xref: /linux/init/main.c (revision 0471440c8061e9a7c0fd292c7c6598d6304efd53)
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/export.h>
17 #include <linux/extable.h>
18 #include <linux/module.h>
19 #include <linux/proc_fs.h>
20 #include <linux/binfmts.h>
21 #include <linux/kernel.h>
22 #include <linux/syscalls.h>
23 #include <linux/stackprotector.h>
24 #include <linux/string.h>
25 #include <linux/ctype.h>
26 #include <linux/delay.h>
27 #include <linux/ioport.h>
28 #include <linux/init.h>
29 #include <linux/initrd.h>
30 #include <linux/memblock.h>
31 #include <linux/acpi.h>
32 #include <linux/bootconfig.h>
33 #include <linux/console.h>
34 #include <linux/nmi.h>
35 #include <linux/percpu.h>
36 #include <linux/kmod.h>
37 #include <linux/kprobes.h>
38 #include <linux/kmsan.h>
39 #include <linux/vmalloc.h>
40 #include <linux/kernel_stat.h>
41 #include <linux/start_kernel.h>
42 #include <linux/security.h>
43 #include <linux/smp.h>
44 #include <linux/profile.h>
45 #include <linux/kfence.h>
46 #include <linux/rcupdate.h>
47 #include <linux/srcu.h>
48 #include <linux/moduleparam.h>
49 #include <linux/kallsyms.h>
50 #include <linux/buildid.h>
51 #include <linux/writeback.h>
52 #include <linux/cpu.h>
53 #include <linux/cpuset.h>
54 #include <linux/memcontrol.h>
55 #include <linux/cgroup.h>
56 #include <linux/tick.h>
57 #include <linux/sched/isolation.h>
58 #include <linux/interrupt.h>
59 #include <linux/taskstats_kern.h>
60 #include <linux/delayacct.h>
61 #include <linux/unistd.h>
62 #include <linux/utsname.h>
63 #include <linux/rmap.h>
64 #include <linux/mempolicy.h>
65 #include <linux/key.h>
66 #include <linux/debug_locks.h>
67 #include <linux/debugobjects.h>
68 #include <linux/lockdep.h>
69 #include <linux/kmemleak.h>
70 #include <linux/padata.h>
71 #include <linux/pid_namespace.h>
72 #include <linux/device/driver.h>
73 #include <linux/kthread.h>
74 #include <linux/sched.h>
75 #include <linux/sched/init.h>
76 #include <linux/signal.h>
77 #include <linux/idr.h>
78 #include <linux/kgdb.h>
79 #include <linux/ftrace.h>
80 #include <linux/async.h>
81 #include <linux/shmem_fs.h>
82 #include <linux/slab.h>
83 #include <linux/perf_event.h>
84 #include <linux/ptrace.h>
85 #include <linux/pti.h>
86 #include <linux/blkdev.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/moduleloader.h>
93 #include <linux/list.h>
94 #include <linux/integrity.h>
95 #include <linux/proc_ns.h>
96 #include <linux/io.h>
97 #include <linux/cache.h>
98 #include <linux/rodata_test.h>
99 #include <linux/jump_label.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
103 #include <linux/randomize_kstack.h>
104 #include <linux/pidfs.h>
105 #include <linux/ptdump.h>
106 #include <net/net_namespace.h>
107 
108 #include <asm/io.h>
109 #include <asm/setup.h>
110 #include <asm/sections.h>
111 #include <asm/cacheflush.h>
112 
113 #define CREATE_TRACE_POINTS
114 #include <trace/events/initcall.h>
115 
116 #include <kunit/test.h>
117 
118 static int kernel_init(void *);
119 
120 /*
121  * Debug helper: via this flag we know that we are in 'early bootup code'
122  * where only the boot processor is running with IRQ disabled.  This means
123  * two things - IRQ must not be enabled before the flag is cleared and some
124  * operations which are not allowed with IRQ disabled are allowed while the
125  * flag is set.
126  */
127 bool early_boot_irqs_disabled __read_mostly;
128 
129 enum system_states system_state __read_mostly;
130 EXPORT_SYMBOL(system_state);
131 
132 /*
133  * Boot command-line arguments
134  */
135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
137 
138 /* Default late time init is NULL. archs can override this later. */
139 void (*__initdata late_time_init)(void);
140 
141 /* Untouched command line saved by arch-specific code. */
142 char __initdata boot_command_line[COMMAND_LINE_SIZE];
143 /* Untouched saved command line (eg. for /proc) */
144 char *saved_command_line __ro_after_init;
145 unsigned int saved_command_line_len __ro_after_init;
146 /* Command line for parameter parsing */
147 static char *static_command_line;
148 /* Untouched extra command line */
149 static char *extra_command_line;
150 /* Extra init arguments */
151 static char *extra_init_args;
152 
153 #ifdef CONFIG_BOOT_CONFIG
154 /* Is bootconfig on command line? */
155 static bool bootconfig_found;
156 static size_t initargs_offs;
157 #else
158 # define bootconfig_found false
159 # define initargs_offs 0
160 #endif
161 
162 static char *execute_command;
163 static char *ramdisk_execute_command = "/init";
164 
165 /*
166  * Used to generate warnings if static_key manipulation functions are used
167  * before jump_label_init is called.
168  */
169 bool static_key_initialized __read_mostly;
170 EXPORT_SYMBOL_GPL(static_key_initialized);
171 
172 /*
173  * If set, this is an indication to the drivers that reset the underlying
174  * device before going ahead with the initialization otherwise driver might
175  * rely on the BIOS and skip the reset operation.
176  *
177  * This is useful if kernel is booting in an unreliable environment.
178  * For ex. kdump situation where previous kernel has crashed, BIOS has been
179  * skipped and devices will be in unknown state.
180  */
181 unsigned int reset_devices;
182 EXPORT_SYMBOL(reset_devices);
183 
184 static int __init set_reset_devices(char *str)
185 {
186 	reset_devices = 1;
187 	return 1;
188 }
189 
190 __setup("reset_devices", set_reset_devices);
191 
192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194 static const char *panic_later, *panic_param;
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(size_t *_size)
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 	if (xbc_calc_checksum(data, size) != csum) {
302 		pr_err("bootconfig checksum failed\n");
303 		return NULL;
304 	}
305 
306 	/* Remove bootconfig from initramfs/initrd */
307 	initrd_end = (unsigned long)data;
308 	if (_size)
309 		*_size = size;
310 
311 	return data;
312 }
313 #else
314 static void * __init get_boot_config_from_initrd(size_t *_size)
315 {
316 	return NULL;
317 }
318 #endif
319 
320 #ifdef CONFIG_BOOT_CONFIG
321 
322 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
323 
324 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
325 
326 static int __init xbc_snprint_cmdline(char *buf, size_t size,
327 				      struct xbc_node *root)
328 {
329 	struct xbc_node *knode, *vnode;
330 	char *end = buf + size;
331 	const char *val, *q;
332 	int ret;
333 
334 	xbc_node_for_each_key_value(root, knode, val) {
335 		ret = xbc_node_compose_key_after(root, knode,
336 					xbc_namebuf, XBC_KEYLEN_MAX);
337 		if (ret < 0)
338 			return ret;
339 
340 		vnode = xbc_node_get_child(knode);
341 		if (!vnode) {
342 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
343 			if (ret < 0)
344 				return ret;
345 			buf += ret;
346 			continue;
347 		}
348 		xbc_array_for_each_value(vnode, val) {
349 			/*
350 			 * For prettier and more readable /proc/cmdline, only
351 			 * quote the value when necessary, i.e. when it contains
352 			 * whitespace.
353 			 */
354 			q = strpbrk(val, " \t\r\n") ? "\"" : "";
355 			ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
356 				       xbc_namebuf, q, val, q);
357 			if (ret < 0)
358 				return ret;
359 			buf += ret;
360 		}
361 	}
362 
363 	return buf - (end - size);
364 }
365 #undef rest
366 
367 /* Make an extra command line under given key word */
368 static char * __init xbc_make_cmdline(const char *key)
369 {
370 	struct xbc_node *root;
371 	char *new_cmdline;
372 	int ret, len = 0;
373 
374 	root = xbc_find_node(key);
375 	if (!root)
376 		return NULL;
377 
378 	/* Count required buffer size */
379 	len = xbc_snprint_cmdline(NULL, 0, root);
380 	if (len <= 0)
381 		return NULL;
382 
383 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
384 	if (!new_cmdline) {
385 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
386 		return NULL;
387 	}
388 
389 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
390 	if (ret < 0 || ret > len) {
391 		pr_err("Failed to print extra kernel cmdline.\n");
392 		memblock_free(new_cmdline, len + 1);
393 		return NULL;
394 	}
395 
396 	return new_cmdline;
397 }
398 
399 static int __init bootconfig_params(char *param, char *val,
400 				    const char *unused, void *arg)
401 {
402 	if (strcmp(param, "bootconfig") == 0) {
403 		bootconfig_found = true;
404 	}
405 	return 0;
406 }
407 
408 static int __init warn_bootconfig(char *str)
409 {
410 	/* The 'bootconfig' has been handled by bootconfig_params(). */
411 	return 0;
412 }
413 
414 static void __init setup_boot_config(void)
415 {
416 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
417 	const char *msg, *data;
418 	int pos, ret;
419 	size_t size;
420 	char *err;
421 
422 	/* Cut out the bootconfig data even if we have no bootconfig option */
423 	data = get_boot_config_from_initrd(&size);
424 	/* If there is no bootconfig in initrd, try embedded one. */
425 	if (!data)
426 		data = xbc_get_embedded_bootconfig(&size);
427 
428 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
429 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
430 			 bootconfig_params);
431 
432 	if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
433 		return;
434 
435 	/* parse_args() stops at the next param of '--' and returns an address */
436 	if (err)
437 		initargs_offs = err - tmp_cmdline;
438 
439 	if (!data) {
440 		/* If user intended to use bootconfig, show an error level message */
441 		if (bootconfig_found)
442 			pr_err("'bootconfig' found on command line, but no bootconfig found\n");
443 		else
444 			pr_info("No bootconfig data provided, so skipping bootconfig");
445 		return;
446 	}
447 
448 	if (size >= XBC_DATA_MAX) {
449 		pr_err("bootconfig size %ld greater than max size %d\n",
450 			(long)size, XBC_DATA_MAX);
451 		return;
452 	}
453 
454 	ret = xbc_init(data, size, &msg, &pos);
455 	if (ret < 0) {
456 		if (pos < 0)
457 			pr_err("Failed to init bootconfig: %s.\n", msg);
458 		else
459 			pr_err("Failed to parse bootconfig: %s at %d.\n",
460 				msg, pos);
461 	} else {
462 		xbc_get_info(&ret, NULL);
463 		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
464 		/* keys starting with "kernel." are passed via cmdline */
465 		extra_command_line = xbc_make_cmdline("kernel");
466 		/* Also, "init." keys are init arguments */
467 		extra_init_args = xbc_make_cmdline("init");
468 	}
469 	return;
470 }
471 
472 static void __init exit_boot_config(void)
473 {
474 	xbc_exit();
475 }
476 
477 #else	/* !CONFIG_BOOT_CONFIG */
478 
479 static void __init setup_boot_config(void)
480 {
481 	/* Remove bootconfig data from initrd */
482 	get_boot_config_from_initrd(NULL);
483 }
484 
485 static int __init warn_bootconfig(char *str)
486 {
487 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
488 	return 0;
489 }
490 
491 #define exit_boot_config()	do {} while (0)
492 
493 #endif	/* CONFIG_BOOT_CONFIG */
494 
495 early_param("bootconfig", warn_bootconfig);
496 
497 bool __init cmdline_has_extra_options(void)
498 {
499 	return extra_command_line || extra_init_args;
500 }
501 
502 /* Change NUL term back to "=", to make "param" the whole string. */
503 static void __init repair_env_string(char *param, char *val)
504 {
505 	if (val) {
506 		/* param=val or param="val"? */
507 		if (val == param+strlen(param)+1)
508 			val[-1] = '=';
509 		else if (val == param+strlen(param)+2) {
510 			val[-2] = '=';
511 			memmove(val-1, val, strlen(val)+1);
512 		} else
513 			BUG();
514 	}
515 }
516 
517 /* Anything after -- gets handed straight to init. */
518 static int __init set_init_arg(char *param, char *val,
519 			       const char *unused, void *arg)
520 {
521 	unsigned int i;
522 
523 	if (panic_later)
524 		return 0;
525 
526 	repair_env_string(param, val);
527 
528 	for (i = 0; argv_init[i]; i++) {
529 		if (i == MAX_INIT_ARGS) {
530 			panic_later = "init";
531 			panic_param = param;
532 			return 0;
533 		}
534 	}
535 	argv_init[i] = param;
536 	return 0;
537 }
538 
539 /*
540  * Unknown boot options get handed to init, unless they look like
541  * unused parameters (modprobe will find them in /proc/cmdline).
542  */
543 static int __init unknown_bootoption(char *param, char *val,
544 				     const char *unused, void *arg)
545 {
546 	size_t len = strlen(param);
547 	/*
548 	 * Well-known bootloader identifiers:
549 	 * 1. LILO/Grub pass "BOOT_IMAGE=...";
550 	 * 2. kexec/kdump (kexec-tools) pass "kexec".
551 	 */
552 	const char *bootloader[] = { "BOOT_IMAGE=", "kexec", NULL };
553 
554 	/* Handle params aliased to sysctls */
555 	if (sysctl_is_alias(param))
556 		return 0;
557 
558 	repair_env_string(param, val);
559 
560 	/* Handle bootloader identifier */
561 	for (int i = 0; bootloader[i]; i++) {
562 		if (strstarts(param, bootloader[i]))
563 			return 0;
564 	}
565 
566 	/* Handle obsolete-style parameters */
567 	if (obsolete_checksetup(param))
568 		return 0;
569 
570 	/* Unused module parameter. */
571 	if (strnchr(param, len, '.'))
572 		return 0;
573 
574 	if (panic_later)
575 		return 0;
576 
577 	if (val) {
578 		/* Environment option */
579 		unsigned int i;
580 		for (i = 0; envp_init[i]; i++) {
581 			if (i == MAX_INIT_ENVS) {
582 				panic_later = "env";
583 				panic_param = param;
584 			}
585 			if (!strncmp(param, envp_init[i], len+1))
586 				break;
587 		}
588 		envp_init[i] = param;
589 	} else {
590 		/* Command line option */
591 		unsigned int i;
592 		for (i = 0; argv_init[i]; i++) {
593 			if (i == MAX_INIT_ARGS) {
594 				panic_later = "init";
595 				panic_param = param;
596 			}
597 		}
598 		argv_init[i] = param;
599 	}
600 	return 0;
601 }
602 
603 static int __init init_setup(char *str)
604 {
605 	unsigned int i;
606 
607 	execute_command = str;
608 	/*
609 	 * In case LILO is going to boot us with default command line,
610 	 * it prepends "auto" before the whole cmdline which makes
611 	 * the shell think it should execute a script with such name.
612 	 * So we ignore all arguments entered _before_ init=... [MJ]
613 	 */
614 	for (i = 1; i < MAX_INIT_ARGS; i++)
615 		argv_init[i] = NULL;
616 	return 1;
617 }
618 __setup("init=", init_setup);
619 
620 static int __init rdinit_setup(char *str)
621 {
622 	unsigned int i;
623 
624 	ramdisk_execute_command = str;
625 	/* See "auto" comment in init_setup */
626 	for (i = 1; i < MAX_INIT_ARGS; i++)
627 		argv_init[i] = NULL;
628 	return 1;
629 }
630 __setup("rdinit=", rdinit_setup);
631 
632 #ifndef CONFIG_SMP
633 static inline void setup_nr_cpu_ids(void) { }
634 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
635 #endif
636 
637 /*
638  * We need to store the untouched command line for future reference.
639  * We also need to store the touched command line since the parameter
640  * parsing is performed in place, and we should allow a component to
641  * store reference of name/value for future reference.
642  */
643 static void __init setup_command_line(char *command_line)
644 {
645 	size_t len, xlen = 0, ilen = 0;
646 
647 	if (extra_command_line)
648 		xlen = strlen(extra_command_line);
649 	if (extra_init_args) {
650 		extra_init_args = strim(extra_init_args); /* remove trailing space */
651 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
652 	}
653 
654 	len = xlen + strlen(boot_command_line) + ilen + 1;
655 
656 	saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
657 
658 	len = xlen + strlen(command_line) + 1;
659 
660 	static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
661 
662 	if (xlen) {
663 		/*
664 		 * We have to put extra_command_line before boot command
665 		 * lines because there could be dashes (separator of init
666 		 * command line) in the command lines.
667 		 */
668 		strcpy(saved_command_line, extra_command_line);
669 		strcpy(static_command_line, extra_command_line);
670 	}
671 	strcpy(saved_command_line + xlen, boot_command_line);
672 	strcpy(static_command_line + xlen, command_line);
673 
674 	if (ilen) {
675 		/*
676 		 * Append supplemental init boot args to saved_command_line
677 		 * so that user can check what command line options passed
678 		 * to init.
679 		 * The order should always be
680 		 * " -- "[bootconfig init-param][cmdline init-param]
681 		 */
682 		if (initargs_offs) {
683 			len = xlen + initargs_offs;
684 			strcpy(saved_command_line + len, extra_init_args);
685 			len += ilen - 4;	/* strlen(extra_init_args) */
686 			strcpy(saved_command_line + len,
687 				boot_command_line + initargs_offs - 1);
688 		} else {
689 			len = strlen(saved_command_line);
690 			strcpy(saved_command_line + len, " -- ");
691 			len += 4;
692 			strcpy(saved_command_line + len, extra_init_args);
693 		}
694 	}
695 
696 	saved_command_line_len = strlen(saved_command_line);
697 }
698 
699 /*
700  * We need to finalize in a non-__init function or else race conditions
701  * between the root thread and the init thread may cause start_kernel to
702  * be reaped by free_initmem before the root thread has proceeded to
703  * cpu_idle.
704  *
705  * gcc-3.4 accidentally inlines this function, so use noinline.
706  */
707 
708 static __initdata DECLARE_COMPLETION(kthreadd_done);
709 
710 static noinline void __ref __noreturn rest_init(void)
711 {
712 	struct task_struct *tsk;
713 	int pid;
714 
715 	rcu_scheduler_starting();
716 	/*
717 	 * We need to spawn init first so that it obtains pid 1, however
718 	 * the init task will end up wanting to create kthreads, which, if
719 	 * we schedule it before we create kthreadd, will OOPS.
720 	 */
721 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
722 	/*
723 	 * Pin init on the boot CPU. Task migration is not properly working
724 	 * until sched_init_smp() has been run. It will set the allowed
725 	 * CPUs for init to the non isolated CPUs.
726 	 */
727 	rcu_read_lock();
728 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
729 	tsk->flags |= PF_NO_SETAFFINITY;
730 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
731 	rcu_read_unlock();
732 
733 	numa_default_policy();
734 	pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
735 	rcu_read_lock();
736 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
737 	rcu_read_unlock();
738 
739 	/*
740 	 * Enable might_sleep() and smp_processor_id() checks.
741 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
742 	 * kernel_thread() would trigger might_sleep() splats. With
743 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
744 	 * already, but it's stuck on the kthreadd_done completion.
745 	 */
746 	system_state = SYSTEM_SCHEDULING;
747 
748 	complete(&kthreadd_done);
749 
750 	/*
751 	 * The boot idle thread must execute schedule()
752 	 * at least once to get things moving:
753 	 */
754 	schedule_preempt_disabled();
755 	/* Call into cpu_idle with preempt disabled */
756 	cpu_startup_entry(CPUHP_ONLINE);
757 }
758 
759 /* Check for early params. */
760 static int __init do_early_param(char *param, char *val,
761 				 const char *unused, void *arg)
762 {
763 	const struct obs_kernel_param *p;
764 
765 	for (p = __setup_start; p < __setup_end; p++) {
766 		if (p->early && parameq(param, p->str)) {
767 			if (p->setup_func(val) != 0)
768 				pr_warn("Malformed early option '%s'\n", param);
769 		}
770 	}
771 	/* We accept everything at this stage. */
772 	return 0;
773 }
774 
775 void __init parse_early_options(char *cmdline)
776 {
777 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
778 		   do_early_param);
779 }
780 
781 /* Arch code calls this early on, or if not, just before other parsing. */
782 void __init parse_early_param(void)
783 {
784 	static int done __initdata;
785 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
786 
787 	if (done)
788 		return;
789 
790 	/* All fall through to do_early_param. */
791 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
792 	parse_early_options(tmp_cmdline);
793 	done = 1;
794 }
795 
796 void __init __weak arch_post_acpi_subsys_init(void) { }
797 
798 void __init __weak smp_setup_processor_id(void)
799 {
800 }
801 
802 void __init __weak smp_prepare_boot_cpu(void)
803 {
804 }
805 
806 # if THREAD_SIZE >= PAGE_SIZE
807 void __init __weak thread_stack_cache_init(void)
808 {
809 }
810 #endif
811 
812 void __init __weak poking_init(void) { }
813 
814 void __init __weak pgtable_cache_init(void) { }
815 
816 void __init __weak trap_init(void) { }
817 
818 bool initcall_debug;
819 core_param(initcall_debug, initcall_debug, bool, 0644);
820 
821 #ifdef TRACEPOINTS_ENABLED
822 static void __init initcall_debug_enable(void);
823 #else
824 static inline void initcall_debug_enable(void)
825 {
826 }
827 #endif
828 
829 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
830 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
831 			   randomize_kstack_offset);
832 DEFINE_PER_CPU(u32, kstack_offset);
833 
834 static int __init early_randomize_kstack_offset(char *buf)
835 {
836 	int ret;
837 	bool bool_result;
838 
839 	ret = kstrtobool(buf, &bool_result);
840 	if (ret)
841 		return ret;
842 
843 	if (bool_result)
844 		static_branch_enable(&randomize_kstack_offset);
845 	else
846 		static_branch_disable(&randomize_kstack_offset);
847 	return 0;
848 }
849 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
850 #endif
851 
852 static void __init print_unknown_bootoptions(void)
853 {
854 	char *unknown_options;
855 	char *end;
856 	const char *const *p;
857 	size_t len;
858 
859 	if (panic_later || (!argv_init[1] && !envp_init[2]))
860 		return;
861 
862 	/*
863 	 * Determine how many options we have to print out, plus a space
864 	 * before each
865 	 */
866 	len = 1; /* null terminator */
867 	for (p = &argv_init[1]; *p; p++) {
868 		len++;
869 		len += strlen(*p);
870 	}
871 	for (p = &envp_init[2]; *p; p++) {
872 		len++;
873 		len += strlen(*p);
874 	}
875 
876 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
877 	if (!unknown_options) {
878 		pr_err("%s: Failed to allocate %zu bytes\n",
879 			__func__, len);
880 		return;
881 	}
882 	end = unknown_options;
883 
884 	for (p = &argv_init[1]; *p; p++)
885 		end += sprintf(end, " %s", *p);
886 	for (p = &envp_init[2]; *p; p++)
887 		end += sprintf(end, " %s", *p);
888 
889 	/* Start at unknown_options[1] to skip the initial space */
890 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
891 		&unknown_options[1]);
892 	memblock_free(unknown_options, len);
893 }
894 
895 static void __init early_numa_node_init(void)
896 {
897 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
898 #ifndef cpu_to_node
899 	int cpu;
900 
901 	/* The early_cpu_to_node() should be ready here. */
902 	for_each_possible_cpu(cpu)
903 		set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
904 #endif
905 #endif
906 }
907 
908 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
909 void start_kernel(void)
910 {
911 	char *command_line;
912 	char *after_dashes;
913 
914 	set_task_stack_end_magic(&init_task);
915 	smp_setup_processor_id();
916 	debug_objects_early_init();
917 	init_vmlinux_build_id();
918 
919 	cgroup_init_early();
920 
921 	local_irq_disable();
922 	early_boot_irqs_disabled = true;
923 
924 	/*
925 	 * Interrupts are still disabled. Do necessary setups, then
926 	 * enable them.
927 	 */
928 	boot_cpu_init();
929 	page_address_init();
930 	pr_notice("%s", linux_banner);
931 	setup_arch(&command_line);
932 	/* Static keys and static calls are needed by LSMs */
933 	jump_label_init();
934 	static_call_init();
935 	early_security_init();
936 	setup_boot_config();
937 	setup_command_line(command_line);
938 	setup_nr_cpu_ids();
939 	setup_per_cpu_areas();
940 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
941 	early_numa_node_init();
942 	boot_cpu_hotplug_init();
943 
944 	pr_notice("Kernel command line: %s\n", saved_command_line);
945 	/* parameters may set static keys */
946 	parse_early_param();
947 	after_dashes = parse_args("Booting kernel",
948 				  static_command_line, __start___param,
949 				  __stop___param - __start___param,
950 				  -1, -1, NULL, &unknown_bootoption);
951 	print_unknown_bootoptions();
952 	if (!IS_ERR_OR_NULL(after_dashes))
953 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
954 			   NULL, set_init_arg);
955 	if (extra_init_args)
956 		parse_args("Setting extra init args", extra_init_args,
957 			   NULL, 0, -1, -1, NULL, set_init_arg);
958 
959 	/* Architectural and non-timekeeping rng init, before allocator init */
960 	random_init_early(command_line);
961 
962 	/*
963 	 * These use large bootmem allocations and must precede
964 	 * initalization of page allocator
965 	 */
966 	setup_log_buf(0);
967 	vfs_caches_init_early();
968 	sort_main_extable();
969 	trap_init();
970 	mm_core_init();
971 	poking_init();
972 	ftrace_init();
973 
974 	/* trace_printk can be enabled here */
975 	early_trace_init();
976 
977 	/*
978 	 * Set up the scheduler prior starting any interrupts (such as the
979 	 * timer interrupt). Full topology setup happens at smp_init()
980 	 * time - but meanwhile we still have a functioning scheduler.
981 	 */
982 	sched_init();
983 
984 	if (WARN(!irqs_disabled(),
985 		 "Interrupts were enabled *very* early, fixing it\n"))
986 		local_irq_disable();
987 	radix_tree_init();
988 	maple_tree_init();
989 
990 	/*
991 	 * Set up housekeeping before setting up workqueues to allow the unbound
992 	 * workqueue to take non-housekeeping into account.
993 	 */
994 	housekeeping_init();
995 
996 	/*
997 	 * Allow workqueue creation and work item queueing/cancelling
998 	 * early.  Work item execution depends on kthreads and starts after
999 	 * workqueue_init().
1000 	 */
1001 	workqueue_init_early();
1002 
1003 	rcu_init();
1004 	kvfree_rcu_init();
1005 
1006 	/* Trace events are available after this */
1007 	trace_init();
1008 
1009 	if (initcall_debug)
1010 		initcall_debug_enable();
1011 
1012 	context_tracking_init();
1013 	/* init some links before init_ISA_irqs() */
1014 	early_irq_init();
1015 	init_IRQ();
1016 	tick_init();
1017 	rcu_init_nohz();
1018 	timers_init();
1019 	srcu_init();
1020 	hrtimers_init();
1021 	softirq_init();
1022 	timekeeping_init();
1023 	time_init();
1024 
1025 	/* This must be after timekeeping is initialized */
1026 	random_init();
1027 
1028 	/* These make use of the fully initialized rng */
1029 	kfence_init();
1030 	boot_init_stack_canary();
1031 
1032 	perf_event_init();
1033 	profile_init();
1034 	call_function_init();
1035 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1036 
1037 	early_boot_irqs_disabled = false;
1038 	local_irq_enable();
1039 
1040 	kmem_cache_init_late();
1041 
1042 	/*
1043 	 * HACK ALERT! This is early. We're enabling the console before
1044 	 * we've done PCI setups etc, and console_init() must be aware of
1045 	 * this. But we do want output early, in case something goes wrong.
1046 	 */
1047 	console_init();
1048 	if (panic_later)
1049 		panic("Too many boot %s vars at `%s'", panic_later,
1050 		      panic_param);
1051 
1052 	lockdep_init();
1053 
1054 	/*
1055 	 * Need to run this when irqs are enabled, because it wants
1056 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1057 	 * too:
1058 	 */
1059 	locking_selftest();
1060 
1061 #ifdef CONFIG_BLK_DEV_INITRD
1062 	if (initrd_start && !initrd_below_start_ok &&
1063 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1064 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1065 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1066 		    min_low_pfn);
1067 		initrd_start = 0;
1068 	}
1069 #endif
1070 	setup_per_cpu_pageset();
1071 	numa_policy_init();
1072 	acpi_early_init();
1073 	if (late_time_init)
1074 		late_time_init();
1075 	sched_clock_init();
1076 	calibrate_delay();
1077 
1078 	arch_cpu_finalize_init();
1079 
1080 	pid_idr_init();
1081 	anon_vma_init();
1082 	thread_stack_cache_init();
1083 	cred_init();
1084 	fork_init();
1085 	proc_caches_init();
1086 	uts_ns_init();
1087 	key_init();
1088 	security_init();
1089 	dbg_late_init();
1090 	net_ns_init();
1091 	vfs_caches_init();
1092 	pagecache_init();
1093 	signals_init();
1094 	seq_file_init();
1095 	proc_root_init();
1096 	nsfs_init();
1097 	pidfs_init();
1098 	cpuset_init();
1099 	mem_cgroup_init();
1100 	cgroup_init();
1101 	taskstats_init_early();
1102 	delayacct_init();
1103 
1104 	acpi_subsystem_init();
1105 	arch_post_acpi_subsys_init();
1106 	kcsan_init();
1107 
1108 	/* Do the rest non-__init'ed, we're now alive */
1109 	rest_init();
1110 
1111 	/*
1112 	 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1113 	 * and older.
1114 	 */
1115 #if !__has_attribute(__no_stack_protector__)
1116 	prevent_tail_call_optimization();
1117 #endif
1118 }
1119 
1120 /* Call all constructor functions linked into the kernel. */
1121 static void __init do_ctors(void)
1122 {
1123 /*
1124  * For UML, the constructors have already been called by the
1125  * normal setup code as it's just a normal ELF binary, so we
1126  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1127  * even on UML for modules.
1128  */
1129 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1130 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1131 
1132 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1133 		(*fn)();
1134 #endif
1135 }
1136 
1137 #ifdef CONFIG_KALLSYMS
1138 struct blacklist_entry {
1139 	struct list_head next;
1140 	char *buf;
1141 };
1142 
1143 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1144 
1145 static int __init initcall_blacklist(char *str)
1146 {
1147 	char *str_entry;
1148 	struct blacklist_entry *entry;
1149 
1150 	/* str argument is a comma-separated list of functions */
1151 	do {
1152 		str_entry = strsep(&str, ",");
1153 		if (str_entry) {
1154 			pr_debug("blacklisting initcall %s\n", str_entry);
1155 			entry = memblock_alloc_or_panic(sizeof(*entry),
1156 					       SMP_CACHE_BYTES);
1157 			entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1,
1158 						    SMP_CACHE_BYTES);
1159 			strcpy(entry->buf, str_entry);
1160 			list_add(&entry->next, &blacklisted_initcalls);
1161 		}
1162 	} while (str_entry);
1163 
1164 	return 1;
1165 }
1166 
1167 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1168 {
1169 	struct blacklist_entry *entry;
1170 	char fn_name[KSYM_SYMBOL_LEN];
1171 	unsigned long addr;
1172 
1173 	if (list_empty(&blacklisted_initcalls))
1174 		return false;
1175 
1176 	addr = (unsigned long) dereference_function_descriptor(fn);
1177 	sprint_symbol_no_offset(fn_name, addr);
1178 
1179 	/*
1180 	 * fn will be "function_name [module_name]" where [module_name] is not
1181 	 * displayed for built-in init functions.  Strip off the [module_name].
1182 	 */
1183 	strreplace(fn_name, ' ', '\0');
1184 
1185 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1186 		if (!strcmp(fn_name, entry->buf)) {
1187 			pr_debug("initcall %s blacklisted\n", fn_name);
1188 			return true;
1189 		}
1190 	}
1191 
1192 	return false;
1193 }
1194 #else
1195 static int __init initcall_blacklist(char *str)
1196 {
1197 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1198 	return 0;
1199 }
1200 
1201 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1202 {
1203 	return false;
1204 }
1205 #endif
1206 __setup("initcall_blacklist=", initcall_blacklist);
1207 
1208 static __init_or_module void
1209 trace_initcall_start_cb(void *data, initcall_t fn)
1210 {
1211 	ktime_t *calltime = data;
1212 
1213 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1214 	*calltime = ktime_get();
1215 }
1216 
1217 static __init_or_module void
1218 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1219 {
1220 	ktime_t rettime, *calltime = data;
1221 
1222 	rettime = ktime_get();
1223 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1224 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1225 }
1226 
1227 static __init_or_module void
1228 trace_initcall_level_cb(void *data, const char *level)
1229 {
1230 	printk(KERN_DEBUG "entering initcall level: %s\n", level);
1231 }
1232 
1233 static ktime_t initcall_calltime;
1234 
1235 #ifdef TRACEPOINTS_ENABLED
1236 static void __init initcall_debug_enable(void)
1237 {
1238 	int ret;
1239 
1240 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1241 					    &initcall_calltime);
1242 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1243 					      &initcall_calltime);
1244 	ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL);
1245 	WARN(ret, "Failed to register initcall tracepoints\n");
1246 }
1247 # define do_trace_initcall_start	trace_initcall_start
1248 # define do_trace_initcall_finish	trace_initcall_finish
1249 # define do_trace_initcall_level	trace_initcall_level
1250 #else
1251 static inline void do_trace_initcall_start(initcall_t fn)
1252 {
1253 	if (!initcall_debug)
1254 		return;
1255 	trace_initcall_start_cb(&initcall_calltime, fn);
1256 }
1257 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1258 {
1259 	if (!initcall_debug)
1260 		return;
1261 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1262 }
1263 static inline void do_trace_initcall_level(const char *level)
1264 {
1265 	if (!initcall_debug)
1266 		return;
1267 	trace_initcall_level_cb(NULL, level);
1268 }
1269 #endif /* !TRACEPOINTS_ENABLED */
1270 
1271 int __init_or_module do_one_initcall(initcall_t fn)
1272 {
1273 	int count = preempt_count();
1274 	char msgbuf[64];
1275 	int ret;
1276 
1277 	if (initcall_blacklisted(fn))
1278 		return -EPERM;
1279 
1280 	do_trace_initcall_start(fn);
1281 	ret = fn();
1282 	do_trace_initcall_finish(fn, ret);
1283 
1284 	msgbuf[0] = 0;
1285 
1286 	if (preempt_count() != count) {
1287 		sprintf(msgbuf, "preemption imbalance ");
1288 		preempt_count_set(count);
1289 	}
1290 	if (irqs_disabled()) {
1291 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1292 		local_irq_enable();
1293 	}
1294 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1295 
1296 	add_latent_entropy();
1297 	return ret;
1298 }
1299 
1300 
1301 static initcall_entry_t *initcall_levels[] __initdata = {
1302 	__initcall0_start,
1303 	__initcall1_start,
1304 	__initcall2_start,
1305 	__initcall3_start,
1306 	__initcall4_start,
1307 	__initcall5_start,
1308 	__initcall6_start,
1309 	__initcall7_start,
1310 	__initcall_end,
1311 };
1312 
1313 /* Keep these in sync with initcalls in include/linux/init.h */
1314 static const char *initcall_level_names[] __initdata = {
1315 	"pure",
1316 	"core",
1317 	"postcore",
1318 	"arch",
1319 	"subsys",
1320 	"fs",
1321 	"device",
1322 	"late",
1323 };
1324 
1325 static int __init ignore_unknown_bootoption(char *param, char *val,
1326 			       const char *unused, void *arg)
1327 {
1328 	return 0;
1329 }
1330 
1331 static void __init do_initcall_level(int level, char *command_line)
1332 {
1333 	initcall_entry_t *fn;
1334 
1335 	parse_args(initcall_level_names[level],
1336 		   command_line, __start___param,
1337 		   __stop___param - __start___param,
1338 		   level, level,
1339 		   NULL, ignore_unknown_bootoption);
1340 
1341 	do_trace_initcall_level(initcall_level_names[level]);
1342 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1343 		do_one_initcall(initcall_from_entry(fn));
1344 }
1345 
1346 static void __init do_initcalls(void)
1347 {
1348 	int level;
1349 	size_t len = saved_command_line_len + 1;
1350 	char *command_line;
1351 
1352 	command_line = kzalloc(len, GFP_KERNEL);
1353 	if (!command_line)
1354 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1355 
1356 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1357 		/* Parser modifies command_line, restore it each time */
1358 		strcpy(command_line, saved_command_line);
1359 		do_initcall_level(level, command_line);
1360 	}
1361 
1362 	kfree(command_line);
1363 }
1364 
1365 /*
1366  * Ok, the machine is now initialized. None of the devices
1367  * have been touched yet, but the CPU subsystem is up and
1368  * running, and memory and process management works.
1369  *
1370  * Now we can finally start doing some real work..
1371  */
1372 static void __init do_basic_setup(void)
1373 {
1374 	cpuset_init_smp();
1375 	driver_init();
1376 	init_irq_proc();
1377 	do_ctors();
1378 	do_initcalls();
1379 }
1380 
1381 static void __init do_pre_smp_initcalls(void)
1382 {
1383 	initcall_entry_t *fn;
1384 
1385 	do_trace_initcall_level("early");
1386 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1387 		do_one_initcall(initcall_from_entry(fn));
1388 }
1389 
1390 static int run_init_process(const char *init_filename)
1391 {
1392 	const char *const *p;
1393 
1394 	argv_init[0] = init_filename;
1395 	pr_info("Run %s as init process\n", init_filename);
1396 	pr_debug("  with arguments:\n");
1397 	for (p = argv_init; *p; p++)
1398 		pr_debug("    %s\n", *p);
1399 	pr_debug("  with environment:\n");
1400 	for (p = envp_init; *p; p++)
1401 		pr_debug("    %s\n", *p);
1402 	return kernel_execve(init_filename, argv_init, envp_init);
1403 }
1404 
1405 static int try_to_run_init_process(const char *init_filename)
1406 {
1407 	int ret;
1408 
1409 	ret = run_init_process(init_filename);
1410 
1411 	if (ret && ret != -ENOENT) {
1412 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1413 		       init_filename, ret);
1414 	}
1415 
1416 	return ret;
1417 }
1418 
1419 static noinline void __init kernel_init_freeable(void);
1420 
1421 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1422 bool rodata_enabled __ro_after_init = true;
1423 
1424 #ifndef arch_parse_debug_rodata
1425 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1426 #endif
1427 
1428 static int __init set_debug_rodata(char *str)
1429 {
1430 	if (arch_parse_debug_rodata(str))
1431 		return 0;
1432 
1433 	if (str && !strcmp(str, "on"))
1434 		rodata_enabled = true;
1435 	else if (str && !strcmp(str, "off"))
1436 		rodata_enabled = false;
1437 	else
1438 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1439 	return 0;
1440 }
1441 early_param("rodata", set_debug_rodata);
1442 #endif
1443 
1444 static void mark_readonly(void)
1445 {
1446 	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1447 		/*
1448 		 * load_module() results in W+X mappings, which are cleaned
1449 		 * up with init_free_wq. Let's make sure that queued work is
1450 		 * flushed so that we don't hit false positives looking for
1451 		 * insecure pages which are W+X.
1452 		 */
1453 		flush_module_init_free_work();
1454 		jump_label_init_ro();
1455 		mark_rodata_ro();
1456 		debug_checkwx();
1457 		rodata_test();
1458 	} else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1459 		pr_info("Kernel memory protection disabled.\n");
1460 	} else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1461 		pr_warn("Kernel memory protection not selected by kernel config.\n");
1462 	} else {
1463 		pr_warn("This architecture does not have kernel memory protection.\n");
1464 	}
1465 }
1466 
1467 void __weak free_initmem(void)
1468 {
1469 	free_initmem_default(POISON_FREE_INITMEM);
1470 }
1471 
1472 static int __ref kernel_init(void *unused)
1473 {
1474 	int ret;
1475 
1476 	/*
1477 	 * Wait until kthreadd is all set-up.
1478 	 */
1479 	wait_for_completion(&kthreadd_done);
1480 
1481 	kernel_init_freeable();
1482 	/* need to finish all async __init code before freeing the memory */
1483 	async_synchronize_full();
1484 
1485 	system_state = SYSTEM_FREEING_INITMEM;
1486 	kprobe_free_init_mem();
1487 	ftrace_free_init_mem();
1488 	kgdb_free_init_mem();
1489 	exit_boot_config();
1490 	free_initmem();
1491 	mark_readonly();
1492 
1493 	/*
1494 	 * Kernel mappings are now finalized - update the userspace page-table
1495 	 * to finalize PTI.
1496 	 */
1497 	pti_finalize();
1498 
1499 	system_state = SYSTEM_RUNNING;
1500 	numa_default_policy();
1501 
1502 	rcu_end_inkernel_boot();
1503 
1504 	do_sysctl_args();
1505 
1506 	if (ramdisk_execute_command) {
1507 		ret = run_init_process(ramdisk_execute_command);
1508 		if (!ret)
1509 			return 0;
1510 		pr_err("Failed to execute %s (error %d)\n",
1511 		       ramdisk_execute_command, ret);
1512 	}
1513 
1514 	/*
1515 	 * We try each of these until one succeeds.
1516 	 *
1517 	 * The Bourne shell can be used instead of init if we are
1518 	 * trying to recover a really broken machine.
1519 	 */
1520 	if (execute_command) {
1521 		ret = run_init_process(execute_command);
1522 		if (!ret)
1523 			return 0;
1524 		panic("Requested init %s failed (error %d).",
1525 		      execute_command, ret);
1526 	}
1527 
1528 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1529 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1530 		if (ret)
1531 			pr_err("Default init %s failed (error %d)\n",
1532 			       CONFIG_DEFAULT_INIT, ret);
1533 		else
1534 			return 0;
1535 	}
1536 
1537 	if (!try_to_run_init_process("/sbin/init") ||
1538 	    !try_to_run_init_process("/etc/init") ||
1539 	    !try_to_run_init_process("/bin/init") ||
1540 	    !try_to_run_init_process("/bin/sh"))
1541 		return 0;
1542 
1543 	panic("No working init found.  Try passing init= option to kernel. "
1544 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1545 }
1546 
1547 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1548 void __init console_on_rootfs(void)
1549 {
1550 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1551 
1552 	if (IS_ERR(file)) {
1553 		pr_err("Warning: unable to open an initial console.\n");
1554 		return;
1555 	}
1556 	init_dup(file);
1557 	init_dup(file);
1558 	init_dup(file);
1559 	fput(file);
1560 }
1561 
1562 static noinline void __init kernel_init_freeable(void)
1563 {
1564 	/* Now the scheduler is fully set up and can do blocking allocations */
1565 	gfp_allowed_mask = __GFP_BITS_MASK;
1566 
1567 	/*
1568 	 * init can allocate pages on any node
1569 	 */
1570 	set_mems_allowed(node_states[N_MEMORY]);
1571 
1572 	cad_pid = get_pid(task_pid(current));
1573 
1574 	smp_prepare_cpus(setup_max_cpus);
1575 
1576 	workqueue_init();
1577 
1578 	init_mm_internals();
1579 
1580 	do_pre_smp_initcalls();
1581 	lockup_detector_init();
1582 
1583 	smp_init();
1584 	sched_init_smp();
1585 
1586 	workqueue_init_topology();
1587 	async_init();
1588 	padata_init();
1589 	page_alloc_init_late();
1590 
1591 	do_basic_setup();
1592 
1593 	kunit_run_all_tests();
1594 
1595 	wait_for_initramfs();
1596 	console_on_rootfs();
1597 
1598 	/*
1599 	 * check if there is an early userspace init.  If yes, let it do all
1600 	 * the work
1601 	 */
1602 	int ramdisk_command_access;
1603 	ramdisk_command_access = init_eaccess(ramdisk_execute_command);
1604 	if (ramdisk_command_access != 0) {
1605 		pr_warn("check access for rdinit=%s failed: %i, ignoring\n",
1606 			ramdisk_execute_command, ramdisk_command_access);
1607 		ramdisk_execute_command = NULL;
1608 		prepare_namespace();
1609 	}
1610 
1611 	/*
1612 	 * Ok, we have completed the initial bootup, and
1613 	 * we're essentially up and running. Get rid of the
1614 	 * initmem segments and start the user-mode stuff..
1615 	 *
1616 	 * rootfs is available now, try loading the public keys
1617 	 * and default modules
1618 	 */
1619 
1620 	integrity_load_keys();
1621 }
1622