noinline void __ref __noreturn rest_init(void)
686 {
687 	struct task_struct *tsk;
688 	int pid;
689 
690 	rcu_scheduler_starting();
691 	/*
692 	 * We need to spawn init first so that it obtains pid 1, however
693 	 * the init task will end up wanting to create kthreads, which, if
694 	 * we schedule it before we create kthreadd, will OOPS.
695 	 */
696 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
697 	/*
698 	 * Pin init on the boot CPU. Task migration is not properly working
699 	 * until sched_init_smp() has been run. It will set the allowed
700 	 * CPUs for init to the non isolated CPUs.
701 	 */
702 	rcu_read_lock();
703 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
704 	tsk->flags |= PF_NO_SETAFFINITY;
705 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
706 	rcu_read_unlock();
707 
708 	numa_default_policy();
709 	pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
710 	rcu_read_lock();
711 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
712 	rcu_read_unlock();
713 
714 	/*
715 	 * Enable might_sleep() and smp_processor_id() checks.
716 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
717 	 * kernel_thread() would trigger might_sleep() splats. With
718 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
719 	 * already, but it's stuck on the kthreadd_done completion.
720 	 */
721 	system_state = SYSTEM_SCHEDULING;
722 
723 	complete(&kthreadd_done);
724 
725 	/*
726 	 * The boot idle thread must execute schedule()
727 	 * at least once to get things moving:
728 	 */
729 	schedule_preempt_disabled();
730 	/* Call into cpu_idle with preempt disabled */
731 	cpu_startup_entry(CPUHP_ONLINE);
732 }
733 
734 /* Check for early params. */
735 static int __init do_early_param(char *param, char *val,
736 				 const char *unused, void *arg)
737 {
738 	const struct obs_kernel_param *p;
739 
740 	for (p = __setup_start; p < __setup_end; p++) {
741 		if ((p->early && parameq(param, p->str)) ||
742 		    (strcmp(param, "console") == 0 &&
743 		     strcmp(p->str, "earlycon") == 0)
744 		) {
745 			if (p->setup_func(val) != 0)
746 				pr_warn("Malformed early option '%s'\n", param);
747 		}
748 	}
749 	/* We accept everything at this stage. */
750 	return 0;
751 }
752 
753 void __init parse_early_options(char *cmdline)
754 {
755 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
756 		   do_early_param);
757 }
758 
759 /* Arch code calls this early on, or if not, just before other parsing. */
760 void __init parse_early_param(void)
761 {
762 	static int done __initdata;
763 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
764 
765 	if (done)
766 		return;
767 
768 	/* All fall through to do_early_param. */
769 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
770 	parse_early_options(tmp_cmdline);
771 	done = 1;
772 }
773 
774 void __init __weak arch_post_acpi_subsys_init(void) { }
775 
776 void __init __weak smp_setup_processor_id(void)
777 {
778 }
779 
780 # if THREAD_SIZE >= PAGE_SIZE
781 void __init __weak thread_stack_cache_init(void)
782 {
783 }
784 #endif
785 
786 void __init __weak poking_init(void) { }
787 
788 void __init __weak pgtable_cache_init(void) { }
789 
790 void __init __weak trap_init(void) { }
791 
792 bool initcall_debug;
793 core_param(initcall_debug, initcall_debug, bool, 0644);
794 
795 #ifdef TRACEPOINTS_ENABLED
796 static void __init initcall_debug_enable(void);
797 #else
798 static inline void initcall_debug_enable(void)
799 {
800 }
801 #endif
802 
803 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
804 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
805 			   randomize_kstack_offset);
806 DEFINE_PER_CPU(u32, kstack_offset);
807 
808 static int __init early_randomize_kstack_offset(char *buf)
809 {
810 	int ret;
811 	bool bool_result;
812 
813 	ret = kstrtobool(buf, &bool_result);
814 	if (ret)
815 		return ret;
816 
817 	if (bool_result)
818 		static_branch_enable(&randomize_kstack_offset);
819 	else
820 		static_branch_disable(&randomize_kstack_offset);
821 	return 0;
822 }
823 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
824 #endif
825 
826 void __init __weak __noreturn arch_call_rest_init(void)
827 {
828 	rest_init();
829 }
830 
831 static void __init print_unknown_bootoptions(void)
832 {
833 	char *unknown_options;
834 	char *end;
835 	const char *const *p;
836 	size_t len;
837 
838 	if (panic_later || (!argv_init[1] && !envp_init[2]))
839 		return;
840 
841 	/*
842 	 * Determine how many options we have to print out, plus a space
843 	 * before each
844 	 */
845 	len = 1; /* null terminator */
846 	for (p = &argv_init[1]; *p; p++) {
847 		len++;
848 		len += strlen(*p);
849 	}
850 	for (p = &envp_init[2]; *p; p++) {
851 		len++;
852 		len += strlen(*p);
853 	}
854 
855 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
856 	if (!unknown_options) {
857 		pr_err("%s: Failed to allocate %zu bytes\n",
858 			__func__, len);
859 		return;
860 	}
861 	end = unknown_options;
862 
863 	for (p = &argv_init[1]; *p; p++)
864 		end += sprintf(end, " %s", *p);
865 	for (p = &envp_init[2]; *p; p++)
866 		end += sprintf(end, " %s", *p);
867 
868 	/* Start at unknown_options[1] to skip the initial space */
869 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
870 		&unknown_options[1]);
871 	memblock_free(unknown_options, len);
872 }
873 
874 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
875 void start_kernel(void)
876 {
877 	char *command_line;
878 	char *after_dashes;
879 
880 	set_task_stack_end_magic(&init_task);
881 	smp_setup_processor_id();
882 	debug_objects_early_init();
883 	init_vmlinux_build_id();
884 
885 	cgroup_init_early();
886 
887 	local_irq_disable();
888 	early_boot_irqs_disabled = true;
889 
890 	/*
891 	 * Interrupts are still disabled. Do necessary setups, then
892 	 * enable them.
893 	 */
894 	boot_cpu_init();
895 	page_address_init();
896 	pr_notice("%s", linux_banner);
897 	early_security_init();
898 	setup_arch(&command_line);
899 	setup_boot_config();
900 	setup_command_line(command_line);
901 	setup_nr_cpu_ids();
902 	setup_per_cpu_areas();
903 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
904 	boot_cpu_hotplug_init();
905 
906 	pr_notice("Kernel command line: %s\n", saved_command_line);
907 	/* parameters may set static keys */
908 	jump_label_init();
909 	parse_early_param();
910 	after_dashes = parse_args("Booting kernel",
911 				  static_command_line, __start___param,
912 				  __stop___param - __start___param,
913 				  -1, -1, NULL, &unknown_bootoption);
914 	print_unknown_bootoptions();
915 	if (!IS_ERR_OR_NULL(after_dashes))
916 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
917 			   NULL, set_init_arg);
918 	if (extra_init_args)
919 		parse_args("Setting extra init args", extra_init_args,
920 			   NULL, 0, -1, -1, NULL, set_init_arg);
921 
922 	/* Architectural and non-timekeeping rng init, before allocator init */
923 	random_init_early(command_line);
924 
925 	/*
926 	 * These use large bootmem allocations and must precede
927 	 * initalization of page allocator
928 	 */
929 	setup_log_buf(0);
930 	vfs_caches_init_early();
931 	sort_main_extable();
932 	trap_init();
933 	mm_core_init();
934 	poking_init();
935 	ftrace_init();
936 
937 	/* trace_printk can be enabled here */
938 	early_trace_init();
939 
940 	/*
941 	 * Set up the scheduler prior starting any interrupts (such as the
942 	 * timer interrupt). Full topology setup happens at smp_init()
943 	 * time - but meanwhile we still have a functioning scheduler.
944 	 */
945 	sched_init();
946 
947 	if (WARN(!irqs_disabled(),
948 		 "Interrupts were enabled *very* early, fixing it\n"))
949 		local_irq_disable();
950 	radix_tree_init();
951 	maple_tree_init();
952 
953 	/*
954 	 * Set up housekeeping before setting up workqueues to allow the unbound
955 	 * workqueue to take non-housekeeping into account.
956 	 */
957 	housekeeping_init();
958 
959 	/*
960 	 * Allow workqueue creation and work item queueing/cancelling
961 	 * early.  Work item execution depends on kthreads and starts after
962 	 * workqueue_init().
963 	 */
964 	workqueue_init_early();
965 
966 	rcu_init();
967 
968 	/* Trace events are available after this */
969 	trace_init();
970 
971 	if (initcall_debug)
972 		initcall_debug_enable();
973 
974 	context_tracking_init();
975 	/* init some links before init_ISA_irqs() */
976 	early_irq_init();
977 	init_IRQ();
978 	tick_init();
979 	rcu_init_nohz();
980 	init_timers();
981 	srcu_init();
982 	hrtimers_init();
983 	softirq_init();
984 	timekeeping_init();
985 	time_init();
986 
987 	/* This must be after timekeeping is initialized */
988 	random_init();
989 
990 	/* These make use of the fully initialized rng */
991 	kfence_init();
992 	boot_init_stack_canary();
993 
994 	perf_event_init();
995 	profile_init();
996 	call_function_init();
997 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
998 
999 	early_boot_irqs_disabled = false;
1000 	local_irq_enable();
1001 
1002 	kmem_cache_init_late();
1003 
1004 	/*
1005 	 * HACK ALERT! This is early. We're enabling the console before
1006 	 * we've done PCI setups etc, and console_init() must be aware of
1007 	 * this. But we do want output early, in case something goes wrong.
1008 	 */
1009 	console_init();
1010 	if (panic_later)
1011 		panic("Too many boot %s vars at `%s'", panic_later,
1012 		      panic_param);
1013 
1014 	lockdep_init();
1015 
1016 	/*
1017 	 * Need to run this when irqs are enabled, because it wants
1018 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1019 	 * too:
1020 	 */
1021 	locking_selftest();
1022 
1023 #ifdef CONFIG_BLK_DEV_INITRD
1024 	if (initrd_start && !initrd_below_start_ok &&
1025 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1026 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1027 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1028 		    min_low_pfn);
1029 		initrd_start = 0;
1030 	}
1031 #endif
1032 	setup_per_cpu_pageset();
1033 	numa_policy_init();
1034 	acpi_early_init();
1035 	if (late_time_init)
1036 		late_time_init();
1037 	sched_clock_init();
1038 	calibrate_delay();
1039 
1040 	arch_cpu_finalize_init();
1041 
1042 	pid_idr_init();
1043 	anon_vma_init();
1044 #ifdef CONFIG_X86
1045 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1046 		efi_enter_virtual_mode();
1047 #endif
1048 	thread_stack_cache_init();
1049 	cred_init();
1050 	fork_init();
1051 	proc_caches_init();
1052 	uts_ns_init();
1053 	key_init();
1054 	security_init();
1055 	dbg_late_init();
1056 	net_ns_init();
1057 	vfs_caches_init();
1058 	pagecache_init();
1059 	signals_init();
1060 	seq_file_init();
1061 	proc_root_init();
1062 	nsfs_init();
1063 	cpuset_init();
1064 	cgroup_init();
1065 	taskstats_init_early();
1066 	delayacct_init();
1067 
1068 	acpi_subsystem_init();
1069 	arch_post_acpi_subsys_init();
1070 	kcsan_init();
1071 
1072 	/* Do the rest non-__init'ed, we're now alive */
1073 	arch_call_rest_init();
1074 
1075 	/*
1076 	 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1077 	 * and older.
1078 	 */
1079 #if !__has_attribute(__no_stack_protector__)
1080 	prevent_tail_call_optimization();
1081 #endif
1082 }
1083 
1084 /* Call all constructor functions linked into the kernel. */
1085 static void __init do_ctors(void)
1086 {
1087 /*
1088  * For UML, the constructors have already been called by the
1089  * normal setup code as it's just a normal ELF binary, so we
1090  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1091  * even on UML for modules.
1092  */
1093 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1094 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1095 
1096 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1097 		(*fn)();
1098 #endif
1099 }
1100 
1101 #ifdef CONFIG_KALLSYMS
1102 struct blacklist_entry {
1103 	struct list_head next;
1104 	char *buf;
1105 };
1106 
1107 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1108 
1109 static int __init initcall_blacklist(char *str)
1110 {
1111 	char *str_entry;
1112 	struct blacklist_entry *entry;
1113 
1114 	/* str argument is a comma-separated list of functions */
1115 	do {
1116 		str_entry = strsep(&str, ",");
1117 		if (str_entry) {
1118 			pr_debug("blacklisting initcall %s\n", str_entry);
1119 			entry = memblock_alloc(sizeof(*entry),
1120 					       SMP_CACHE_BYTES);
1121 			if (!entry)
1122 				panic("%s: Failed to allocate %zu bytes\n",
1123 				      __func__, sizeof(*entry));
1124 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1125 						    SMP_CACHE_BYTES);
1126 			if (!entry->buf)
1127 				panic("%s: Failed to allocate %zu bytes\n",
1128 				      __func__, strlen(str_entry) + 1);
1129 			strcpy(entry->buf, str_entry);
1130 			list_add(&entry->next, &blacklisted_initcalls);
1131 		}
1132 	} while (str_entry);
1133 
1134 	return 1;
1135 }
1136 
1137 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1138 {
1139 	struct blacklist_entry *entry;
1140 	char fn_name[KSYM_SYMBOL_LEN];
1141 	unsigned long addr;
1142 
1143 	if (list_empty(&blacklisted_initcalls))
1144 		return false;
1145 
1146 	addr = (unsigned long) dereference_function_descriptor(fn);
1147 	sprint_symbol_no_offset(fn_name, addr);
1148 
1149 	/*
1150 	 * fn will be "function_name [module_name]" where [module_name] is not
1151 	 * displayed for built-in init functions.  Strip off the [module_name].
1152 	 */
1153 	strreplace(fn_name, ' ', '\0');
1154 
1155 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1156 		if (!strcmp(fn_name, entry->buf)) {
1157 			pr_debug("initcall %s blacklisted\n", fn_name);
1158 			return true;
1159 		}
1160 	}
1161 
1162 	return false;
1163 }
1164 #else
1165 static int __init initcall_blacklist(char *str)
1166 {
1167 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1168 	return 0;
1169 }
1170 
1171 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1172 {
1173 	return false;
1174 }
1175 #endif
1176 __setup("initcall_blacklist=", initcall_blacklist);
1177 
1178 static __init_or_module void
1179 trace_initcall_start_cb(void *data, initcall_t fn)
1180 {
1181 	ktime_t *calltime = data;
1182 
1183 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1184 	*calltime = ktime_get();
1185 }
1186 
1187 static __init_or_module void
1188 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1189 {
1190 	ktime_t rettime, *calltime = data;
1191 
1192 	rettime = ktime_get();
1193 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1194 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1195 }
1196 
1197 static ktime_t initcall_calltime;
1198 
1199 #ifdef TRACEPOINTS_ENABLED
1200 static void __init initcall_debug_enable(void)
1201 {
1202 	int ret;
1203 
1204 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1205 					    &initcall_calltime);
1206 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1207 					      &initcall_calltime);
1208 	WARN(ret, "Failed to register initcall tracepoints\n");
1209 }
1210 # define do_trace_initcall_start	trace_initcall_start
1211 # define do_trace_initcall_finish	trace_initcall_finish
1212 #else
1213 static inline void do_trace_initcall_start(initcall_t fn)
1214 {
1215 	if (!initcall_debug)
1216 		return;
1217 	trace_initcall_start_cb(&initcall_calltime, fn);
1218 }
1219 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1220 {
1221 	if (!initcall_debug)
1222 		return;
1223 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1224 }
1225 #endif /* !TRACEPOINTS_ENABLED */
1226 
1227 int __init_or_module do_one_initcall(initcall_t fn)
1228 {
1229 	int count = preempt_count();
1230 	char msgbuf[64];
1231 	int ret;
1232 
1233 	if (initcall_blacklisted(fn))
1234 		return -EPERM;
1235 
1236 	do_trace_initcall_start(fn);
1237 	ret = fn();
1238 	do_trace_initcall_finish(fn, ret);
1239 
1240 	msgbuf[0] = 0;
1241 
1242 	if (preempt_count() != count) {
1243 		sprintf(msgbuf, "preemption imbalance ");
1244 		preempt_count_set(count);
1245 	}
1246 	if (irqs_disabled()) {
1247 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1248 		local_irq_enable();
1249 	}
1250 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1251 
1252 	add_latent_entropy();
1253 	return ret;
1254 }
1255 
1256 
1257 static initcall_entry_t *initcall_levels[] __initdata = {
1258 	__initcall0_start,
1259 	__initcall1_start,
1260 	__initcall2_start,
1261 	__initcall3_start,
1262 	__initcall4_start,
1263 	__initcall5_start,
1264 	__initcall6_start,
1265 	__initcall7_start,
1266 	__initcall_end,
1267 };
1268 
1269 /* Keep these in sync with initcalls in include/linux/init.h */
1270 static const char *initcall_level_names[] __initdata = {
1271 	"pure",
1272 	"core",
1273 	"postcore",
1274 	"arch",
1275 	"subsys",
1276 	"fs",
1277 	"device",
1278 	"late",
1279 };
1280 
1281 static int __init ignore_unknown_bootoption(char *param, char *val,
1282 			       const char *unused, void *arg)
1283 {
1284 	return 0;
1285 }
1286 
1287 static void __init do_initcall_level(int level, char *command_line)
1288 {
1289 	initcall_entry_t *fn;
1290 
1291 	parse_args(initcall_level_names[level],
1292 		   command_line, __start___param,
1293 		   __stop___param - __start___param,
1294 		   level, level,
1295 		   NULL, ignore_unknown_bootoption);
1296 
1297 	trace_initcall_level(initcall_level_names[level]);
1298 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1299 		do_one_initcall(initcall_from_entry(fn));
1300 }
1301 
1302 static void __init do_initcalls(void)
1303 {
1304 	int level;
1305 	size_t len = saved_command_line_len + 1;
1306 	char *command_line;
1307 
1308 	command_line = kzalloc(len, GFP_KERNEL);
1309 	if (!command_line)
1310 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1311 
1312 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1313 		/* Parser modifies command_line, restore it each time */
1314 		strcpy(command_line, saved_command_line);
1315 		do_initcall_level(level, command_line);
1316 	}
1317 
1318 	kfree(command_line);
1319 }
1320 
1321 /*
1322  * Ok, the machine is now initialized. None of the devices
1323  * have been touched yet, but the CPU subsystem is up and
1324  * running, and memory and process management works.
1325  *
1326  * Now we can finally start doing some real work..
1327  */
1328 static void __init do_basic_setup(void)
1329 {
1330 	cpuset_init_smp();
1331 	driver_init();
1332 	init_irq_proc();
1333 	do_ctors();
1334 	do_initcalls();
1335 }
1336 
1337 static void __init do_pre_smp_initcalls(void)
1338 {
1339 	initcall_entry_t *fn;
1340 
1341 	trace_initcall_level("early");
1342 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1343 		do_one_initcall(initcall_from_entry(fn));
1344 }
1345 
1346 static int run_init_process(const char *init_filename)
1347 {
1348 	const char *const *p;
1349 
1350 	argv_init[0] = init_filename;
1351 	pr_info("Run %s as init process\n", init_filename);
1352 	pr_debug("  with arguments:\n");
1353 	for (p = argv_init; *p; p++)
1354 		pr_debug("    %s\n", *p);
1355 	pr_debug("  with environment:\n");
1356 	for (p = envp_init; *p; p++)
1357 		pr_debug("    %s\n", *p);
1358 	return kernel_execve(init_filename, argv_init, envp_init);
1359 }
1360 
1361 static int try_to_run_init_process(const char *init_filename)
1362 {
1363 	int ret;
1364 
1365 	ret = run_init_process(init_filename);
1366 
1367 	if (ret && ret != -ENOENT) {
1368 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1369 		       init_filename, ret);
1370 	}
1371 
1372 	return ret;
1373 }
1374 
1375 static noinline void __init kernel_init_freeable(void);
1376 
1377 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1378 bool rodata_enabled __ro_after_init = true;
1379 
1380 #ifndef arch_parse_debug_rodata
1381 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1382 #endif
1383 
1384 static int __init set_debug_rodata(char *str)
1385 {
1386 	if (arch_parse_debug_rodata(str))
1387 		return 0;
1388 
1389 	if (str && !strcmp(str, "on"))
1390 		rodata_enabled = true;
1391 	else if (str && !strcmp(str, "off"))
1392 		rodata_enabled = false;
1393 	else
1394 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1395 	return 0;
1396 }
1397 early_param("rodata", set_debug_rodata);
1398 #endif
1399 
1400 #ifdef CONFIG_STRICT_KERNEL_RWX
1401 static void mark_readonly(void)
1402 {
1403 	if (rodata_enabled) {
1404 		/*
1405 		 * load_module() results in W+X mappings, which are cleaned
1406 		 * up with init_free_wq. Let's make sure that queued work is
1407 		 * flushed so that we don't hit false positives looking for
1408 		 * insecure pages which are W+X.
1409 		 */
1410 		flush_module_init_free_work();
1411 		mark_rodata_ro();
1412 		rodata_test();
1413 	} else
1414 		pr_info("Kernel memory protection disabled.\n");
1415 }
1416 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1417 static inline void mark_readonly(void)
1418 {
1419 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1420 }
1421 #else
1422 static inline void mark_readonly(void)
1423 {
1424 	pr_warn("This architecture does not have kernel memory protection.\n");
1425 }
1426 #endif
1427 
1428 void __weak free_initmem(void)
1429 {
1430 	free_initmem_default(POISON_FREE_INITMEM);
1431 }
1432 
1433 static int __ref kernel_init(void *unused)
1434 {
1435 	int ret;
1436 
1437 	/*
1438 	 * Wait until kthreadd is all set-up.
1439 	 */
1440 	wait_for_completion(&kthreadd_done);
1441 
1442 	kernel_init_freeable();
1443 	/* need to finish all async __init code before freeing the memory */
1444 	async_synchronize_full();
1445 
1446 	system_state = SYSTEM_FREEING_INITMEM;
1447 	kprobe_free_init_mem();
1448 	ftrace_free_init_mem();
1449 	kgdb_free_init_mem();
1450 	exit_boot_config();
1451 	free_initmem();
1452 	mark_readonly();
1453 
1454 	/*
1455 	 * Kernel mappings are now finalized - update the userspace page-table
1456 	 * to finalize PTI.
1457 	 */
1458 	pti_finalize();
1459 
1460 	system_state = SYSTEM_RUNNING;
1461 	numa_default_policy();
1462 
1463 	rcu_end_inkernel_boot();
1464 
1465 	do_sysctl_args();
1466 
1467 	if (ramdisk_execute_command) {
1468 		ret = run_init_process(ramdisk_execute_command);
1469 		if (!ret)
1470 			return 0;
1471 		pr_err("Failed to execute %s (error %d)\n",
1472 		       ramdisk_execute_command, ret);
1473 	}
1474 
1475 	/*
1476 	 * We try each of these until one succeeds.
1477 	 *
1478 	 * The Bourne shell can be used instead of init if we are
1479 	 * trying to recover a really broken machine.
1480 	 */
1481 	if (execute_command) {
1482 		ret = run_init_process(execute_command);
1483 		if (!ret)
1484 			return 0;
1485 		panic("Requested init %s failed (error %d).",
1486 		      execute_command, ret);
1487 	}
1488 
1489 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1490 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1491 		if (ret)
1492 			pr_err("Default init %s failed (error %d)\n",
1493 			       CONFIG_DEFAULT_INIT, ret);
1494 		else
1495 			return 0;
1496 	}
1497 
1498 	if (!try_to_run_init_process("/sbin/init") ||
1499 	    !try_to_run_init_process("/etc/init") ||
1500 	    !try_to_run_init_process("/bin/init") ||
1501 	    !try_to_run_init_process("/bin/sh"))
1502 		return 0;
1503 
1504 	panic("No working init found.  Try passing init= option to kernel. "
1505 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1506 }
1507 
1508 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1509 void __init console_on_rootfs(void)
1510 {
1511 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1512 
1513 	if (IS_ERR(file)) {
1514 		pr_err("Warning: unable to open an initial console.\n");
1515 		return;
1516 	}
1517 	init_dup(file);
1518 	init_dup(file);
1519 	init_dup(file);
1520 	fput(file);
1521 }
1522 
1523 static noinline void __init kernel_init_freeable(void)
1524 {
1525 	/* Now the scheduler is fully set up and can do blocking allocations */
1526 	gfp_allowed_mask = __GFP_BITS_MASK;
1527 
1528 	/*
1529 	 * init can allocate pages on any node
1530 	 */
1531 	set_mems_allowed(node_states[N_MEMORY]);
1532 
1533 	cad_pid = get_pid(task_pid(current));
1534 
1535 	smp_prepare_cpus(setup_max_cpus);
1536 
1537 	workqueue_init();
1538 
1539 	init_mm_internals();
1540 
1541 	rcu_init_tasks_generic();
1542 	do_pre_smp_initcalls();
1543 	lockup_detector_init();
1544 
1545 	smp_init();
1546 	sched_init_smp();
1547 
1548 	workqueue_init_topology();
1549 	padata_init();
1550 	page_alloc_init_late();
1551 
1552 	do_basic_setup();
1553 
1554 	kunit_run_all_tests();
1555 
1556 	wait_for_initramfs();
1557 	console_on_rootfs();
1558 
1559 	/*
1560 	 * check if there is an early userspace init.  If yes, let it do all
1561 	 * the work
1562 	 */
1563 	if (init_eaccess(ramdisk_execute_command) != 0) {
1564 		ramdisk_execute_command = NULL;
1565 		prepare_namespace();
1566 	}
1567 
1568 	/*
1569 	 * Ok, we have completed the initial bootup, and
1570 	 * we're essentially up and running. Get rid of the
1571 	 * initmem segments and start the user-mode stuff..
1572 	 *
1573 	 * rootfs is available now, try loading the public keys
1574 	 * and default modules
1575 	 */
1576 
1577 	integrity_load_keys();
1578 }
1579