xref: /linux/init/main.c (revision f58f45c1e5b92975e91754f5407250085a6ae7cf)
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 <linux/pidfs.h>
104 #include <linux/ptdump.h>
105 #include <net/net_namespace.h>
106 
107 #include <asm/io.h>
108 #include <asm/setup.h>
109 #include <asm/sections.h>
110 #include <asm/cacheflush.h>
111 
112 #define CREATE_TRACE_POINTS
113 #include <trace/events/initcall.h>
114 
115 #include <kunit/test.h>
116 
117 static int kernel_init(void *);
118 
119 /*
120  * Debug helper: via this flag we know that we are in 'early bootup code'
121  * where only the boot processor is running with IRQ disabled.  This means
122  * two things - IRQ must not be enabled before the flag is cleared and some
123  * operations which are not allowed with IRQ disabled are allowed while the
124  * flag is set.
125  */
126 bool early_boot_irqs_disabled __read_mostly;
127 
128 enum system_states system_state __read_mostly;
129 EXPORT_SYMBOL(system_state);
130 
131 /*
132  * Boot command-line arguments
133  */
134 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
135 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
136 
137 /* Default late time init is NULL. archs can override this later. */
138 void (*__initdata late_time_init)(void);
139 
140 /* Untouched command line saved by arch-specific code. */
141 char __initdata boot_command_line[COMMAND_LINE_SIZE];
142 /* Untouched saved command line (eg. for /proc) */
143 char *saved_command_line __ro_after_init;
144 unsigned int saved_command_line_len __ro_after_init;
145 /* Command line for parameter parsing */
146 static char *static_command_line;
147 /* Untouched extra command line */
148 static char *extra_command_line;
149 /* Extra init arguments */
150 static char *extra_init_args;
151 
152 #ifdef CONFIG_BOOT_CONFIG
153 /* Is bootconfig on command line? */
154 static bool bootconfig_found;
155 static size_t initargs_offs;
156 #else
157 # define bootconfig_found false
158 # define initargs_offs 0
159 #endif
160 
161 static char *execute_command;
162 static char *ramdisk_execute_command = "/init";
163 
164 /*
165  * Used to generate warnings if static_key manipulation functions are used
166  * before jump_label_init is called.
167  */
168 bool static_key_initialized __read_mostly;
169 EXPORT_SYMBOL_GPL(static_key_initialized);
170 
171 /*
172  * If set, this is an indication to the drivers that reset the underlying
173  * device before going ahead with the initialization otherwise driver might
174  * rely on the BIOS and skip the reset operation.
175  *
176  * This is useful if kernel is booting in an unreliable environment.
177  * For ex. kdump situation where previous kernel has crashed, BIOS has been
178  * skipped and devices will be in unknown state.
179  */
180 unsigned int reset_devices;
181 EXPORT_SYMBOL(reset_devices);
182 
183 static int __init set_reset_devices(char *str)
184 {
185 	reset_devices = 1;
186 	return 1;
187 }
188 
189 __setup("reset_devices", set_reset_devices);
190 
191 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
192 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
193 static const char *panic_later, *panic_param;
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(size_t *_size)
267 {
268 	u32 size, csum;
269 	char *data;
270 	u32 *hdr;
271 	int i;
272 
273 	if (!initrd_end)
274 		return NULL;
275 
276 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
277 	/*
278 	 * Since Grub may align the size of initrd to 4, we must
279 	 * check the preceding 3 bytes as well.
280 	 */
281 	for (i = 0; i < 4; i++) {
282 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
283 			goto found;
284 		data--;
285 	}
286 	return NULL;
287 
288 found:
289 	hdr = (u32 *)(data - 8);
290 	size = le32_to_cpu(hdr[0]);
291 	csum = le32_to_cpu(hdr[1]);
292 
293 	data = ((void *)hdr) - size;
294 	if ((unsigned long)data < initrd_start) {
295 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
296 			size, initrd_end - initrd_start);
297 		return NULL;
298 	}
299 
300 	if (xbc_calc_checksum(data, size) != csum) {
301 		pr_err("bootconfig checksum failed\n");
302 		return NULL;
303 	}
304 
305 	/* Remove bootconfig from initramfs/initrd */
306 	initrd_end = (unsigned long)data;
307 	if (_size)
308 		*_size = size;
309 
310 	return data;
311 }
312 #else
313 static void * __init get_boot_config_from_initrd(size_t *_size)
314 {
315 	return NULL;
316 }
317 #endif
318 
319 #ifdef CONFIG_BOOT_CONFIG
320 
321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
322 
323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
324 
325 static int __init xbc_snprint_cmdline(char *buf, size_t size,
326 				      struct xbc_node *root)
327 {
328 	struct xbc_node *knode, *vnode;
329 	char *end = buf + size;
330 	const char *val;
331 	int ret;
332 
333 	xbc_node_for_each_key_value(root, knode, val) {
334 		ret = xbc_node_compose_key_after(root, knode,
335 					xbc_namebuf, XBC_KEYLEN_MAX);
336 		if (ret < 0)
337 			return ret;
338 
339 		vnode = xbc_node_get_child(knode);
340 		if (!vnode) {
341 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
342 			if (ret < 0)
343 				return ret;
344 			buf += ret;
345 			continue;
346 		}
347 		xbc_array_for_each_value(vnode, val) {
348 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
349 				       xbc_namebuf, val);
350 			if (ret < 0)
351 				return ret;
352 			buf += ret;
353 		}
354 	}
355 
356 	return buf - (end - size);
357 }
358 #undef rest
359 
360 /* Make an extra command line under given key word */
361 static char * __init xbc_make_cmdline(const char *key)
362 {
363 	struct xbc_node *root;
364 	char *new_cmdline;
365 	int ret, len = 0;
366 
367 	root = xbc_find_node(key);
368 	if (!root)
369 		return NULL;
370 
371 	/* Count required buffer size */
372 	len = xbc_snprint_cmdline(NULL, 0, root);
373 	if (len <= 0)
374 		return NULL;
375 
376 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
377 	if (!new_cmdline) {
378 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
379 		return NULL;
380 	}
381 
382 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
383 	if (ret < 0 || ret > len) {
384 		pr_err("Failed to print extra kernel cmdline.\n");
385 		memblock_free(new_cmdline, len + 1);
386 		return NULL;
387 	}
388 
389 	return new_cmdline;
390 }
391 
392 static int __init bootconfig_params(char *param, char *val,
393 				    const char *unused, void *arg)
394 {
395 	if (strcmp(param, "bootconfig") == 0) {
396 		bootconfig_found = true;
397 	}
398 	return 0;
399 }
400 
401 static int __init warn_bootconfig(char *str)
402 {
403 	/* The 'bootconfig' has been handled by bootconfig_params(). */
404 	return 0;
405 }
406 
407 static void __init setup_boot_config(void)
408 {
409 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
410 	const char *msg, *data;
411 	int pos, ret;
412 	size_t size;
413 	char *err;
414 
415 	/* Cut out the bootconfig data even if we have no bootconfig option */
416 	data = get_boot_config_from_initrd(&size);
417 	/* If there is no bootconfig in initrd, try embedded one. */
418 	if (!data)
419 		data = xbc_get_embedded_bootconfig(&size);
420 
421 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
422 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
423 			 bootconfig_params);
424 
425 	if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
426 		return;
427 
428 	/* parse_args() stops at the next param of '--' and returns an address */
429 	if (err)
430 		initargs_offs = err - tmp_cmdline;
431 
432 	if (!data) {
433 		/* If user intended to use bootconfig, show an error level message */
434 		if (bootconfig_found)
435 			pr_err("'bootconfig' found on command line, but no bootconfig found\n");
436 		else
437 			pr_info("No bootconfig data provided, so skipping bootconfig");
438 		return;
439 	}
440 
441 	if (size >= XBC_DATA_MAX) {
442 		pr_err("bootconfig size %ld greater than max size %d\n",
443 			(long)size, XBC_DATA_MAX);
444 		return;
445 	}
446 
447 	ret = xbc_init(data, size, &msg, &pos);
448 	if (ret < 0) {
449 		if (pos < 0)
450 			pr_err("Failed to init bootconfig: %s.\n", msg);
451 		else
452 			pr_err("Failed to parse bootconfig: %s at %d.\n",
453 				msg, pos);
454 	} else {
455 		xbc_get_info(&ret, NULL);
456 		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
457 		/* keys starting with "kernel." are passed via cmdline */
458 		extra_command_line = xbc_make_cmdline("kernel");
459 		/* Also, "init." keys are init arguments */
460 		extra_init_args = xbc_make_cmdline("init");
461 	}
462 	return;
463 }
464 
465 static void __init exit_boot_config(void)
466 {
467 	xbc_exit();
468 }
469 
470 #else	/* !CONFIG_BOOT_CONFIG */
471 
472 static void __init setup_boot_config(void)
473 {
474 	/* Remove bootconfig data from initrd */
475 	get_boot_config_from_initrd(NULL);
476 }
477 
478 static int __init warn_bootconfig(char *str)
479 {
480 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
481 	return 0;
482 }
483 
484 #define exit_boot_config()	do {} while (0)
485 
486 #endif	/* CONFIG_BOOT_CONFIG */
487 
488 early_param("bootconfig", warn_bootconfig);
489 
490 bool __init cmdline_has_extra_options(void)
491 {
492 	return extra_command_line || extra_init_args;
493 }
494 
495 /* Change NUL term back to "=", to make "param" the whole string. */
496 static void __init repair_env_string(char *param, char *val)
497 {
498 	if (val) {
499 		/* param=val or param="val"? */
500 		if (val == param+strlen(param)+1)
501 			val[-1] = '=';
502 		else if (val == param+strlen(param)+2) {
503 			val[-2] = '=';
504 			memmove(val-1, val, strlen(val)+1);
505 		} else
506 			BUG();
507 	}
508 }
509 
510 /* Anything after -- gets handed straight to init. */
511 static int __init set_init_arg(char *param, char *val,
512 			       const char *unused, void *arg)
513 {
514 	unsigned int i;
515 
516 	if (panic_later)
517 		return 0;
518 
519 	repair_env_string(param, val);
520 
521 	for (i = 0; argv_init[i]; i++) {
522 		if (i == MAX_INIT_ARGS) {
523 			panic_later = "init";
524 			panic_param = param;
525 			return 0;
526 		}
527 	}
528 	argv_init[i] = param;
529 	return 0;
530 }
531 
532 /*
533  * Unknown boot options get handed to init, unless they look like
534  * unused parameters (modprobe will find them in /proc/cmdline).
535  */
536 static int __init unknown_bootoption(char *param, char *val,
537 				     const char *unused, void *arg)
538 {
539 	size_t len = strlen(param);
540 
541 	/* Handle params aliased to sysctls */
542 	if (sysctl_is_alias(param))
543 		return 0;
544 
545 	repair_env_string(param, val);
546 
547 	/* Handle obsolete-style parameters */
548 	if (obsolete_checksetup(param))
549 		return 0;
550 
551 	/* Unused module parameter. */
552 	if (strnchr(param, len, '.'))
553 		return 0;
554 
555 	if (panic_later)
556 		return 0;
557 
558 	if (val) {
559 		/* Environment option */
560 		unsigned int i;
561 		for (i = 0; envp_init[i]; i++) {
562 			if (i == MAX_INIT_ENVS) {
563 				panic_later = "env";
564 				panic_param = param;
565 			}
566 			if (!strncmp(param, envp_init[i], len+1))
567 				break;
568 		}
569 		envp_init[i] = param;
570 	} else {
571 		/* Command line option */
572 		unsigned int i;
573 		for (i = 0; argv_init[i]; i++) {
574 			if (i == MAX_INIT_ARGS) {
575 				panic_later = "init";
576 				panic_param = param;
577 			}
578 		}
579 		argv_init[i] = param;
580 	}
581 	return 0;
582 }
583 
584 static int __init init_setup(char *str)
585 {
586 	unsigned int i;
587 
588 	execute_command = str;
589 	/*
590 	 * In case LILO is going to boot us with default command line,
591 	 * it prepends "auto" before the whole cmdline which makes
592 	 * the shell think it should execute a script with such name.
593 	 * So we ignore all arguments entered _before_ init=... [MJ]
594 	 */
595 	for (i = 1; i < MAX_INIT_ARGS; i++)
596 		argv_init[i] = NULL;
597 	return 1;
598 }
599 __setup("init=", init_setup);
600 
601 static int __init rdinit_setup(char *str)
602 {
603 	unsigned int i;
604 
605 	ramdisk_execute_command = str;
606 	/* See "auto" comment in init_setup */
607 	for (i = 1; i < MAX_INIT_ARGS; i++)
608 		argv_init[i] = NULL;
609 	return 1;
610 }
611 __setup("rdinit=", rdinit_setup);
612 
613 #ifndef CONFIG_SMP
614 static inline void setup_nr_cpu_ids(void) { }
615 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
616 #endif
617 
618 /*
619  * We need to store the untouched command line for future reference.
620  * We also need to store the touched command line since the parameter
621  * parsing is performed in place, and we should allow a component to
622  * store reference of name/value for future reference.
623  */
624 static void __init setup_command_line(char *command_line)
625 {
626 	size_t len, xlen = 0, ilen = 0;
627 
628 	if (extra_command_line)
629 		xlen = strlen(extra_command_line);
630 	if (extra_init_args)
631 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
632 
633 	len = xlen + strlen(boot_command_line) + 1;
634 
635 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
636 	if (!saved_command_line)
637 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
638 
639 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
640 	if (!static_command_line)
641 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
642 
643 	if (xlen) {
644 		/*
645 		 * We have to put extra_command_line before boot command
646 		 * lines because there could be dashes (separator of init
647 		 * command line) in the command lines.
648 		 */
649 		strcpy(saved_command_line, extra_command_line);
650 		strcpy(static_command_line, extra_command_line);
651 	}
652 	strcpy(saved_command_line + xlen, boot_command_line);
653 	strcpy(static_command_line + xlen, command_line);
654 
655 	if (ilen) {
656 		/*
657 		 * Append supplemental init boot args to saved_command_line
658 		 * so that user can check what command line options passed
659 		 * to init.
660 		 * The order should always be
661 		 * " -- "[bootconfig init-param][cmdline init-param]
662 		 */
663 		if (initargs_offs) {
664 			len = xlen + initargs_offs;
665 			strcpy(saved_command_line + len, extra_init_args);
666 			len += ilen - 4;	/* strlen(extra_init_args) */
667 			strcpy(saved_command_line + len,
668 				boot_command_line + initargs_offs - 1);
669 		} else {
670 			len = strlen(saved_command_line);
671 			strcpy(saved_command_line + len, " -- ");
672 			len += 4;
673 			strcpy(saved_command_line + len, extra_init_args);
674 		}
675 	}
676 
677 	saved_command_line_len = strlen(saved_command_line);
678 }
679 
680 /*
681  * We need to finalize in a non-__init function or else race conditions
682  * between the root thread and the init thread may cause start_kernel to
683  * be reaped by free_initmem before the root thread has proceeded to
684  * cpu_idle.
685  *
686  * gcc-3.4 accidentally inlines this function, so use noinline.
687  */
688 
689 static __initdata DECLARE_COMPLETION(kthreadd_done);
690 
691 static noinline void __ref __noreturn rest_init(void)
692 {
693 	struct task_struct *tsk;
694 	int pid;
695 
696 	rcu_scheduler_starting();
697 	/*
698 	 * We need to spawn init first so that it obtains pid 1, however
699 	 * the init task will end up wanting to create kthreads, which, if
700 	 * we schedule it before we create kthreadd, will OOPS.
701 	 */
702 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
703 	/*
704 	 * Pin init on the boot CPU. Task migration is not properly working
705 	 * until sched_init_smp() has been run. It will set the allowed
706 	 * CPUs for init to the non isolated CPUs.
707 	 */
708 	rcu_read_lock();
709 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
710 	tsk->flags |= PF_NO_SETAFFINITY;
711 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
712 	rcu_read_unlock();
713 
714 	numa_default_policy();
715 	pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
716 	rcu_read_lock();
717 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
718 	rcu_read_unlock();
719 
720 	/*
721 	 * Enable might_sleep() and smp_processor_id() checks.
722 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
723 	 * kernel_thread() would trigger might_sleep() splats. With
724 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
725 	 * already, but it's stuck on the kthreadd_done completion.
726 	 */
727 	system_state = SYSTEM_SCHEDULING;
728 
729 	complete(&kthreadd_done);
730 
731 	/*
732 	 * The boot idle thread must execute schedule()
733 	 * at least once to get things moving:
734 	 */
735 	schedule_preempt_disabled();
736 	/* Call into cpu_idle with preempt disabled */
737 	cpu_startup_entry(CPUHP_ONLINE);
738 }
739 
740 /* Check for early params. */
741 static int __init do_early_param(char *param, char *val,
742 				 const char *unused, void *arg)
743 {
744 	const struct obs_kernel_param *p;
745 
746 	for (p = __setup_start; p < __setup_end; p++) {
747 		if ((p->early && parameq(param, p->str)) ||
748 		    (strcmp(param, "console") == 0 &&
749 		     strcmp(p->str, "earlycon") == 0)
750 		) {
751 			if (p->setup_func(val) != 0)
752 				pr_warn("Malformed early option '%s'\n", param);
753 		}
754 	}
755 	/* We accept everything at this stage. */
756 	return 0;
757 }
758 
759 void __init parse_early_options(char *cmdline)
760 {
761 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
762 		   do_early_param);
763 }
764 
765 /* Arch code calls this early on, or if not, just before other parsing. */
766 void __init parse_early_param(void)
767 {
768 	static int done __initdata;
769 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
770 
771 	if (done)
772 		return;
773 
774 	/* All fall through to do_early_param. */
775 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
776 	parse_early_options(tmp_cmdline);
777 	done = 1;
778 }
779 
780 void __init __weak arch_post_acpi_subsys_init(void) { }
781 
782 void __init __weak smp_setup_processor_id(void)
783 {
784 }
785 
786 void __init __weak smp_prepare_boot_cpu(void)
787 {
788 }
789 
790 # if THREAD_SIZE >= PAGE_SIZE
791 void __init __weak thread_stack_cache_init(void)
792 {
793 }
794 #endif
795 
796 void __init __weak poking_init(void) { }
797 
798 void __init __weak pgtable_cache_init(void) { }
799 
800 void __init __weak trap_init(void) { }
801 
802 bool initcall_debug;
803 core_param(initcall_debug, initcall_debug, bool, 0644);
804 
805 #ifdef TRACEPOINTS_ENABLED
806 static void __init initcall_debug_enable(void);
807 #else
808 static inline void initcall_debug_enable(void)
809 {
810 }
811 #endif
812 
813 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
814 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
815 			   randomize_kstack_offset);
816 DEFINE_PER_CPU(u32, kstack_offset);
817 
818 static int __init early_randomize_kstack_offset(char *buf)
819 {
820 	int ret;
821 	bool bool_result;
822 
823 	ret = kstrtobool(buf, &bool_result);
824 	if (ret)
825 		return ret;
826 
827 	if (bool_result)
828 		static_branch_enable(&randomize_kstack_offset);
829 	else
830 		static_branch_disable(&randomize_kstack_offset);
831 	return 0;
832 }
833 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
834 #endif
835 
836 static void __init print_unknown_bootoptions(void)
837 {
838 	char *unknown_options;
839 	char *end;
840 	const char *const *p;
841 	size_t len;
842 
843 	if (panic_later || (!argv_init[1] && !envp_init[2]))
844 		return;
845 
846 	/*
847 	 * Determine how many options we have to print out, plus a space
848 	 * before each
849 	 */
850 	len = 1; /* null terminator */
851 	for (p = &argv_init[1]; *p; p++) {
852 		len++;
853 		len += strlen(*p);
854 	}
855 	for (p = &envp_init[2]; *p; p++) {
856 		len++;
857 		len += strlen(*p);
858 	}
859 
860 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
861 	if (!unknown_options) {
862 		pr_err("%s: Failed to allocate %zu bytes\n",
863 			__func__, len);
864 		return;
865 	}
866 	end = unknown_options;
867 
868 	for (p = &argv_init[1]; *p; p++)
869 		end += sprintf(end, " %s", *p);
870 	for (p = &envp_init[2]; *p; p++)
871 		end += sprintf(end, " %s", *p);
872 
873 	/* Start at unknown_options[1] to skip the initial space */
874 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
875 		&unknown_options[1]);
876 	memblock_free(unknown_options, len);
877 }
878 
879 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
880 void start_kernel(void)
881 {
882 	char *command_line;
883 	char *after_dashes;
884 
885 	set_task_stack_end_magic(&init_task);
886 	smp_setup_processor_id();
887 	debug_objects_early_init();
888 	init_vmlinux_build_id();
889 
890 	cgroup_init_early();
891 
892 	local_irq_disable();
893 	early_boot_irqs_disabled = true;
894 
895 	/*
896 	 * Interrupts are still disabled. Do necessary setups, then
897 	 * enable them.
898 	 */
899 	boot_cpu_init();
900 	page_address_init();
901 	pr_notice("%s", linux_banner);
902 	early_security_init();
903 	setup_arch(&command_line);
904 	setup_boot_config();
905 	setup_command_line(command_line);
906 	setup_nr_cpu_ids();
907 	setup_per_cpu_areas();
908 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
909 	boot_cpu_hotplug_init();
910 
911 	pr_notice("Kernel command line: %s\n", saved_command_line);
912 	/* parameters may set static keys */
913 	jump_label_init();
914 	parse_early_param();
915 	after_dashes = parse_args("Booting kernel",
916 				  static_command_line, __start___param,
917 				  __stop___param - __start___param,
918 				  -1, -1, NULL, &unknown_bootoption);
919 	print_unknown_bootoptions();
920 	if (!IS_ERR_OR_NULL(after_dashes))
921 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
922 			   NULL, set_init_arg);
923 	if (extra_init_args)
924 		parse_args("Setting extra init args", extra_init_args,
925 			   NULL, 0, -1, -1, NULL, set_init_arg);
926 
927 	/* Architectural and non-timekeeping rng init, before allocator init */
928 	random_init_early(command_line);
929 
930 	/*
931 	 * These use large bootmem allocations and must precede
932 	 * initalization of page allocator
933 	 */
934 	setup_log_buf(0);
935 	vfs_caches_init_early();
936 	sort_main_extable();
937 	trap_init();
938 	mm_core_init();
939 	poking_init();
940 	ftrace_init();
941 
942 	/* trace_printk can be enabled here */
943 	early_trace_init();
944 
945 	/*
946 	 * Set up the scheduler prior starting any interrupts (such as the
947 	 * timer interrupt). Full topology setup happens at smp_init()
948 	 * time - but meanwhile we still have a functioning scheduler.
949 	 */
950 	sched_init();
951 
952 	if (WARN(!irqs_disabled(),
953 		 "Interrupts were enabled *very* early, fixing it\n"))
954 		local_irq_disable();
955 	radix_tree_init();
956 	maple_tree_init();
957 
958 	/*
959 	 * Set up housekeeping before setting up workqueues to allow the unbound
960 	 * workqueue to take non-housekeeping into account.
961 	 */
962 	housekeeping_init();
963 
964 	/*
965 	 * Allow workqueue creation and work item queueing/cancelling
966 	 * early.  Work item execution depends on kthreads and starts after
967 	 * workqueue_init().
968 	 */
969 	workqueue_init_early();
970 
971 	rcu_init();
972 
973 	/* Trace events are available after this */
974 	trace_init();
975 
976 	if (initcall_debug)
977 		initcall_debug_enable();
978 
979 	context_tracking_init();
980 	/* init some links before init_ISA_irqs() */
981 	early_irq_init();
982 	init_IRQ();
983 	tick_init();
984 	rcu_init_nohz();
985 	init_timers();
986 	srcu_init();
987 	hrtimers_init();
988 	softirq_init();
989 	timekeeping_init();
990 	time_init();
991 
992 	/* This must be after timekeeping is initialized */
993 	random_init();
994 
995 	/* These make use of the fully initialized rng */
996 	kfence_init();
997 	boot_init_stack_canary();
998 
999 	perf_event_init();
1000 	profile_init();
1001 	call_function_init();
1002 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1003 
1004 	early_boot_irqs_disabled = false;
1005 	local_irq_enable();
1006 
1007 	kmem_cache_init_late();
1008 
1009 	/*
1010 	 * HACK ALERT! This is early. We're enabling the console before
1011 	 * we've done PCI setups etc, and console_init() must be aware of
1012 	 * this. But we do want output early, in case something goes wrong.
1013 	 */
1014 	console_init();
1015 	if (panic_later)
1016 		panic("Too many boot %s vars at `%s'", panic_later,
1017 		      panic_param);
1018 
1019 	lockdep_init();
1020 
1021 	/*
1022 	 * Need to run this when irqs are enabled, because it wants
1023 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1024 	 * too:
1025 	 */
1026 	locking_selftest();
1027 
1028 #ifdef CONFIG_BLK_DEV_INITRD
1029 	if (initrd_start && !initrd_below_start_ok &&
1030 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1031 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1032 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1033 		    min_low_pfn);
1034 		initrd_start = 0;
1035 	}
1036 #endif
1037 	setup_per_cpu_pageset();
1038 	numa_policy_init();
1039 	acpi_early_init();
1040 	if (late_time_init)
1041 		late_time_init();
1042 	sched_clock_init();
1043 	calibrate_delay();
1044 
1045 	arch_cpu_finalize_init();
1046 
1047 	pid_idr_init();
1048 	anon_vma_init();
1049 #ifdef CONFIG_X86
1050 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1051 		efi_enter_virtual_mode();
1052 #endif
1053 	thread_stack_cache_init();
1054 	cred_init();
1055 	fork_init();
1056 	proc_caches_init();
1057 	uts_ns_init();
1058 	key_init();
1059 	security_init();
1060 	dbg_late_init();
1061 	net_ns_init();
1062 	vfs_caches_init();
1063 	pagecache_init();
1064 	signals_init();
1065 	seq_file_init();
1066 	proc_root_init();
1067 	nsfs_init();
1068 	pidfs_init();
1069 	cpuset_init();
1070 	cgroup_init();
1071 	taskstats_init_early();
1072 	delayacct_init();
1073 
1074 	acpi_subsystem_init();
1075 	arch_post_acpi_subsys_init();
1076 	kcsan_init();
1077 
1078 	/* Do the rest non-__init'ed, we're now alive */
1079 	rest_init();
1080 
1081 	/*
1082 	 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1083 	 * and older.
1084 	 */
1085 #if !__has_attribute(__no_stack_protector__)
1086 	prevent_tail_call_optimization();
1087 #endif
1088 }
1089 
1090 /* Call all constructor functions linked into the kernel. */
1091 static void __init do_ctors(void)
1092 {
1093 /*
1094  * For UML, the constructors have already been called by the
1095  * normal setup code as it's just a normal ELF binary, so we
1096  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1097  * even on UML for modules.
1098  */
1099 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1100 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1101 
1102 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1103 		(*fn)();
1104 #endif
1105 }
1106 
1107 #ifdef CONFIG_KALLSYMS
1108 struct blacklist_entry {
1109 	struct list_head next;
1110 	char *buf;
1111 };
1112 
1113 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1114 
1115 static int __init initcall_blacklist(char *str)
1116 {
1117 	char *str_entry;
1118 	struct blacklist_entry *entry;
1119 
1120 	/* str argument is a comma-separated list of functions */
1121 	do {
1122 		str_entry = strsep(&str, ",");
1123 		if (str_entry) {
1124 			pr_debug("blacklisting initcall %s\n", str_entry);
1125 			entry = memblock_alloc(sizeof(*entry),
1126 					       SMP_CACHE_BYTES);
1127 			if (!entry)
1128 				panic("%s: Failed to allocate %zu bytes\n",
1129 				      __func__, sizeof(*entry));
1130 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1131 						    SMP_CACHE_BYTES);
1132 			if (!entry->buf)
1133 				panic("%s: Failed to allocate %zu bytes\n",
1134 				      __func__, strlen(str_entry) + 1);
1135 			strcpy(entry->buf, str_entry);
1136 			list_add(&entry->next, &blacklisted_initcalls);
1137 		}
1138 	} while (str_entry);
1139 
1140 	return 1;
1141 }
1142 
1143 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1144 {
1145 	struct blacklist_entry *entry;
1146 	char fn_name[KSYM_SYMBOL_LEN];
1147 	unsigned long addr;
1148 
1149 	if (list_empty(&blacklisted_initcalls))
1150 		return false;
1151 
1152 	addr = (unsigned long) dereference_function_descriptor(fn);
1153 	sprint_symbol_no_offset(fn_name, addr);
1154 
1155 	/*
1156 	 * fn will be "function_name [module_name]" where [module_name] is not
1157 	 * displayed for built-in init functions.  Strip off the [module_name].
1158 	 */
1159 	strreplace(fn_name, ' ', '\0');
1160 
1161 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1162 		if (!strcmp(fn_name, entry->buf)) {
1163 			pr_debug("initcall %s blacklisted\n", fn_name);
1164 			return true;
1165 		}
1166 	}
1167 
1168 	return false;
1169 }
1170 #else
1171 static int __init initcall_blacklist(char *str)
1172 {
1173 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1174 	return 0;
1175 }
1176 
1177 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1178 {
1179 	return false;
1180 }
1181 #endif
1182 __setup("initcall_blacklist=", initcall_blacklist);
1183 
1184 static __init_or_module void
1185 trace_initcall_start_cb(void *data, initcall_t fn)
1186 {
1187 	ktime_t *calltime = data;
1188 
1189 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1190 	*calltime = ktime_get();
1191 }
1192 
1193 static __init_or_module void
1194 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1195 {
1196 	ktime_t rettime, *calltime = data;
1197 
1198 	rettime = ktime_get();
1199 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1200 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1201 }
1202 
1203 static ktime_t initcall_calltime;
1204 
1205 #ifdef TRACEPOINTS_ENABLED
1206 static void __init initcall_debug_enable(void)
1207 {
1208 	int ret;
1209 
1210 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1211 					    &initcall_calltime);
1212 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1213 					      &initcall_calltime);
1214 	WARN(ret, "Failed to register initcall tracepoints\n");
1215 }
1216 # define do_trace_initcall_start	trace_initcall_start
1217 # define do_trace_initcall_finish	trace_initcall_finish
1218 #else
1219 static inline void do_trace_initcall_start(initcall_t fn)
1220 {
1221 	if (!initcall_debug)
1222 		return;
1223 	trace_initcall_start_cb(&initcall_calltime, fn);
1224 }
1225 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1226 {
1227 	if (!initcall_debug)
1228 		return;
1229 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1230 }
1231 #endif /* !TRACEPOINTS_ENABLED */
1232 
1233 int __init_or_module do_one_initcall(initcall_t fn)
1234 {
1235 	int count = preempt_count();
1236 	char msgbuf[64];
1237 	int ret;
1238 
1239 	if (initcall_blacklisted(fn))
1240 		return -EPERM;
1241 
1242 	do_trace_initcall_start(fn);
1243 	ret = fn();
1244 	do_trace_initcall_finish(fn, ret);
1245 
1246 	msgbuf[0] = 0;
1247 
1248 	if (preempt_count() != count) {
1249 		sprintf(msgbuf, "preemption imbalance ");
1250 		preempt_count_set(count);
1251 	}
1252 	if (irqs_disabled()) {
1253 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1254 		local_irq_enable();
1255 	}
1256 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1257 
1258 	add_latent_entropy();
1259 	return ret;
1260 }
1261 
1262 
1263 static initcall_entry_t *initcall_levels[] __initdata = {
1264 	__initcall0_start,
1265 	__initcall1_start,
1266 	__initcall2_start,
1267 	__initcall3_start,
1268 	__initcall4_start,
1269 	__initcall5_start,
1270 	__initcall6_start,
1271 	__initcall7_start,
1272 	__initcall_end,
1273 };
1274 
1275 /* Keep these in sync with initcalls in include/linux/init.h */
1276 static const char *initcall_level_names[] __initdata = {
1277 	"pure",
1278 	"core",
1279 	"postcore",
1280 	"arch",
1281 	"subsys",
1282 	"fs",
1283 	"device",
1284 	"late",
1285 };
1286 
1287 static int __init ignore_unknown_bootoption(char *param, char *val,
1288 			       const char *unused, void *arg)
1289 {
1290 	return 0;
1291 }
1292 
1293 static void __init do_initcall_level(int level, char *command_line)
1294 {
1295 	initcall_entry_t *fn;
1296 
1297 	parse_args(initcall_level_names[level],
1298 		   command_line, __start___param,
1299 		   __stop___param - __start___param,
1300 		   level, level,
1301 		   NULL, ignore_unknown_bootoption);
1302 
1303 	trace_initcall_level(initcall_level_names[level]);
1304 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1305 		do_one_initcall(initcall_from_entry(fn));
1306 }
1307 
1308 static void __init do_initcalls(void)
1309 {
1310 	int level;
1311 	size_t len = saved_command_line_len + 1;
1312 	char *command_line;
1313 
1314 	command_line = kzalloc(len, GFP_KERNEL);
1315 	if (!command_line)
1316 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1317 
1318 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1319 		/* Parser modifies command_line, restore it each time */
1320 		strcpy(command_line, saved_command_line);
1321 		do_initcall_level(level, command_line);
1322 	}
1323 
1324 	kfree(command_line);
1325 }
1326 
1327 /*
1328  * Ok, the machine is now initialized. None of the devices
1329  * have been touched yet, but the CPU subsystem is up and
1330  * running, and memory and process management works.
1331  *
1332  * Now we can finally start doing some real work..
1333  */
1334 static void __init do_basic_setup(void)
1335 {
1336 	cpuset_init_smp();
1337 	driver_init();
1338 	init_irq_proc();
1339 	do_ctors();
1340 	do_initcalls();
1341 }
1342 
1343 static void __init do_pre_smp_initcalls(void)
1344 {
1345 	initcall_entry_t *fn;
1346 
1347 	trace_initcall_level("early");
1348 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1349 		do_one_initcall(initcall_from_entry(fn));
1350 }
1351 
1352 static int run_init_process(const char *init_filename)
1353 {
1354 	const char *const *p;
1355 
1356 	argv_init[0] = init_filename;
1357 	pr_info("Run %s as init process\n", init_filename);
1358 	pr_debug("  with arguments:\n");
1359 	for (p = argv_init; *p; p++)
1360 		pr_debug("    %s\n", *p);
1361 	pr_debug("  with environment:\n");
1362 	for (p = envp_init; *p; p++)
1363 		pr_debug("    %s\n", *p);
1364 	return kernel_execve(init_filename, argv_init, envp_init);
1365 }
1366 
1367 static int try_to_run_init_process(const char *init_filename)
1368 {
1369 	int ret;
1370 
1371 	ret = run_init_process(init_filename);
1372 
1373 	if (ret && ret != -ENOENT) {
1374 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1375 		       init_filename, ret);
1376 	}
1377 
1378 	return ret;
1379 }
1380 
1381 static noinline void __init kernel_init_freeable(void);
1382 
1383 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1384 bool rodata_enabled __ro_after_init = true;
1385 
1386 #ifndef arch_parse_debug_rodata
1387 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1388 #endif
1389 
1390 static int __init set_debug_rodata(char *str)
1391 {
1392 	if (arch_parse_debug_rodata(str))
1393 		return 0;
1394 
1395 	if (str && !strcmp(str, "on"))
1396 		rodata_enabled = true;
1397 	else if (str && !strcmp(str, "off"))
1398 		rodata_enabled = false;
1399 	else
1400 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1401 	return 0;
1402 }
1403 early_param("rodata", set_debug_rodata);
1404 #endif
1405 
1406 static void mark_readonly(void)
1407 {
1408 	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1409 		/*
1410 		 * load_module() results in W+X mappings, which are cleaned
1411 		 * up with init_free_wq. Let's make sure that queued work is
1412 		 * flushed so that we don't hit false positives looking for
1413 		 * insecure pages which are W+X.
1414 		 */
1415 		flush_module_init_free_work();
1416 		mark_rodata_ro();
1417 		debug_checkwx();
1418 		rodata_test();
1419 	} else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1420 		pr_info("Kernel memory protection disabled.\n");
1421 	} else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1422 		pr_warn("Kernel memory protection not selected by kernel config.\n");
1423 	} else {
1424 		pr_warn("This architecture does not have kernel memory protection.\n");
1425 	}
1426 }
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 	async_init();
1550 	padata_init();
1551 	page_alloc_init_late();
1552 
1553 	do_basic_setup();
1554 
1555 	kunit_run_all_tests();
1556 
1557 	wait_for_initramfs();
1558 	console_on_rootfs();
1559 
1560 	/*
1561 	 * check if there is an early userspace init.  If yes, let it do all
1562 	 * the work
1563 	 */
1564 	if (init_eaccess(ramdisk_execute_command) != 0) {
1565 		ramdisk_execute_command = NULL;
1566 		prepare_namespace();
1567 	}
1568 
1569 	/*
1570 	 * Ok, we have completed the initial bootup, and
1571 	 * we're essentially up and running. Get rid of the
1572 	 * initmem segments and start the user-mode stuff..
1573 	 *
1574 	 * rootfs is available now, try loading the public keys
1575 	 * and default modules
1576 	 */
1577 
1578 	integrity_load_keys();
1579 }
1580