xref: /openbmc/linux/init/main.c (revision e4c881d2)
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 	len = xlen + strlen(command_line) + 1;
634 
635 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
636 	if (!static_command_line)
637 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
638 
639 	if (xlen) {
640 		/*
641 		 * We have to put extra_command_line before boot command
642 		 * lines because there could be dashes (separator of init
643 		 * command line) in the command lines.
644 		 */
645 		strcpy(saved_command_line, extra_command_line);
646 		strcpy(static_command_line, extra_command_line);
647 	}
648 	strcpy(saved_command_line + xlen, boot_command_line);
649 	strcpy(static_command_line + xlen, command_line);
650 
651 	if (ilen) {
652 		/*
653 		 * Append supplemental init boot args to saved_command_line
654 		 * so that user can check what command line options passed
655 		 * to init.
656 		 * The order should always be
657 		 * " -- "[bootconfig init-param][cmdline init-param]
658 		 */
659 		if (initargs_offs) {
660 			len = xlen + initargs_offs;
661 			strcpy(saved_command_line + len, extra_init_args);
662 			len += ilen - 4;	/* strlen(extra_init_args) */
663 			strcpy(saved_command_line + len,
664 				boot_command_line + initargs_offs - 1);
665 		} else {
666 			len = strlen(saved_command_line);
667 			strcpy(saved_command_line + len, " -- ");
668 			len += 4;
669 			strcpy(saved_command_line + len, extra_init_args);
670 		}
671 	}
672 
673 	saved_command_line_len = strlen(saved_command_line);
674 }
675 
676 /*
677  * We need to finalize in a non-__init function or else race conditions
678  * between the root thread and the init thread may cause start_kernel to
679  * be reaped by free_initmem before the root thread has proceeded to
680  * cpu_idle.
681  *
682  * gcc-3.4 accidentally inlines this function, so use noinline.
683  */
684 
685 static __initdata DECLARE_COMPLETION(kthreadd_done);
686 
687 noinline void __ref __noreturn rest_init(void)
688 {
689 	struct task_struct *tsk;
690 	int pid;
691 
692 	rcu_scheduler_starting();
693 	/*
694 	 * We need to spawn init first so that it obtains pid 1, however
695 	 * the init task will end up wanting to create kthreads, which, if
696 	 * we schedule it before we create kthreadd, will OOPS.
697 	 */
698 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
699 	/*
700 	 * Pin init on the boot CPU. Task migration is not properly working
701 	 * until sched_init_smp() has been run. It will set the allowed
702 	 * CPUs for init to the non isolated CPUs.
703 	 */
704 	rcu_read_lock();
705 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
706 	tsk->flags |= PF_NO_SETAFFINITY;
707 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
708 	rcu_read_unlock();
709 
710 	numa_default_policy();
711 	pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
712 	rcu_read_lock();
713 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
714 	rcu_read_unlock();
715 
716 	/*
717 	 * Enable might_sleep() and smp_processor_id() checks.
718 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
719 	 * kernel_thread() would trigger might_sleep() splats. With
720 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
721 	 * already, but it's stuck on the kthreadd_done completion.
722 	 */
723 	system_state = SYSTEM_SCHEDULING;
724 
725 	complete(&kthreadd_done);
726 
727 	/*
728 	 * The boot idle thread must execute schedule()
729 	 * at least once to get things moving:
730 	 */
731 	schedule_preempt_disabled();
732 	/* Call into cpu_idle with preempt disabled */
733 	cpu_startup_entry(CPUHP_ONLINE);
734 }
735 
736 /* Check for early params. */
737 static int __init do_early_param(char *param, char *val,
738 				 const char *unused, void *arg)
739 {
740 	const struct obs_kernel_param *p;
741 
742 	for (p = __setup_start; p < __setup_end; p++) {
743 		if ((p->early && parameq(param, p->str)) ||
744 		    (strcmp(param, "console") == 0 &&
745 		     strcmp(p->str, "earlycon") == 0)
746 		) {
747 			if (p->setup_func(val) != 0)
748 				pr_warn("Malformed early option '%s'\n", param);
749 		}
750 	}
751 	/* We accept everything at this stage. */
752 	return 0;
753 }
754 
755 void __init parse_early_options(char *cmdline)
756 {
757 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
758 		   do_early_param);
759 }
760 
761 /* Arch code calls this early on, or if not, just before other parsing. */
762 void __init parse_early_param(void)
763 {
764 	static int done __initdata;
765 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
766 
767 	if (done)
768 		return;
769 
770 	/* All fall through to do_early_param. */
771 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
772 	parse_early_options(tmp_cmdline);
773 	done = 1;
774 }
775 
776 void __init __weak arch_post_acpi_subsys_init(void) { }
777 
778 void __init __weak smp_setup_processor_id(void)
779 {
780 }
781 
782 # if THREAD_SIZE >= PAGE_SIZE
783 void __init __weak thread_stack_cache_init(void)
784 {
785 }
786 #endif
787 
788 void __init __weak poking_init(void) { }
789 
790 void __init __weak pgtable_cache_init(void) { }
791 
792 void __init __weak trap_init(void) { }
793 
794 bool initcall_debug;
795 core_param(initcall_debug, initcall_debug, bool, 0644);
796 
797 #ifdef TRACEPOINTS_ENABLED
798 static void __init initcall_debug_enable(void);
799 #else
800 static inline void initcall_debug_enable(void)
801 {
802 }
803 #endif
804 
805 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
806 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
807 			   randomize_kstack_offset);
808 DEFINE_PER_CPU(u32, kstack_offset);
809 
810 static int __init early_randomize_kstack_offset(char *buf)
811 {
812 	int ret;
813 	bool bool_result;
814 
815 	ret = kstrtobool(buf, &bool_result);
816 	if (ret)
817 		return ret;
818 
819 	if (bool_result)
820 		static_branch_enable(&randomize_kstack_offset);
821 	else
822 		static_branch_disable(&randomize_kstack_offset);
823 	return 0;
824 }
825 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
826 #endif
827 
828 void __init __weak __noreturn arch_call_rest_init(void)
829 {
830 	rest_init();
831 }
832 
833 static void __init print_unknown_bootoptions(void)
834 {
835 	char *unknown_options;
836 	char *end;
837 	const char *const *p;
838 	size_t len;
839 
840 	if (panic_later || (!argv_init[1] && !envp_init[2]))
841 		return;
842 
843 	/*
844 	 * Determine how many options we have to print out, plus a space
845 	 * before each
846 	 */
847 	len = 1; /* null terminator */
848 	for (p = &argv_init[1]; *p; p++) {
849 		len++;
850 		len += strlen(*p);
851 	}
852 	for (p = &envp_init[2]; *p; p++) {
853 		len++;
854 		len += strlen(*p);
855 	}
856 
857 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
858 	if (!unknown_options) {
859 		pr_err("%s: Failed to allocate %zu bytes\n",
860 			__func__, len);
861 		return;
862 	}
863 	end = unknown_options;
864 
865 	for (p = &argv_init[1]; *p; p++)
866 		end += sprintf(end, " %s", *p);
867 	for (p = &envp_init[2]; *p; p++)
868 		end += sprintf(end, " %s", *p);
869 
870 	/* Start at unknown_options[1] to skip the initial space */
871 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
872 		&unknown_options[1]);
873 	memblock_free(unknown_options, len);
874 }
875 
876 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
877 void start_kernel(void)
878 {
879 	char *command_line;
880 	char *after_dashes;
881 
882 	set_task_stack_end_magic(&init_task);
883 	smp_setup_processor_id();
884 	debug_objects_early_init();
885 	init_vmlinux_build_id();
886 
887 	cgroup_init_early();
888 
889 	local_irq_disable();
890 	early_boot_irqs_disabled = true;
891 
892 	/*
893 	 * Interrupts are still disabled. Do necessary setups, then
894 	 * enable them.
895 	 */
896 	boot_cpu_init();
897 	page_address_init();
898 	pr_notice("%s", linux_banner);
899 	early_security_init();
900 	setup_arch(&command_line);
901 	setup_boot_config();
902 	setup_command_line(command_line);
903 	setup_nr_cpu_ids();
904 	setup_per_cpu_areas();
905 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
906 	boot_cpu_hotplug_init();
907 
908 	pr_notice("Kernel command line: %s\n", saved_command_line);
909 	/* parameters may set static keys */
910 	jump_label_init();
911 	parse_early_param();
912 	after_dashes = parse_args("Booting kernel",
913 				  static_command_line, __start___param,
914 				  __stop___param - __start___param,
915 				  -1, -1, NULL, &unknown_bootoption);
916 	print_unknown_bootoptions();
917 	if (!IS_ERR_OR_NULL(after_dashes))
918 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
919 			   NULL, set_init_arg);
920 	if (extra_init_args)
921 		parse_args("Setting extra init args", extra_init_args,
922 			   NULL, 0, -1, -1, NULL, set_init_arg);
923 
924 	/* Architectural and non-timekeeping rng init, before allocator init */
925 	random_init_early(command_line);
926 
927 	/*
928 	 * These use large bootmem allocations and must precede
929 	 * initalization of page allocator
930 	 */
931 	setup_log_buf(0);
932 	vfs_caches_init_early();
933 	sort_main_extable();
934 	trap_init();
935 	mm_core_init();
936 	poking_init();
937 	ftrace_init();
938 
939 	/* trace_printk can be enabled here */
940 	early_trace_init();
941 
942 	/*
943 	 * Set up the scheduler prior starting any interrupts (such as the
944 	 * timer interrupt). Full topology setup happens at smp_init()
945 	 * time - but meanwhile we still have a functioning scheduler.
946 	 */
947 	sched_init();
948 
949 	if (WARN(!irqs_disabled(),
950 		 "Interrupts were enabled *very* early, fixing it\n"))
951 		local_irq_disable();
952 	radix_tree_init();
953 	maple_tree_init();
954 
955 	/*
956 	 * Set up housekeeping before setting up workqueues to allow the unbound
957 	 * workqueue to take non-housekeeping into account.
958 	 */
959 	housekeeping_init();
960 
961 	/*
962 	 * Allow workqueue creation and work item queueing/cancelling
963 	 * early.  Work item execution depends on kthreads and starts after
964 	 * workqueue_init().
965 	 */
966 	workqueue_init_early();
967 
968 	rcu_init();
969 
970 	/* Trace events are available after this */
971 	trace_init();
972 
973 	if (initcall_debug)
974 		initcall_debug_enable();
975 
976 	context_tracking_init();
977 	/* init some links before init_ISA_irqs() */
978 	early_irq_init();
979 	init_IRQ();
980 	tick_init();
981 	rcu_init_nohz();
982 	init_timers();
983 	srcu_init();
984 	hrtimers_init();
985 	softirq_init();
986 	timekeeping_init();
987 	time_init();
988 
989 	/* This must be after timekeeping is initialized */
990 	random_init();
991 
992 	/* These make use of the fully initialized rng */
993 	kfence_init();
994 	boot_init_stack_canary();
995 
996 	perf_event_init();
997 	profile_init();
998 	call_function_init();
999 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1000 
1001 	early_boot_irqs_disabled = false;
1002 	local_irq_enable();
1003 
1004 	kmem_cache_init_late();
1005 
1006 	/*
1007 	 * HACK ALERT! This is early. We're enabling the console before
1008 	 * we've done PCI setups etc, and console_init() must be aware of
1009 	 * this. But we do want output early, in case something goes wrong.
1010 	 */
1011 	console_init();
1012 	if (panic_later)
1013 		panic("Too many boot %s vars at `%s'", panic_later,
1014 		      panic_param);
1015 
1016 	lockdep_init();
1017 
1018 	/*
1019 	 * Need to run this when irqs are enabled, because it wants
1020 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1021 	 * too:
1022 	 */
1023 	locking_selftest();
1024 
1025 #ifdef CONFIG_BLK_DEV_INITRD
1026 	if (initrd_start && !initrd_below_start_ok &&
1027 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1028 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1029 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1030 		    min_low_pfn);
1031 		initrd_start = 0;
1032 	}
1033 #endif
1034 	setup_per_cpu_pageset();
1035 	numa_policy_init();
1036 	acpi_early_init();
1037 	if (late_time_init)
1038 		late_time_init();
1039 	sched_clock_init();
1040 	calibrate_delay();
1041 
1042 	arch_cpu_finalize_init();
1043 
1044 	pid_idr_init();
1045 	anon_vma_init();
1046 #ifdef CONFIG_X86
1047 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1048 		efi_enter_virtual_mode();
1049 #endif
1050 	thread_stack_cache_init();
1051 	cred_init();
1052 	fork_init();
1053 	proc_caches_init();
1054 	uts_ns_init();
1055 	key_init();
1056 	security_init();
1057 	dbg_late_init();
1058 	net_ns_init();
1059 	vfs_caches_init();
1060 	pagecache_init();
1061 	signals_init();
1062 	seq_file_init();
1063 	proc_root_init();
1064 	nsfs_init();
1065 	cpuset_init();
1066 	cgroup_init();
1067 	taskstats_init_early();
1068 	delayacct_init();
1069 
1070 	acpi_subsystem_init();
1071 	arch_post_acpi_subsys_init();
1072 	kcsan_init();
1073 
1074 	/* Do the rest non-__init'ed, we're now alive */
1075 	arch_call_rest_init();
1076 
1077 	/*
1078 	 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1079 	 * and older.
1080 	 */
1081 #if !__has_attribute(__no_stack_protector__)
1082 	prevent_tail_call_optimization();
1083 #endif
1084 }
1085 
1086 /* Call all constructor functions linked into the kernel. */
1087 static void __init do_ctors(void)
1088 {
1089 /*
1090  * For UML, the constructors have already been called by the
1091  * normal setup code as it's just a normal ELF binary, so we
1092  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1093  * even on UML for modules.
1094  */
1095 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1096 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1097 
1098 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1099 		(*fn)();
1100 #endif
1101 }
1102 
1103 #ifdef CONFIG_KALLSYMS
1104 struct blacklist_entry {
1105 	struct list_head next;
1106 	char *buf;
1107 };
1108 
1109 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1110 
1111 static int __init initcall_blacklist(char *str)
1112 {
1113 	char *str_entry;
1114 	struct blacklist_entry *entry;
1115 
1116 	/* str argument is a comma-separated list of functions */
1117 	do {
1118 		str_entry = strsep(&str, ",");
1119 		if (str_entry) {
1120 			pr_debug("blacklisting initcall %s\n", str_entry);
1121 			entry = memblock_alloc(sizeof(*entry),
1122 					       SMP_CACHE_BYTES);
1123 			if (!entry)
1124 				panic("%s: Failed to allocate %zu bytes\n",
1125 				      __func__, sizeof(*entry));
1126 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1127 						    SMP_CACHE_BYTES);
1128 			if (!entry->buf)
1129 				panic("%s: Failed to allocate %zu bytes\n",
1130 				      __func__, strlen(str_entry) + 1);
1131 			strcpy(entry->buf, str_entry);
1132 			list_add(&entry->next, &blacklisted_initcalls);
1133 		}
1134 	} while (str_entry);
1135 
1136 	return 1;
1137 }
1138 
1139 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1140 {
1141 	struct blacklist_entry *entry;
1142 	char fn_name[KSYM_SYMBOL_LEN];
1143 	unsigned long addr;
1144 
1145 	if (list_empty(&blacklisted_initcalls))
1146 		return false;
1147 
1148 	addr = (unsigned long) dereference_function_descriptor(fn);
1149 	sprint_symbol_no_offset(fn_name, addr);
1150 
1151 	/*
1152 	 * fn will be "function_name [module_name]" where [module_name] is not
1153 	 * displayed for built-in init functions.  Strip off the [module_name].
1154 	 */
1155 	strreplace(fn_name, ' ', '\0');
1156 
1157 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1158 		if (!strcmp(fn_name, entry->buf)) {
1159 			pr_debug("initcall %s blacklisted\n", fn_name);
1160 			return true;
1161 		}
1162 	}
1163 
1164 	return false;
1165 }
1166 #else
1167 static int __init initcall_blacklist(char *str)
1168 {
1169 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1170 	return 0;
1171 }
1172 
1173 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1174 {
1175 	return false;
1176 }
1177 #endif
1178 __setup("initcall_blacklist=", initcall_blacklist);
1179 
1180 static __init_or_module void
1181 trace_initcall_start_cb(void *data, initcall_t fn)
1182 {
1183 	ktime_t *calltime = data;
1184 
1185 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1186 	*calltime = ktime_get();
1187 }
1188 
1189 static __init_or_module void
1190 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1191 {
1192 	ktime_t rettime, *calltime = data;
1193 
1194 	rettime = ktime_get();
1195 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1196 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1197 }
1198 
1199 static ktime_t initcall_calltime;
1200 
1201 #ifdef TRACEPOINTS_ENABLED
1202 static void __init initcall_debug_enable(void)
1203 {
1204 	int ret;
1205 
1206 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1207 					    &initcall_calltime);
1208 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1209 					      &initcall_calltime);
1210 	WARN(ret, "Failed to register initcall tracepoints\n");
1211 }
1212 # define do_trace_initcall_start	trace_initcall_start
1213 # define do_trace_initcall_finish	trace_initcall_finish
1214 #else
1215 static inline void do_trace_initcall_start(initcall_t fn)
1216 {
1217 	if (!initcall_debug)
1218 		return;
1219 	trace_initcall_start_cb(&initcall_calltime, fn);
1220 }
1221 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1222 {
1223 	if (!initcall_debug)
1224 		return;
1225 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1226 }
1227 #endif /* !TRACEPOINTS_ENABLED */
1228 
1229 int __init_or_module do_one_initcall(initcall_t fn)
1230 {
1231 	int count = preempt_count();
1232 	char msgbuf[64];
1233 	int ret;
1234 
1235 	if (initcall_blacklisted(fn))
1236 		return -EPERM;
1237 
1238 	do_trace_initcall_start(fn);
1239 	ret = fn();
1240 	do_trace_initcall_finish(fn, ret);
1241 
1242 	msgbuf[0] = 0;
1243 
1244 	if (preempt_count() != count) {
1245 		sprintf(msgbuf, "preemption imbalance ");
1246 		preempt_count_set(count);
1247 	}
1248 	if (irqs_disabled()) {
1249 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1250 		local_irq_enable();
1251 	}
1252 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1253 
1254 	add_latent_entropy();
1255 	return ret;
1256 }
1257 
1258 
1259 static initcall_entry_t *initcall_levels[] __initdata = {
1260 	__initcall0_start,
1261 	__initcall1_start,
1262 	__initcall2_start,
1263 	__initcall3_start,
1264 	__initcall4_start,
1265 	__initcall5_start,
1266 	__initcall6_start,
1267 	__initcall7_start,
1268 	__initcall_end,
1269 };
1270 
1271 /* Keep these in sync with initcalls in include/linux/init.h */
1272 static const char *initcall_level_names[] __initdata = {
1273 	"pure",
1274 	"core",
1275 	"postcore",
1276 	"arch",
1277 	"subsys",
1278 	"fs",
1279 	"device",
1280 	"late",
1281 };
1282 
1283 static int __init ignore_unknown_bootoption(char *param, char *val,
1284 			       const char *unused, void *arg)
1285 {
1286 	return 0;
1287 }
1288 
1289 static void __init do_initcall_level(int level, char *command_line)
1290 {
1291 	initcall_entry_t *fn;
1292 
1293 	parse_args(initcall_level_names[level],
1294 		   command_line, __start___param,
1295 		   __stop___param - __start___param,
1296 		   level, level,
1297 		   NULL, ignore_unknown_bootoption);
1298 
1299 	trace_initcall_level(initcall_level_names[level]);
1300 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1301 		do_one_initcall(initcall_from_entry(fn));
1302 }
1303 
1304 static void __init do_initcalls(void)
1305 {
1306 	int level;
1307 	size_t len = saved_command_line_len + 1;
1308 	char *command_line;
1309 
1310 	command_line = kzalloc(len, GFP_KERNEL);
1311 	if (!command_line)
1312 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1313 
1314 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1315 		/* Parser modifies command_line, restore it each time */
1316 		strcpy(command_line, saved_command_line);
1317 		do_initcall_level(level, command_line);
1318 	}
1319 
1320 	kfree(command_line);
1321 }
1322 
1323 /*
1324  * Ok, the machine is now initialized. None of the devices
1325  * have been touched yet, but the CPU subsystem is up and
1326  * running, and memory and process management works.
1327  *
1328  * Now we can finally start doing some real work..
1329  */
1330 static void __init do_basic_setup(void)
1331 {
1332 	cpuset_init_smp();
1333 	driver_init();
1334 	init_irq_proc();
1335 	do_ctors();
1336 	do_initcalls();
1337 }
1338 
1339 static void __init do_pre_smp_initcalls(void)
1340 {
1341 	initcall_entry_t *fn;
1342 
1343 	trace_initcall_level("early");
1344 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1345 		do_one_initcall(initcall_from_entry(fn));
1346 }
1347 
1348 static int run_init_process(const char *init_filename)
1349 {
1350 	const char *const *p;
1351 
1352 	argv_init[0] = init_filename;
1353 	pr_info("Run %s as init process\n", init_filename);
1354 	pr_debug("  with arguments:\n");
1355 	for (p = argv_init; *p; p++)
1356 		pr_debug("    %s\n", *p);
1357 	pr_debug("  with environment:\n");
1358 	for (p = envp_init; *p; p++)
1359 		pr_debug("    %s\n", *p);
1360 	return kernel_execve(init_filename, argv_init, envp_init);
1361 }
1362 
1363 static int try_to_run_init_process(const char *init_filename)
1364 {
1365 	int ret;
1366 
1367 	ret = run_init_process(init_filename);
1368 
1369 	if (ret && ret != -ENOENT) {
1370 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1371 		       init_filename, ret);
1372 	}
1373 
1374 	return ret;
1375 }
1376 
1377 static noinline void __init kernel_init_freeable(void);
1378 
1379 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1380 bool rodata_enabled __ro_after_init = true;
1381 
1382 #ifndef arch_parse_debug_rodata
1383 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1384 #endif
1385 
1386 static int __init set_debug_rodata(char *str)
1387 {
1388 	if (arch_parse_debug_rodata(str))
1389 		return 0;
1390 
1391 	if (str && !strcmp(str, "on"))
1392 		rodata_enabled = true;
1393 	else if (str && !strcmp(str, "off"))
1394 		rodata_enabled = false;
1395 	else
1396 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1397 	return 0;
1398 }
1399 early_param("rodata", set_debug_rodata);
1400 #endif
1401 
1402 #ifdef CONFIG_STRICT_KERNEL_RWX
1403 static void mark_readonly(void)
1404 {
1405 	if (rodata_enabled) {
1406 		/*
1407 		 * load_module() results in W+X mappings, which are cleaned
1408 		 * up with init_free_wq. Let's make sure that queued work is
1409 		 * flushed so that we don't hit false positives looking for
1410 		 * insecure pages which are W+X.
1411 		 */
1412 		flush_module_init_free_work();
1413 		mark_rodata_ro();
1414 		rodata_test();
1415 	} else
1416 		pr_info("Kernel memory protection disabled.\n");
1417 }
1418 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1419 static inline void mark_readonly(void)
1420 {
1421 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1422 }
1423 #else
1424 static inline void mark_readonly(void)
1425 {
1426 	pr_warn("This architecture does not have kernel memory protection.\n");
1427 }
1428 #endif
1429 
1430 void __weak free_initmem(void)
1431 {
1432 	free_initmem_default(POISON_FREE_INITMEM);
1433 }
1434 
1435 static int __ref kernel_init(void *unused)
1436 {
1437 	int ret;
1438 
1439 	/*
1440 	 * Wait until kthreadd is all set-up.
1441 	 */
1442 	wait_for_completion(&kthreadd_done);
1443 
1444 	kernel_init_freeable();
1445 	/* need to finish all async __init code before freeing the memory */
1446 	async_synchronize_full();
1447 
1448 	system_state = SYSTEM_FREEING_INITMEM;
1449 	kprobe_free_init_mem();
1450 	ftrace_free_init_mem();
1451 	kgdb_free_init_mem();
1452 	exit_boot_config();
1453 	free_initmem();
1454 	mark_readonly();
1455 
1456 	/*
1457 	 * Kernel mappings are now finalized - update the userspace page-table
1458 	 * to finalize PTI.
1459 	 */
1460 	pti_finalize();
1461 
1462 	system_state = SYSTEM_RUNNING;
1463 	numa_default_policy();
1464 
1465 	rcu_end_inkernel_boot();
1466 
1467 	do_sysctl_args();
1468 
1469 	if (ramdisk_execute_command) {
1470 		ret = run_init_process(ramdisk_execute_command);
1471 		if (!ret)
1472 			return 0;
1473 		pr_err("Failed to execute %s (error %d)\n",
1474 		       ramdisk_execute_command, ret);
1475 	}
1476 
1477 	/*
1478 	 * We try each of these until one succeeds.
1479 	 *
1480 	 * The Bourne shell can be used instead of init if we are
1481 	 * trying to recover a really broken machine.
1482 	 */
1483 	if (execute_command) {
1484 		ret = run_init_process(execute_command);
1485 		if (!ret)
1486 			return 0;
1487 		panic("Requested init %s failed (error %d).",
1488 		      execute_command, ret);
1489 	}
1490 
1491 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1492 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1493 		if (ret)
1494 			pr_err("Default init %s failed (error %d)\n",
1495 			       CONFIG_DEFAULT_INIT, ret);
1496 		else
1497 			return 0;
1498 	}
1499 
1500 	if (!try_to_run_init_process("/sbin/init") ||
1501 	    !try_to_run_init_process("/etc/init") ||
1502 	    !try_to_run_init_process("/bin/init") ||
1503 	    !try_to_run_init_process("/bin/sh"))
1504 		return 0;
1505 
1506 	panic("No working init found.  Try passing init= option to kernel. "
1507 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1508 }
1509 
1510 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1511 void __init console_on_rootfs(void)
1512 {
1513 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1514 
1515 	if (IS_ERR(file)) {
1516 		pr_err("Warning: unable to open an initial console.\n");
1517 		return;
1518 	}
1519 	init_dup(file);
1520 	init_dup(file);
1521 	init_dup(file);
1522 	fput(file);
1523 }
1524 
1525 static noinline void __init kernel_init_freeable(void)
1526 {
1527 	/* Now the scheduler is fully set up and can do blocking allocations */
1528 	gfp_allowed_mask = __GFP_BITS_MASK;
1529 
1530 	/*
1531 	 * init can allocate pages on any node
1532 	 */
1533 	set_mems_allowed(node_states[N_MEMORY]);
1534 
1535 	cad_pid = get_pid(task_pid(current));
1536 
1537 	smp_prepare_cpus(setup_max_cpus);
1538 
1539 	workqueue_init();
1540 
1541 	init_mm_internals();
1542 
1543 	rcu_init_tasks_generic();
1544 	do_pre_smp_initcalls();
1545 	lockup_detector_init();
1546 
1547 	smp_init();
1548 	sched_init_smp();
1549 
1550 	workqueue_init_topology();
1551 	padata_init();
1552 	page_alloc_init_late();
1553 
1554 	do_basic_setup();
1555 
1556 	kunit_run_all_tests();
1557 
1558 	wait_for_initramfs();
1559 	console_on_rootfs();
1560 
1561 	/*
1562 	 * check if there is an early userspace init.  If yes, let it do all
1563 	 * the work
1564 	 */
1565 	if (init_eaccess(ramdisk_execute_command) != 0) {
1566 		ramdisk_execute_command = NULL;
1567 		prepare_namespace();
1568 	}
1569 
1570 	/*
1571 	 * Ok, we have completed the initial bootup, and
1572 	 * we're essentially up and running. Get rid of the
1573 	 * initmem segments and start the user-mode stuff..
1574 	 *
1575 	 * rootfs is available now, try loading the public keys
1576 	 * and default modules
1577 	 */
1578 
1579 	integrity_load_keys();
1580 }
1581