xref: /linux/init/main.c (revision fa73514d5e620c2bb48b63e1dd509ce0942c39cf)
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/vmalloc.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/start_kernel.h>
40 #include <linux/security.h>
41 #include <linux/smp.h>
42 #include <linux/profile.h>
43 #include <linux/kfence.h>
44 #include <linux/rcupdate.h>
45 #include <linux/moduleparam.h>
46 #include <linux/kallsyms.h>
47 #include <linux/writeback.h>
48 #include <linux/cpu.h>
49 #include <linux/cpuset.h>
50 #include <linux/cgroup.h>
51 #include <linux/efi.h>
52 #include <linux/tick.h>
53 #include <linux/sched/isolation.h>
54 #include <linux/interrupt.h>
55 #include <linux/taskstats_kern.h>
56 #include <linux/delayacct.h>
57 #include <linux/unistd.h>
58 #include <linux/utsname.h>
59 #include <linux/rmap.h>
60 #include <linux/mempolicy.h>
61 #include <linux/key.h>
62 #include <linux/page_ext.h>
63 #include <linux/debug_locks.h>
64 #include <linux/debugobjects.h>
65 #include <linux/lockdep.h>
66 #include <linux/kmemleak.h>
67 #include <linux/padata.h>
68 #include <linux/pid_namespace.h>
69 #include <linux/device/driver.h>
70 #include <linux/kthread.h>
71 #include <linux/sched.h>
72 #include <linux/sched/init.h>
73 #include <linux/signal.h>
74 #include <linux/idr.h>
75 #include <linux/kgdb.h>
76 #include <linux/ftrace.h>
77 #include <linux/async.h>
78 #include <linux/shmem_fs.h>
79 #include <linux/slab.h>
80 #include <linux/perf_event.h>
81 #include <linux/ptrace.h>
82 #include <linux/pti.h>
83 #include <linux/blkdev.h>
84 #include <linux/elevator.h>
85 #include <linux/sched/clock.h>
86 #include <linux/sched/task.h>
87 #include <linux/sched/task_stack.h>
88 #include <linux/context_tracking.h>
89 #include <linux/random.h>
90 #include <linux/list.h>
91 #include <linux/integrity.h>
92 #include <linux/proc_ns.h>
93 #include <linux/io.h>
94 #include <linux/cache.h>
95 #include <linux/rodata_test.h>
96 #include <linux/jump_label.h>
97 #include <linux/mem_encrypt.h>
98 #include <linux/kcsan.h>
99 #include <linux/init_syscalls.h>
100 #include <linux/stackdepot.h>
101 
102 #include <asm/io.h>
103 #include <asm/bugs.h>
104 #include <asm/setup.h>
105 #include <asm/sections.h>
106 #include <asm/cacheflush.h>
107 
108 #define CREATE_TRACE_POINTS
109 #include <trace/events/initcall.h>
110 
111 #include <kunit/test.h>
112 
113 static int kernel_init(void *);
114 
115 extern void init_IRQ(void);
116 extern void radix_tree_init(void);
117 
118 /*
119  * Debug helper: via this flag we know that we are in 'early bootup code'
120  * where only the boot processor is running with IRQ disabled.  This means
121  * two things - IRQ must not be enabled before the flag is cleared and some
122  * operations which are not allowed with IRQ disabled are allowed while the
123  * flag is set.
124  */
125 bool early_boot_irqs_disabled __read_mostly;
126 
127 enum system_states system_state __read_mostly;
128 EXPORT_SYMBOL(system_state);
129 
130 /*
131  * Boot command-line arguments
132  */
133 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
134 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
135 
136 extern void time_init(void);
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;
144 /* Command line for parameter parsing */
145 static char *static_command_line;
146 /* Untouched extra command line */
147 static char *extra_command_line;
148 /* Extra init arguments */
149 static char *extra_init_args;
150 
151 #ifdef CONFIG_BOOT_CONFIG
152 /* Is bootconfig on command line? */
153 static bool bootconfig_found;
154 static bool initargs_found;
155 #else
156 # define bootconfig_found false
157 # define initargs_found false
158 #endif
159 
160 static char *execute_command;
161 static char *ramdisk_execute_command = "/init";
162 
163 /*
164  * Used to generate warnings if static_key manipulation functions are used
165  * before jump_label_init is called.
166  */
167 bool static_key_initialized __read_mostly;
168 EXPORT_SYMBOL_GPL(static_key_initialized);
169 
170 /*
171  * If set, this is an indication to the drivers that reset the underlying
172  * device before going ahead with the initialization otherwise driver might
173  * rely on the BIOS and skip the reset operation.
174  *
175  * This is useful if kernel is booting in an unreliable environment.
176  * For ex. kdump situation where previous kernel has crashed, BIOS has been
177  * skipped and devices will be in unknown state.
178  */
179 unsigned int reset_devices;
180 EXPORT_SYMBOL(reset_devices);
181 
182 static int __init set_reset_devices(char *str)
183 {
184 	reset_devices = 1;
185 	return 1;
186 }
187 
188 __setup("reset_devices", set_reset_devices);
189 
190 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
191 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
192 static const char *panic_later, *panic_param;
193 
194 extern const struct obs_kernel_param __setup_start[], __setup_end[];
195 
196 static bool __init obsolete_checksetup(char *line)
197 {
198 	const struct obs_kernel_param *p;
199 	bool had_early_param = false;
200 
201 	p = __setup_start;
202 	do {
203 		int n = strlen(p->str);
204 		if (parameqn(line, p->str, n)) {
205 			if (p->early) {
206 				/* Already done in parse_early_param?
207 				 * (Needs exact match on param part).
208 				 * Keep iterating, as we can have early
209 				 * params and __setups of same names 8( */
210 				if (line[n] == '\0' || line[n] == '=')
211 					had_early_param = true;
212 			} else if (!p->setup_func) {
213 				pr_warn("Parameter %s is obsolete, ignored\n",
214 					p->str);
215 				return true;
216 			} else if (p->setup_func(line + n))
217 				return true;
218 		}
219 		p++;
220 	} while (p < __setup_end);
221 
222 	return had_early_param;
223 }
224 
225 /*
226  * This should be approx 2 Bo*oMips to start (note initial shift), and will
227  * still work even if initially too large, it will just take slightly longer
228  */
229 unsigned long loops_per_jiffy = (1<<12);
230 EXPORT_SYMBOL(loops_per_jiffy);
231 
232 static int __init debug_kernel(char *str)
233 {
234 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
235 	return 0;
236 }
237 
238 static int __init quiet_kernel(char *str)
239 {
240 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
241 	return 0;
242 }
243 
244 early_param("debug", debug_kernel);
245 early_param("quiet", quiet_kernel);
246 
247 static int __init loglevel(char *str)
248 {
249 	int newlevel;
250 
251 	/*
252 	 * Only update loglevel value when a correct setting was passed,
253 	 * to prevent blind crashes (when loglevel being set to 0) that
254 	 * are quite hard to debug
255 	 */
256 	if (get_option(&str, &newlevel)) {
257 		console_loglevel = newlevel;
258 		return 0;
259 	}
260 
261 	return -EINVAL;
262 }
263 
264 early_param("loglevel", loglevel);
265 
266 #ifdef CONFIG_BLK_DEV_INITRD
267 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
268 {
269 	u32 size, csum;
270 	char *data;
271 	u32 *hdr;
272 	int i;
273 
274 	if (!initrd_end)
275 		return NULL;
276 
277 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
278 	/*
279 	 * Since Grub may align the size of initrd to 4, we must
280 	 * check the preceding 3 bytes as well.
281 	 */
282 	for (i = 0; i < 4; i++) {
283 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
284 			goto found;
285 		data--;
286 	}
287 	return NULL;
288 
289 found:
290 	hdr = (u32 *)(data - 8);
291 	size = le32_to_cpu(hdr[0]);
292 	csum = le32_to_cpu(hdr[1]);
293 
294 	data = ((void *)hdr) - size;
295 	if ((unsigned long)data < initrd_start) {
296 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
297 			size, initrd_end - initrd_start);
298 		return NULL;
299 	}
300 
301 	/* Remove bootconfig from initramfs/initrd */
302 	initrd_end = (unsigned long)data;
303 	if (_size)
304 		*_size = size;
305 	if (_csum)
306 		*_csum = csum;
307 
308 	return data;
309 }
310 #else
311 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
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 		return NULL;
384 	}
385 
386 	return new_cmdline;
387 }
388 
389 static int __init bootconfig_params(char *param, char *val,
390 				    const char *unused, void *arg)
391 {
392 	if (strcmp(param, "bootconfig") == 0) {
393 		bootconfig_found = true;
394 	}
395 	return 0;
396 }
397 
398 static void __init setup_boot_config(void)
399 {
400 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
401 	const char *msg;
402 	int pos;
403 	u32 size, csum;
404 	char *data, *copy, *err;
405 	int ret;
406 
407 	/* Cut out the bootconfig data even if we have no bootconfig option */
408 	data = get_boot_config_from_initrd(&size, &csum);
409 
410 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
411 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
412 			 bootconfig_params);
413 
414 	if (IS_ERR(err) || !bootconfig_found)
415 		return;
416 
417 	/* parse_args() stops at '--' and returns an address */
418 	if (err)
419 		initargs_found = true;
420 
421 	if (!data) {
422 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
423 		return;
424 	}
425 
426 	if (size >= XBC_DATA_MAX) {
427 		pr_err("bootconfig size %d greater than max size %d\n",
428 			size, XBC_DATA_MAX);
429 		return;
430 	}
431 
432 	if (xbc_calc_checksum(data, size) != csum) {
433 		pr_err("bootconfig checksum failed\n");
434 		return;
435 	}
436 
437 	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
438 	if (!copy) {
439 		pr_err("Failed to allocate memory for bootconfig\n");
440 		return;
441 	}
442 
443 	memcpy(copy, data, size);
444 	copy[size] = '\0';
445 
446 	ret = xbc_init(copy, &msg, &pos);
447 	if (ret < 0) {
448 		if (pos < 0)
449 			pr_err("Failed to init bootconfig: %s.\n", msg);
450 		else
451 			pr_err("Failed to parse bootconfig: %s at %d.\n",
452 				msg, pos);
453 	} else {
454 		pr_info("Load bootconfig: %d bytes %d nodes\n", 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 #else
464 
465 static void __init setup_boot_config(void)
466 {
467 	/* Remove bootconfig data from initrd */
468 	get_boot_config_from_initrd(NULL, NULL);
469 }
470 
471 static int __init warn_bootconfig(char *str)
472 {
473 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
474 	return 0;
475 }
476 early_param("bootconfig", warn_bootconfig);
477 
478 #endif
479 
480 /* Change NUL term back to "=", to make "param" the whole string. */
481 static void __init repair_env_string(char *param, char *val)
482 {
483 	if (val) {
484 		/* param=val or param="val"? */
485 		if (val == param+strlen(param)+1)
486 			val[-1] = '=';
487 		else if (val == param+strlen(param)+2) {
488 			val[-2] = '=';
489 			memmove(val-1, val, strlen(val)+1);
490 		} else
491 			BUG();
492 	}
493 }
494 
495 /* Anything after -- gets handed straight to init. */
496 static int __init set_init_arg(char *param, char *val,
497 			       const char *unused, void *arg)
498 {
499 	unsigned int i;
500 
501 	if (panic_later)
502 		return 0;
503 
504 	repair_env_string(param, val);
505 
506 	for (i = 0; argv_init[i]; i++) {
507 		if (i == MAX_INIT_ARGS) {
508 			panic_later = "init";
509 			panic_param = param;
510 			return 0;
511 		}
512 	}
513 	argv_init[i] = param;
514 	return 0;
515 }
516 
517 /*
518  * Unknown boot options get handed to init, unless they look like
519  * unused parameters (modprobe will find them in /proc/cmdline).
520  */
521 static int __init unknown_bootoption(char *param, char *val,
522 				     const char *unused, void *arg)
523 {
524 	size_t len = strlen(param);
525 
526 	repair_env_string(param, val);
527 
528 	/* Handle obsolete-style parameters */
529 	if (obsolete_checksetup(param))
530 		return 0;
531 
532 	/* Unused module parameter. */
533 	if (strnchr(param, len, '.'))
534 		return 0;
535 
536 	if (panic_later)
537 		return 0;
538 
539 	if (val) {
540 		/* Environment option */
541 		unsigned int i;
542 		for (i = 0; envp_init[i]; i++) {
543 			if (i == MAX_INIT_ENVS) {
544 				panic_later = "env";
545 				panic_param = param;
546 			}
547 			if (!strncmp(param, envp_init[i], len+1))
548 				break;
549 		}
550 		envp_init[i] = param;
551 	} else {
552 		/* Command line option */
553 		unsigned int i;
554 		for (i = 0; argv_init[i]; i++) {
555 			if (i == MAX_INIT_ARGS) {
556 				panic_later = "init";
557 				panic_param = param;
558 			}
559 		}
560 		argv_init[i] = param;
561 	}
562 	return 0;
563 }
564 
565 static int __init init_setup(char *str)
566 {
567 	unsigned int i;
568 
569 	execute_command = str;
570 	/*
571 	 * In case LILO is going to boot us with default command line,
572 	 * it prepends "auto" before the whole cmdline which makes
573 	 * the shell think it should execute a script with such name.
574 	 * So we ignore all arguments entered _before_ init=... [MJ]
575 	 */
576 	for (i = 1; i < MAX_INIT_ARGS; i++)
577 		argv_init[i] = NULL;
578 	return 1;
579 }
580 __setup("init=", init_setup);
581 
582 static int __init rdinit_setup(char *str)
583 {
584 	unsigned int i;
585 
586 	ramdisk_execute_command = str;
587 	/* See "auto" comment in init_setup */
588 	for (i = 1; i < MAX_INIT_ARGS; i++)
589 		argv_init[i] = NULL;
590 	return 1;
591 }
592 __setup("rdinit=", rdinit_setup);
593 
594 #ifndef CONFIG_SMP
595 static const unsigned int setup_max_cpus = NR_CPUS;
596 static inline void setup_nr_cpu_ids(void) { }
597 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
598 #endif
599 
600 /*
601  * We need to store the untouched command line for future reference.
602  * We also need to store the touched command line since the parameter
603  * parsing is performed in place, and we should allow a component to
604  * store reference of name/value for future reference.
605  */
606 static void __init setup_command_line(char *command_line)
607 {
608 	size_t len, xlen = 0, ilen = 0;
609 
610 	if (extra_command_line)
611 		xlen = strlen(extra_command_line);
612 	if (extra_init_args)
613 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
614 
615 	len = xlen + strlen(boot_command_line) + 1;
616 
617 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
618 	if (!saved_command_line)
619 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
620 
621 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
622 	if (!static_command_line)
623 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
624 
625 	if (xlen) {
626 		/*
627 		 * We have to put extra_command_line before boot command
628 		 * lines because there could be dashes (separator of init
629 		 * command line) in the command lines.
630 		 */
631 		strcpy(saved_command_line, extra_command_line);
632 		strcpy(static_command_line, extra_command_line);
633 	}
634 	strcpy(saved_command_line + xlen, boot_command_line);
635 	strcpy(static_command_line + xlen, command_line);
636 
637 	if (ilen) {
638 		/*
639 		 * Append supplemental init boot args to saved_command_line
640 		 * so that user can check what command line options passed
641 		 * to init.
642 		 */
643 		len = strlen(saved_command_line);
644 		if (initargs_found) {
645 			saved_command_line[len++] = ' ';
646 		} else {
647 			strcpy(saved_command_line + len, " -- ");
648 			len += 4;
649 		}
650 
651 		strcpy(saved_command_line + len, extra_init_args);
652 	}
653 }
654 
655 /*
656  * We need to finalize in a non-__init function or else race conditions
657  * between the root thread and the init thread may cause start_kernel to
658  * be reaped by free_initmem before the root thread has proceeded to
659  * cpu_idle.
660  *
661  * gcc-3.4 accidentally inlines this function, so use noinline.
662  */
663 
664 static __initdata DECLARE_COMPLETION(kthreadd_done);
665 
666 noinline void __ref rest_init(void)
667 {
668 	struct task_struct *tsk;
669 	int pid;
670 
671 	rcu_scheduler_starting();
672 	/*
673 	 * We need to spawn init first so that it obtains pid 1, however
674 	 * the init task will end up wanting to create kthreads, which, if
675 	 * we schedule it before we create kthreadd, will OOPS.
676 	 */
677 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
678 	/*
679 	 * Pin init on the boot CPU. Task migration is not properly working
680 	 * until sched_init_smp() has been run. It will set the allowed
681 	 * CPUs for init to the non isolated CPUs.
682 	 */
683 	rcu_read_lock();
684 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
685 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
686 	rcu_read_unlock();
687 
688 	numa_default_policy();
689 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
690 	rcu_read_lock();
691 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
692 	rcu_read_unlock();
693 
694 	/*
695 	 * Enable might_sleep() and smp_processor_id() checks.
696 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
697 	 * kernel_thread() would trigger might_sleep() splats. With
698 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
699 	 * already, but it's stuck on the kthreadd_done completion.
700 	 */
701 	system_state = SYSTEM_SCHEDULING;
702 
703 	complete(&kthreadd_done);
704 
705 	/*
706 	 * The boot idle thread must execute schedule()
707 	 * at least once to get things moving:
708 	 */
709 	schedule_preempt_disabled();
710 	/* Call into cpu_idle with preempt disabled */
711 	cpu_startup_entry(CPUHP_ONLINE);
712 }
713 
714 /* Check for early params. */
715 static int __init do_early_param(char *param, char *val,
716 				 const char *unused, void *arg)
717 {
718 	const struct obs_kernel_param *p;
719 
720 	for (p = __setup_start; p < __setup_end; p++) {
721 		if ((p->early && parameq(param, p->str)) ||
722 		    (strcmp(param, "console") == 0 &&
723 		     strcmp(p->str, "earlycon") == 0)
724 		) {
725 			if (p->setup_func(val) != 0)
726 				pr_warn("Malformed early option '%s'\n", param);
727 		}
728 	}
729 	/* We accept everything at this stage. */
730 	return 0;
731 }
732 
733 void __init parse_early_options(char *cmdline)
734 {
735 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
736 		   do_early_param);
737 }
738 
739 /* Arch code calls this early on, or if not, just before other parsing. */
740 void __init parse_early_param(void)
741 {
742 	static int done __initdata;
743 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
744 
745 	if (done)
746 		return;
747 
748 	/* All fall through to do_early_param. */
749 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
750 	parse_early_options(tmp_cmdline);
751 	done = 1;
752 }
753 
754 void __init __weak arch_post_acpi_subsys_init(void) { }
755 
756 void __init __weak smp_setup_processor_id(void)
757 {
758 }
759 
760 # if THREAD_SIZE >= PAGE_SIZE
761 void __init __weak thread_stack_cache_init(void)
762 {
763 }
764 #endif
765 
766 void __init __weak mem_encrypt_init(void) { }
767 
768 void __init __weak poking_init(void) { }
769 
770 void __init __weak pgtable_cache_init(void) { }
771 
772 bool initcall_debug;
773 core_param(initcall_debug, initcall_debug, bool, 0644);
774 
775 #ifdef TRACEPOINTS_ENABLED
776 static void __init initcall_debug_enable(void);
777 #else
778 static inline void initcall_debug_enable(void)
779 {
780 }
781 #endif
782 
783 /* Report memory auto-initialization states for this boot. */
784 static void __init report_meminit(void)
785 {
786 	const char *stack;
787 
788 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
789 		stack = "all(pattern)";
790 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
791 		stack = "all(zero)";
792 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
793 		stack = "byref_all(zero)";
794 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
795 		stack = "byref(zero)";
796 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
797 		stack = "__user(zero)";
798 	else
799 		stack = "off";
800 
801 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
802 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
803 		want_init_on_free() ? "on" : "off");
804 	if (want_init_on_free())
805 		pr_info("mem auto-init: clearing system memory may take some time...\n");
806 }
807 
808 /*
809  * Set up kernel memory allocators
810  */
811 static void __init mm_init(void)
812 {
813 	/*
814 	 * page_ext requires contiguous pages,
815 	 * bigger than MAX_ORDER unless SPARSEMEM.
816 	 */
817 	page_ext_init_flatmem();
818 	init_mem_debugging_and_hardening();
819 	kfence_alloc_pool();
820 	report_meminit();
821 	stack_depot_init();
822 	mem_init();
823 	mem_init_print_info();
824 	/* page_owner must be initialized after buddy is ready */
825 	page_ext_init_flatmem_late();
826 	kmem_cache_init();
827 	kmemleak_init();
828 	pgtable_init();
829 	debug_objects_mem_init();
830 	vmalloc_init();
831 	/* Should be run before the first non-init thread is created */
832 	init_espfix_bsp();
833 	/* Should be run after espfix64 is set up. */
834 	pti_init();
835 }
836 
837 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
838 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
839 			   randomize_kstack_offset);
840 DEFINE_PER_CPU(u32, kstack_offset);
841 
842 static int __init early_randomize_kstack_offset(char *buf)
843 {
844 	int ret;
845 	bool bool_result;
846 
847 	ret = kstrtobool(buf, &bool_result);
848 	if (ret)
849 		return ret;
850 
851 	if (bool_result)
852 		static_branch_enable(&randomize_kstack_offset);
853 	else
854 		static_branch_disable(&randomize_kstack_offset);
855 	return 0;
856 }
857 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
858 #endif
859 
860 void __init __weak arch_call_rest_init(void)
861 {
862 	rest_init();
863 }
864 
865 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
866 {
867 	char *command_line;
868 	char *after_dashes;
869 
870 	set_task_stack_end_magic(&init_task);
871 	smp_setup_processor_id();
872 	debug_objects_early_init();
873 
874 	cgroup_init_early();
875 
876 	local_irq_disable();
877 	early_boot_irqs_disabled = true;
878 
879 	/*
880 	 * Interrupts are still disabled. Do necessary setups, then
881 	 * enable them.
882 	 */
883 	boot_cpu_init();
884 	page_address_init();
885 	pr_notice("%s", linux_banner);
886 	early_security_init();
887 	setup_arch(&command_line);
888 	setup_boot_config();
889 	setup_command_line(command_line);
890 	setup_nr_cpu_ids();
891 	setup_per_cpu_areas();
892 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
893 	boot_cpu_hotplug_init();
894 
895 	build_all_zonelists(NULL);
896 	page_alloc_init();
897 
898 	pr_notice("Kernel command line: %s\n", saved_command_line);
899 	/* parameters may set static keys */
900 	jump_label_init();
901 	parse_early_param();
902 	after_dashes = parse_args("Booting kernel",
903 				  static_command_line, __start___param,
904 				  __stop___param - __start___param,
905 				  -1, -1, NULL, &unknown_bootoption);
906 	if (!IS_ERR_OR_NULL(after_dashes))
907 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
908 			   NULL, set_init_arg);
909 	if (extra_init_args)
910 		parse_args("Setting extra init args", extra_init_args,
911 			   NULL, 0, -1, -1, NULL, set_init_arg);
912 
913 	/*
914 	 * These use large bootmem allocations and must precede
915 	 * kmem_cache_init()
916 	 */
917 	setup_log_buf(0);
918 	vfs_caches_init_early();
919 	sort_main_extable();
920 	trap_init();
921 	mm_init();
922 
923 	ftrace_init();
924 
925 	/* trace_printk can be enabled here */
926 	early_trace_init();
927 
928 	/*
929 	 * Set up the scheduler prior starting any interrupts (such as the
930 	 * timer interrupt). Full topology setup happens at smp_init()
931 	 * time - but meanwhile we still have a functioning scheduler.
932 	 */
933 	sched_init();
934 	/*
935 	 * Disable preemption - early bootup scheduling is extremely
936 	 * fragile until we cpu_idle() for the first time.
937 	 */
938 	preempt_disable();
939 	if (WARN(!irqs_disabled(),
940 		 "Interrupts were enabled *very* early, fixing it\n"))
941 		local_irq_disable();
942 	radix_tree_init();
943 
944 	/*
945 	 * Set up housekeeping before setting up workqueues to allow the unbound
946 	 * workqueue to take non-housekeeping into account.
947 	 */
948 	housekeeping_init();
949 
950 	/*
951 	 * Allow workqueue creation and work item queueing/cancelling
952 	 * early.  Work item execution depends on kthreads and starts after
953 	 * workqueue_init().
954 	 */
955 	workqueue_init_early();
956 
957 	rcu_init();
958 
959 	/* Trace events are available after this */
960 	trace_init();
961 
962 	if (initcall_debug)
963 		initcall_debug_enable();
964 
965 	context_tracking_init();
966 	/* init some links before init_ISA_irqs() */
967 	early_irq_init();
968 	init_IRQ();
969 	tick_init();
970 	rcu_init_nohz();
971 	init_timers();
972 	hrtimers_init();
973 	softirq_init();
974 	timekeeping_init();
975 	kfence_init();
976 
977 	/*
978 	 * For best initial stack canary entropy, prepare it after:
979 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
980 	 * - timekeeping_init() for ktime entropy used in rand_initialize()
981 	 * - rand_initialize() to get any arch-specific entropy like RDRAND
982 	 * - add_latent_entropy() to get any latent entropy
983 	 * - adding command line entropy
984 	 */
985 	rand_initialize();
986 	add_latent_entropy();
987 	add_device_randomness(command_line, strlen(command_line));
988 	boot_init_stack_canary();
989 
990 	time_init();
991 	perf_event_init();
992 	profile_init();
993 	call_function_init();
994 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
995 
996 	early_boot_irqs_disabled = false;
997 	local_irq_enable();
998 
999 	kmem_cache_init_late();
1000 
1001 	/*
1002 	 * HACK ALERT! This is early. We're enabling the console before
1003 	 * we've done PCI setups etc, and console_init() must be aware of
1004 	 * this. But we do want output early, in case something goes wrong.
1005 	 */
1006 	console_init();
1007 	if (panic_later)
1008 		panic("Too many boot %s vars at `%s'", panic_later,
1009 		      panic_param);
1010 
1011 	lockdep_init();
1012 
1013 	/*
1014 	 * Need to run this when irqs are enabled, because it wants
1015 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1016 	 * too:
1017 	 */
1018 	locking_selftest();
1019 
1020 	/*
1021 	 * This needs to be called before any devices perform DMA
1022 	 * operations that might use the SWIOTLB bounce buffers. It will
1023 	 * mark the bounce buffers as decrypted so that their usage will
1024 	 * not cause "plain-text" data to be decrypted when accessed.
1025 	 */
1026 	mem_encrypt_init();
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 	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 	vfs_caches_init();
1059 	pagecache_init();
1060 	signals_init();
1061 	seq_file_init();
1062 	proc_root_init();
1063 	nsfs_init();
1064 	cpuset_init();
1065 	cgroup_init();
1066 	taskstats_init_early();
1067 	delayacct_init();
1068 
1069 	poking_init();
1070 	check_bugs();
1071 
1072 	acpi_subsystem_init();
1073 	arch_post_acpi_subsys_init();
1074 	kcsan_init();
1075 
1076 	/* Do the rest non-__init'ed, we're now alive */
1077 	arch_call_rest_init();
1078 
1079 	prevent_tail_call_optimization();
1080 }
1081 
1082 /* Call all constructor functions linked into the kernel. */
1083 static void __init do_ctors(void)
1084 {
1085 /*
1086  * For UML, the constructors have already been called by the
1087  * normal setup code as it's just a normal ELF binary, so we
1088  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1089  * even on UML for modules.
1090  */
1091 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1092 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1093 
1094 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1095 		(*fn)();
1096 #endif
1097 }
1098 
1099 #ifdef CONFIG_KALLSYMS
1100 struct blacklist_entry {
1101 	struct list_head next;
1102 	char *buf;
1103 };
1104 
1105 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1106 
1107 static int __init initcall_blacklist(char *str)
1108 {
1109 	char *str_entry;
1110 	struct blacklist_entry *entry;
1111 
1112 	/* str argument is a comma-separated list of functions */
1113 	do {
1114 		str_entry = strsep(&str, ",");
1115 		if (str_entry) {
1116 			pr_debug("blacklisting initcall %s\n", str_entry);
1117 			entry = memblock_alloc(sizeof(*entry),
1118 					       SMP_CACHE_BYTES);
1119 			if (!entry)
1120 				panic("%s: Failed to allocate %zu bytes\n",
1121 				      __func__, sizeof(*entry));
1122 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1123 						    SMP_CACHE_BYTES);
1124 			if (!entry->buf)
1125 				panic("%s: Failed to allocate %zu bytes\n",
1126 				      __func__, strlen(str_entry) + 1);
1127 			strcpy(entry->buf, str_entry);
1128 			list_add(&entry->next, &blacklisted_initcalls);
1129 		}
1130 	} while (str_entry);
1131 
1132 	return 0;
1133 }
1134 
1135 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1136 {
1137 	struct blacklist_entry *entry;
1138 	char fn_name[KSYM_SYMBOL_LEN];
1139 	unsigned long addr;
1140 
1141 	if (list_empty(&blacklisted_initcalls))
1142 		return false;
1143 
1144 	addr = (unsigned long) dereference_function_descriptor(fn);
1145 	sprint_symbol_no_offset(fn_name, addr);
1146 
1147 	/*
1148 	 * fn will be "function_name [module_name]" where [module_name] is not
1149 	 * displayed for built-in init functions.  Strip off the [module_name].
1150 	 */
1151 	strreplace(fn_name, ' ', '\0');
1152 
1153 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1154 		if (!strcmp(fn_name, entry->buf)) {
1155 			pr_debug("initcall %s blacklisted\n", fn_name);
1156 			return true;
1157 		}
1158 	}
1159 
1160 	return false;
1161 }
1162 #else
1163 static int __init initcall_blacklist(char *str)
1164 {
1165 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1166 	return 0;
1167 }
1168 
1169 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1170 {
1171 	return false;
1172 }
1173 #endif
1174 __setup("initcall_blacklist=", initcall_blacklist);
1175 
1176 static __init_or_module void
1177 trace_initcall_start_cb(void *data, initcall_t fn)
1178 {
1179 	ktime_t *calltime = (ktime_t *)data;
1180 
1181 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1182 	*calltime = ktime_get();
1183 }
1184 
1185 static __init_or_module void
1186 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1187 {
1188 	ktime_t *calltime = (ktime_t *)data;
1189 	ktime_t delta, rettime;
1190 	unsigned long long duration;
1191 
1192 	rettime = ktime_get();
1193 	delta = ktime_sub(rettime, *calltime);
1194 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1195 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1196 		 fn, ret, duration);
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 extern initcall_entry_t __initcall_start[];
1260 extern initcall_entry_t __initcall0_start[];
1261 extern initcall_entry_t __initcall1_start[];
1262 extern initcall_entry_t __initcall2_start[];
1263 extern initcall_entry_t __initcall3_start[];
1264 extern initcall_entry_t __initcall4_start[];
1265 extern initcall_entry_t __initcall5_start[];
1266 extern initcall_entry_t __initcall6_start[];
1267 extern initcall_entry_t __initcall7_start[];
1268 extern initcall_entry_t __initcall_end[];
1269 
1270 static initcall_entry_t *initcall_levels[] __initdata = {
1271 	__initcall0_start,
1272 	__initcall1_start,
1273 	__initcall2_start,
1274 	__initcall3_start,
1275 	__initcall4_start,
1276 	__initcall5_start,
1277 	__initcall6_start,
1278 	__initcall7_start,
1279 	__initcall_end,
1280 };
1281 
1282 /* Keep these in sync with initcalls in include/linux/init.h */
1283 static const char *initcall_level_names[] __initdata = {
1284 	"pure",
1285 	"core",
1286 	"postcore",
1287 	"arch",
1288 	"subsys",
1289 	"fs",
1290 	"device",
1291 	"late",
1292 };
1293 
1294 static int __init ignore_unknown_bootoption(char *param, char *val,
1295 			       const char *unused, void *arg)
1296 {
1297 	return 0;
1298 }
1299 
1300 static void __init do_initcall_level(int level, char *command_line)
1301 {
1302 	initcall_entry_t *fn;
1303 
1304 	parse_args(initcall_level_names[level],
1305 		   command_line, __start___param,
1306 		   __stop___param - __start___param,
1307 		   level, level,
1308 		   NULL, ignore_unknown_bootoption);
1309 
1310 	trace_initcall_level(initcall_level_names[level]);
1311 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1312 		do_one_initcall(initcall_from_entry(fn));
1313 }
1314 
1315 static void __init do_initcalls(void)
1316 {
1317 	int level;
1318 	size_t len = strlen(saved_command_line) + 1;
1319 	char *command_line;
1320 
1321 	command_line = kzalloc(len, GFP_KERNEL);
1322 	if (!command_line)
1323 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1324 
1325 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1326 		/* Parser modifies command_line, restore it each time */
1327 		strcpy(command_line, saved_command_line);
1328 		do_initcall_level(level, command_line);
1329 	}
1330 
1331 	kfree(command_line);
1332 }
1333 
1334 /*
1335  * Ok, the machine is now initialized. None of the devices
1336  * have been touched yet, but the CPU subsystem is up and
1337  * running, and memory and process management works.
1338  *
1339  * Now we can finally start doing some real work..
1340  */
1341 static void __init do_basic_setup(void)
1342 {
1343 	cpuset_init_smp();
1344 	driver_init();
1345 	init_irq_proc();
1346 	do_ctors();
1347 	usermodehelper_enable();
1348 	do_initcalls();
1349 }
1350 
1351 static void __init do_pre_smp_initcalls(void)
1352 {
1353 	initcall_entry_t *fn;
1354 
1355 	trace_initcall_level("early");
1356 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1357 		do_one_initcall(initcall_from_entry(fn));
1358 }
1359 
1360 static int run_init_process(const char *init_filename)
1361 {
1362 	const char *const *p;
1363 
1364 	argv_init[0] = init_filename;
1365 	pr_info("Run %s as init process\n", init_filename);
1366 	pr_debug("  with arguments:\n");
1367 	for (p = argv_init; *p; p++)
1368 		pr_debug("    %s\n", *p);
1369 	pr_debug("  with environment:\n");
1370 	for (p = envp_init; *p; p++)
1371 		pr_debug("    %s\n", *p);
1372 	return kernel_execve(init_filename, argv_init, envp_init);
1373 }
1374 
1375 static int try_to_run_init_process(const char *init_filename)
1376 {
1377 	int ret;
1378 
1379 	ret = run_init_process(init_filename);
1380 
1381 	if (ret && ret != -ENOENT) {
1382 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1383 		       init_filename, ret);
1384 	}
1385 
1386 	return ret;
1387 }
1388 
1389 static noinline void __init kernel_init_freeable(void);
1390 
1391 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1392 bool rodata_enabled __ro_after_init = true;
1393 static int __init set_debug_rodata(char *str)
1394 {
1395 	return strtobool(str, &rodata_enabled);
1396 }
1397 __setup("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 call_rcu().  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 		rcu_barrier();
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 	kernel_init_freeable();
1438 	/* need to finish all async __init code before freeing the memory */
1439 	async_synchronize_full();
1440 	kprobe_free_init_mem();
1441 	ftrace_free_init_mem();
1442 	kgdb_free_init_mem();
1443 	free_initmem();
1444 	mark_readonly();
1445 
1446 	/*
1447 	 * Kernel mappings are now finalized - update the userspace page-table
1448 	 * to finalize PTI.
1449 	 */
1450 	pti_finalize();
1451 
1452 	system_state = SYSTEM_RUNNING;
1453 	numa_default_policy();
1454 
1455 	rcu_end_inkernel_boot();
1456 
1457 	do_sysctl_args();
1458 
1459 	if (ramdisk_execute_command) {
1460 		ret = run_init_process(ramdisk_execute_command);
1461 		if (!ret)
1462 			return 0;
1463 		pr_err("Failed to execute %s (error %d)\n",
1464 		       ramdisk_execute_command, ret);
1465 	}
1466 
1467 	/*
1468 	 * We try each of these until one succeeds.
1469 	 *
1470 	 * The Bourne shell can be used instead of init if we are
1471 	 * trying to recover a really broken machine.
1472 	 */
1473 	if (execute_command) {
1474 		ret = run_init_process(execute_command);
1475 		if (!ret)
1476 			return 0;
1477 		panic("Requested init %s failed (error %d).",
1478 		      execute_command, ret);
1479 	}
1480 
1481 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1482 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1483 		if (ret)
1484 			pr_err("Default init %s failed (error %d)\n",
1485 			       CONFIG_DEFAULT_INIT, ret);
1486 		else
1487 			return 0;
1488 	}
1489 
1490 	if (!try_to_run_init_process("/sbin/init") ||
1491 	    !try_to_run_init_process("/etc/init") ||
1492 	    !try_to_run_init_process("/bin/init") ||
1493 	    !try_to_run_init_process("/bin/sh"))
1494 		return 0;
1495 
1496 	panic("No working init found.  Try passing init= option to kernel. "
1497 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1498 }
1499 
1500 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1501 void __init console_on_rootfs(void)
1502 {
1503 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1504 
1505 	if (IS_ERR(file)) {
1506 		pr_err("Warning: unable to open an initial console.\n");
1507 		return;
1508 	}
1509 	init_dup(file);
1510 	init_dup(file);
1511 	init_dup(file);
1512 	fput(file);
1513 }
1514 
1515 static noinline void __init kernel_init_freeable(void)
1516 {
1517 	/*
1518 	 * Wait until kthreadd is all set-up.
1519 	 */
1520 	wait_for_completion(&kthreadd_done);
1521 
1522 	/* Now the scheduler is fully set up and can do blocking allocations */
1523 	gfp_allowed_mask = __GFP_BITS_MASK;
1524 
1525 	/*
1526 	 * init can allocate pages on any node
1527 	 */
1528 	set_mems_allowed(node_states[N_MEMORY]);
1529 
1530 	cad_pid = get_pid(task_pid(current));
1531 
1532 	smp_prepare_cpus(setup_max_cpus);
1533 
1534 	workqueue_init();
1535 
1536 	init_mm_internals();
1537 
1538 	rcu_init_tasks_generic();
1539 	do_pre_smp_initcalls();
1540 	lockup_detector_init();
1541 
1542 	smp_init();
1543 	sched_init_smp();
1544 
1545 	padata_init();
1546 	page_alloc_init_late();
1547 	/* Initialize page ext after all struct pages are initialized. */
1548 	page_ext_init();
1549 
1550 	do_basic_setup();
1551 
1552 	kunit_run_all_tests();
1553 
1554 	wait_for_initramfs();
1555 	console_on_rootfs();
1556 
1557 	/*
1558 	 * check if there is an early userspace init.  If yes, let it do all
1559 	 * the work
1560 	 */
1561 	if (init_eaccess(ramdisk_execute_command) != 0) {
1562 		ramdisk_execute_command = NULL;
1563 		prepare_namespace();
1564 	}
1565 
1566 	/*
1567 	 * Ok, we have completed the initial bootup, and
1568 	 * we're essentially up and running. Get rid of the
1569 	 * initmem segments and start the user-mode stuff..
1570 	 *
1571 	 * rootfs is available now, try loading the public keys
1572 	 * and default modules
1573 	 */
1574 
1575 	integrity_load_keys();
1576 }
1577