xref: /linux/init/main.c (revision 27d9a0fdb53f05c93ed9c674b870c8add451697e)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/init/main.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
8  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
11  */
12 
13 #define DEBUG		/* Enable initcall_debug */
14 
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/cgroup.h>
54 #include <linux/efi.h>
55 #include <linux/tick.h>
56 #include <linux/sched/isolation.h>
57 #include <linux/interrupt.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/delayacct.h>
60 #include <linux/unistd.h>
61 #include <linux/utsname.h>
62 #include <linux/rmap.h>
63 #include <linux/mempolicy.h>
64 #include <linux/key.h>
65 #include <linux/debug_locks.h>
66 #include <linux/debugobjects.h>
67 #include <linux/lockdep.h>
68 #include <linux/kmemleak.h>
69 #include <linux/padata.h>
70 #include <linux/pid_namespace.h>
71 #include <linux/device/driver.h>
72 #include <linux/kthread.h>
73 #include <linux/sched.h>
74 #include <linux/sched/init.h>
75 #include <linux/signal.h>
76 #include <linux/idr.h>
77 #include <linux/kgdb.h>
78 #include <linux/ftrace.h>
79 #include <linux/async.h>
80 #include <linux/shmem_fs.h>
81 #include <linux/slab.h>
82 #include <linux/perf_event.h>
83 #include <linux/ptrace.h>
84 #include <linux/pti.h>
85 #include <linux/blkdev.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/list.h>
92 #include <linux/integrity.h>
93 #include <linux/proc_ns.h>
94 #include <linux/io.h>
95 #include <linux/cache.h>
96 #include <linux/rodata_test.h>
97 #include <linux/jump_label.h>
98 #include <linux/mem_encrypt.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #include <linux/stackdepot.h>
102 #include <linux/randomize_kstack.h>
103 #include <net/net_namespace.h>
104 
105 #include <asm/io.h>
106 #include <asm/bugs.h>
107 #include <asm/setup.h>
108 #include <asm/sections.h>
109 #include <asm/cacheflush.h>
110 
111 #define CREATE_TRACE_POINTS
112 #include <trace/events/initcall.h>
113 
114 #include <kunit/test.h>
115 
116 static int kernel_init(void *);
117 
118 extern void init_IRQ(void);
119 extern void radix_tree_init(void);
120 extern void maple_tree_init(void);
121 
122 /*
123  * Debug helper: via this flag we know that we are in 'early bootup code'
124  * where only the boot processor is running with IRQ disabled.  This means
125  * two things - IRQ must not be enabled before the flag is cleared and some
126  * operations which are not allowed with IRQ disabled are allowed while the
127  * flag is set.
128  */
129 bool early_boot_irqs_disabled __read_mostly;
130 
131 enum system_states system_state __read_mostly;
132 EXPORT_SYMBOL(system_state);
133 
134 /*
135  * Boot command-line arguments
136  */
137 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
138 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
139 
140 extern void time_init(void);
141 /* Default late time init is NULL. archs can override this later. */
142 void (*__initdata late_time_init)(void);
143 
144 /* Untouched command line saved by arch-specific code. */
145 char __initdata boot_command_line[COMMAND_LINE_SIZE];
146 /* Untouched saved command line (eg. for /proc) */
147 char *saved_command_line __ro_after_init;
148 unsigned int saved_command_line_len __ro_after_init;
149 /* Command line for parameter parsing */
150 static char *static_command_line;
151 /* Untouched extra command line */
152 static char *extra_command_line;
153 /* Extra init arguments */
154 static char *extra_init_args;
155 
156 #ifdef CONFIG_BOOT_CONFIG
157 /* Is bootconfig on command line? */
158 static bool bootconfig_found;
159 static size_t initargs_offs;
160 #else
161 # define bootconfig_found false
162 # define initargs_offs 0
163 #endif
164 
165 static char *execute_command;
166 static char *ramdisk_execute_command = "/init";
167 
168 /*
169  * Used to generate warnings if static_key manipulation functions are used
170  * before jump_label_init is called.
171  */
172 bool static_key_initialized __read_mostly;
173 EXPORT_SYMBOL_GPL(static_key_initialized);
174 
175 /*
176  * If set, this is an indication to the drivers that reset the underlying
177  * device before going ahead with the initialization otherwise driver might
178  * rely on the BIOS and skip the reset operation.
179  *
180  * This is useful if kernel is booting in an unreliable environment.
181  * For ex. kdump situation where previous kernel has crashed, BIOS has been
182  * skipped and devices will be in unknown state.
183  */
184 unsigned int reset_devices;
185 EXPORT_SYMBOL(reset_devices);
186 
187 static int __init set_reset_devices(char *str)
188 {
189 	reset_devices = 1;
190 	return 1;
191 }
192 
193 __setup("reset_devices", set_reset_devices);
194 
195 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
196 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
197 static const char *panic_later, *panic_param;
198 
199 extern const struct obs_kernel_param __setup_start[], __setup_end[];
200 
201 static bool __init obsolete_checksetup(char *line)
202 {
203 	const struct obs_kernel_param *p;
204 	bool had_early_param = false;
205 
206 	p = __setup_start;
207 	do {
208 		int n = strlen(p->str);
209 		if (parameqn(line, p->str, n)) {
210 			if (p->early) {
211 				/* Already done in parse_early_param?
212 				 * (Needs exact match on param part).
213 				 * Keep iterating, as we can have early
214 				 * params and __setups of same names 8( */
215 				if (line[n] == '\0' || line[n] == '=')
216 					had_early_param = true;
217 			} else if (!p->setup_func) {
218 				pr_warn("Parameter %s is obsolete, ignored\n",
219 					p->str);
220 				return true;
221 			} else if (p->setup_func(line + n))
222 				return true;
223 		}
224 		p++;
225 	} while (p < __setup_end);
226 
227 	return had_early_param;
228 }
229 
230 /*
231  * This should be approx 2 Bo*oMips to start (note initial shift), and will
232  * still work even if initially too large, it will just take slightly longer
233  */
234 unsigned long loops_per_jiffy = (1<<12);
235 EXPORT_SYMBOL(loops_per_jiffy);
236 
237 static int __init debug_kernel(char *str)
238 {
239 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
240 	return 0;
241 }
242 
243 static int __init quiet_kernel(char *str)
244 {
245 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
246 	return 0;
247 }
248 
249 early_param("debug", debug_kernel);
250 early_param("quiet", quiet_kernel);
251 
252 static int __init loglevel(char *str)
253 {
254 	int newlevel;
255 
256 	/*
257 	 * Only update loglevel value when a correct setting was passed,
258 	 * to prevent blind crashes (when loglevel being set to 0) that
259 	 * are quite hard to debug
260 	 */
261 	if (get_option(&str, &newlevel)) {
262 		console_loglevel = newlevel;
263 		return 0;
264 	}
265 
266 	return -EINVAL;
267 }
268 
269 early_param("loglevel", loglevel);
270 
271 #ifdef CONFIG_BLK_DEV_INITRD
272 static void * __init get_boot_config_from_initrd(size_t *_size)
273 {
274 	u32 size, csum;
275 	char *data;
276 	u32 *hdr;
277 	int i;
278 
279 	if (!initrd_end)
280 		return NULL;
281 
282 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
283 	/*
284 	 * Since Grub may align the size of initrd to 4, we must
285 	 * check the preceding 3 bytes as well.
286 	 */
287 	for (i = 0; i < 4; i++) {
288 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
289 			goto found;
290 		data--;
291 	}
292 	return NULL;
293 
294 found:
295 	hdr = (u32 *)(data - 8);
296 	size = le32_to_cpu(hdr[0]);
297 	csum = le32_to_cpu(hdr[1]);
298 
299 	data = ((void *)hdr) - size;
300 	if ((unsigned long)data < initrd_start) {
301 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
302 			size, initrd_end - initrd_start);
303 		return NULL;
304 	}
305 
306 	if (xbc_calc_checksum(data, size) != csum) {
307 		pr_err("bootconfig checksum failed\n");
308 		return NULL;
309 	}
310 
311 	/* Remove bootconfig from initramfs/initrd */
312 	initrd_end = (unsigned long)data;
313 	if (_size)
314 		*_size = size;
315 
316 	return data;
317 }
318 #else
319 static void * __init get_boot_config_from_initrd(size_t *_size)
320 {
321 	return NULL;
322 }
323 #endif
324 
325 #ifdef CONFIG_BOOT_CONFIG
326 
327 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
328 
329 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
330 
331 static int __init xbc_snprint_cmdline(char *buf, size_t size,
332 				      struct xbc_node *root)
333 {
334 	struct xbc_node *knode, *vnode;
335 	char *end = buf + size;
336 	const char *val;
337 	int ret;
338 
339 	xbc_node_for_each_key_value(root, knode, val) {
340 		ret = xbc_node_compose_key_after(root, knode,
341 					xbc_namebuf, XBC_KEYLEN_MAX);
342 		if (ret < 0)
343 			return ret;
344 
345 		vnode = xbc_node_get_child(knode);
346 		if (!vnode) {
347 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
348 			if (ret < 0)
349 				return ret;
350 			buf += ret;
351 			continue;
352 		}
353 		xbc_array_for_each_value(vnode, val) {
354 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
355 				       xbc_namebuf, val);
356 			if (ret < 0)
357 				return ret;
358 			buf += ret;
359 		}
360 	}
361 
362 	return buf - (end - size);
363 }
364 #undef rest
365 
366 /* Make an extra command line under given key word */
367 static char * __init xbc_make_cmdline(const char *key)
368 {
369 	struct xbc_node *root;
370 	char *new_cmdline;
371 	int ret, len = 0;
372 
373 	root = xbc_find_node(key);
374 	if (!root)
375 		return NULL;
376 
377 	/* Count required buffer size */
378 	len = xbc_snprint_cmdline(NULL, 0, root);
379 	if (len <= 0)
380 		return NULL;
381 
382 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
383 	if (!new_cmdline) {
384 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
385 		return NULL;
386 	}
387 
388 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
389 	if (ret < 0 || ret > len) {
390 		pr_err("Failed to print extra kernel cmdline.\n");
391 		memblock_free(new_cmdline, len + 1);
392 		return NULL;
393 	}
394 
395 	return new_cmdline;
396 }
397 
398 static int __init bootconfig_params(char *param, char *val,
399 				    const char *unused, void *arg)
400 {
401 	if (strcmp(param, "bootconfig") == 0) {
402 		bootconfig_found = true;
403 	}
404 	return 0;
405 }
406 
407 static int __init warn_bootconfig(char *str)
408 {
409 	/* The 'bootconfig' has been handled by bootconfig_params(). */
410 	return 0;
411 }
412 
413 static void __init setup_boot_config(void)
414 {
415 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
416 	const char *msg, *data;
417 	int pos, ret;
418 	size_t size;
419 	char *err;
420 
421 	/* Cut out the bootconfig data even if we have no bootconfig option */
422 	data = get_boot_config_from_initrd(&size);
423 	/* If there is no bootconfig in initrd, try embedded one. */
424 	if (!data)
425 		data = xbc_get_embedded_bootconfig(&size);
426 
427 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
428 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
429 			 bootconfig_params);
430 
431 	if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
432 		return;
433 
434 	/* parse_args() stops at the next param of '--' and returns an address */
435 	if (err)
436 		initargs_offs = err - tmp_cmdline;
437 
438 	if (!data) {
439 		/* If user intended to use bootconfig, show an error level message */
440 		if (bootconfig_found)
441 			pr_err("'bootconfig' found on command line, but no bootconfig found\n");
442 		else
443 			pr_info("No bootconfig data provided, so skipping bootconfig");
444 		return;
445 	}
446 
447 	if (size >= XBC_DATA_MAX) {
448 		pr_err("bootconfig size %ld greater than max size %d\n",
449 			(long)size, XBC_DATA_MAX);
450 		return;
451 	}
452 
453 	ret = xbc_init(data, size, &msg, &pos);
454 	if (ret < 0) {
455 		if (pos < 0)
456 			pr_err("Failed to init bootconfig: %s.\n", msg);
457 		else
458 			pr_err("Failed to parse bootconfig: %s at %d.\n",
459 				msg, pos);
460 	} else {
461 		xbc_get_info(&ret, NULL);
462 		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
463 		/* keys starting with "kernel." are passed via cmdline */
464 		extra_command_line = xbc_make_cmdline("kernel");
465 		/* Also, "init." keys are init arguments */
466 		extra_init_args = xbc_make_cmdline("init");
467 	}
468 	return;
469 }
470 
471 static void __init exit_boot_config(void)
472 {
473 	xbc_exit();
474 }
475 
476 #else	/* !CONFIG_BOOT_CONFIG */
477 
478 static void __init setup_boot_config(void)
479 {
480 	/* Remove bootconfig data from initrd */
481 	get_boot_config_from_initrd(NULL);
482 }
483 
484 static int __init warn_bootconfig(char *str)
485 {
486 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
487 	return 0;
488 }
489 
490 #define exit_boot_config()	do {} while (0)
491 
492 #endif	/* CONFIG_BOOT_CONFIG */
493 
494 early_param("bootconfig", warn_bootconfig);
495 
496 /* Change NUL term back to "=", to make "param" the whole string. */
497 static void __init repair_env_string(char *param, char *val)
498 {
499 	if (val) {
500 		/* param=val or param="val"? */
501 		if (val == param+strlen(param)+1)
502 			val[-1] = '=';
503 		else if (val == param+strlen(param)+2) {
504 			val[-2] = '=';
505 			memmove(val-1, val, strlen(val)+1);
506 		} else
507 			BUG();
508 	}
509 }
510 
511 /* Anything after -- gets handed straight to init. */
512 static int __init set_init_arg(char *param, char *val,
513 			       const char *unused, void *arg)
514 {
515 	unsigned int i;
516 
517 	if (panic_later)
518 		return 0;
519 
520 	repair_env_string(param, val);
521 
522 	for (i = 0; argv_init[i]; i++) {
523 		if (i == MAX_INIT_ARGS) {
524 			panic_later = "init";
525 			panic_param = param;
526 			return 0;
527 		}
528 	}
529 	argv_init[i] = param;
530 	return 0;
531 }
532 
533 /*
534  * Unknown boot options get handed to init, unless they look like
535  * unused parameters (modprobe will find them in /proc/cmdline).
536  */
537 static int __init unknown_bootoption(char *param, char *val,
538 				     const char *unused, void *arg)
539 {
540 	size_t len = strlen(param);
541 
542 	repair_env_string(param, val);
543 
544 	/* Handle obsolete-style parameters */
545 	if (obsolete_checksetup(param))
546 		return 0;
547 
548 	/* Unused module parameter. */
549 	if (strnchr(param, len, '.'))
550 		return 0;
551 
552 	if (panic_later)
553 		return 0;
554 
555 	if (val) {
556 		/* Environment option */
557 		unsigned int i;
558 		for (i = 0; envp_init[i]; i++) {
559 			if (i == MAX_INIT_ENVS) {
560 				panic_later = "env";
561 				panic_param = param;
562 			}
563 			if (!strncmp(param, envp_init[i], len+1))
564 				break;
565 		}
566 		envp_init[i] = param;
567 	} else {
568 		/* Command line option */
569 		unsigned int i;
570 		for (i = 0; argv_init[i]; i++) {
571 			if (i == MAX_INIT_ARGS) {
572 				panic_later = "init";
573 				panic_param = param;
574 			}
575 		}
576 		argv_init[i] = param;
577 	}
578 	return 0;
579 }
580 
581 static int __init init_setup(char *str)
582 {
583 	unsigned int i;
584 
585 	execute_command = str;
586 	/*
587 	 * In case LILO is going to boot us with default command line,
588 	 * it prepends "auto" before the whole cmdline which makes
589 	 * the shell think it should execute a script with such name.
590 	 * So we ignore all arguments entered _before_ init=... [MJ]
591 	 */
592 	for (i = 1; i < MAX_INIT_ARGS; i++)
593 		argv_init[i] = NULL;
594 	return 1;
595 }
596 __setup("init=", init_setup);
597 
598 static int __init rdinit_setup(char *str)
599 {
600 	unsigned int i;
601 
602 	ramdisk_execute_command = str;
603 	/* See "auto" comment in init_setup */
604 	for (i = 1; i < MAX_INIT_ARGS; i++)
605 		argv_init[i] = NULL;
606 	return 1;
607 }
608 __setup("rdinit=", rdinit_setup);
609 
610 #ifndef CONFIG_SMP
611 static const unsigned int setup_max_cpus = NR_CPUS;
612 static inline void setup_nr_cpu_ids(void) { }
613 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
614 #endif
615 
616 /*
617  * We need to store the untouched command line for future reference.
618  * We also need to store the touched command line since the parameter
619  * parsing is performed in place, and we should allow a component to
620  * store reference of name/value for future reference.
621  */
622 static void __init setup_command_line(char *command_line)
623 {
624 	size_t len, xlen = 0, ilen = 0;
625 
626 	if (extra_command_line)
627 		xlen = strlen(extra_command_line);
628 	if (extra_init_args)
629 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
630 
631 	len = xlen + strlen(boot_command_line) + 1;
632 
633 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
634 	if (!saved_command_line)
635 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
636 
637 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
638 	if (!static_command_line)
639 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
640 
641 	if (xlen) {
642 		/*
643 		 * We have to put extra_command_line before boot command
644 		 * lines because there could be dashes (separator of init
645 		 * command line) in the command lines.
646 		 */
647 		strcpy(saved_command_line, extra_command_line);
648 		strcpy(static_command_line, extra_command_line);
649 	}
650 	strcpy(saved_command_line + xlen, boot_command_line);
651 	strcpy(static_command_line + xlen, command_line);
652 
653 	if (ilen) {
654 		/*
655 		 * Append supplemental init boot args to saved_command_line
656 		 * so that user can check what command line options passed
657 		 * to init.
658 		 * The order should always be
659 		 * " -- "[bootconfig init-param][cmdline init-param]
660 		 */
661 		if (initargs_offs) {
662 			len = xlen + initargs_offs;
663 			strcpy(saved_command_line + len, extra_init_args);
664 			len += ilen - 4;	/* strlen(extra_init_args) */
665 			strcpy(saved_command_line + len,
666 				boot_command_line + initargs_offs - 1);
667 		} else {
668 			len = strlen(saved_command_line);
669 			strcpy(saved_command_line + len, " -- ");
670 			len += 4;
671 			strcpy(saved_command_line + len, extra_init_args);
672 		}
673 	}
674 
675 	saved_command_line_len = strlen(saved_command_line);
676 }
677 
678 /*
679  * We need to finalize in a non-__init function or else race conditions
680  * between the root thread and the init thread may cause start_kernel to
681  * be reaped by free_initmem before the root thread has proceeded to
682  * cpu_idle.
683  *
684  * gcc-3.4 accidentally inlines this function, so use noinline.
685  */
686 
687 static __initdata DECLARE_COMPLETION(kthreadd_done);
688 
689 noinline void __ref rest_init(void)
690 {
691 	struct task_struct *tsk;
692 	int pid;
693 
694 	rcu_scheduler_starting();
695 	/*
696 	 * We need to spawn init first so that it obtains pid 1, however
697 	 * the init task will end up wanting to create kthreads, which, if
698 	 * we schedule it before we create kthreadd, will OOPS.
699 	 */
700 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
701 	/*
702 	 * Pin init on the boot CPU. Task migration is not properly working
703 	 * until sched_init_smp() has been run. It will set the allowed
704 	 * CPUs for init to the non isolated CPUs.
705 	 */
706 	rcu_read_lock();
707 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
708 	tsk->flags |= PF_NO_SETAFFINITY;
709 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
710 	rcu_read_unlock();
711 
712 	numa_default_policy();
713 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
714 	rcu_read_lock();
715 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
716 	rcu_read_unlock();
717 
718 	/*
719 	 * Enable might_sleep() and smp_processor_id() checks.
720 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
721 	 * kernel_thread() would trigger might_sleep() splats. With
722 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
723 	 * already, but it's stuck on the kthreadd_done completion.
724 	 */
725 	system_state = SYSTEM_SCHEDULING;
726 
727 	complete(&kthreadd_done);
728 
729 	/*
730 	 * The boot idle thread must execute schedule()
731 	 * at least once to get things moving:
732 	 */
733 	schedule_preempt_disabled();
734 	/* Call into cpu_idle with preempt disabled */
735 	cpu_startup_entry(CPUHP_ONLINE);
736 }
737 
738 /* Check for early params. */
739 static int __init do_early_param(char *param, char *val,
740 				 const char *unused, void *arg)
741 {
742 	const struct obs_kernel_param *p;
743 
744 	for (p = __setup_start; p < __setup_end; p++) {
745 		if ((p->early && parameq(param, p->str)) ||
746 		    (strcmp(param, "console") == 0 &&
747 		     strcmp(p->str, "earlycon") == 0)
748 		) {
749 			if (p->setup_func(val) != 0)
750 				pr_warn("Malformed early option '%s'\n", param);
751 		}
752 	}
753 	/* We accept everything at this stage. */
754 	return 0;
755 }
756 
757 void __init parse_early_options(char *cmdline)
758 {
759 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
760 		   do_early_param);
761 }
762 
763 /* Arch code calls this early on, or if not, just before other parsing. */
764 void __init parse_early_param(void)
765 {
766 	static int done __initdata;
767 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
768 
769 	if (done)
770 		return;
771 
772 	/* All fall through to do_early_param. */
773 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
774 	parse_early_options(tmp_cmdline);
775 	done = 1;
776 }
777 
778 void __init __weak arch_post_acpi_subsys_init(void) { }
779 
780 void __init __weak smp_setup_processor_id(void)
781 {
782 }
783 
784 # if THREAD_SIZE >= PAGE_SIZE
785 void __init __weak thread_stack_cache_init(void)
786 {
787 }
788 #endif
789 
790 void __init __weak mem_encrypt_init(void) { }
791 
792 void __init __weak poking_init(void) { }
793 
794 void __init __weak pgtable_cache_init(void) { }
795 
796 void __init __weak trap_init(void) { }
797 
798 bool initcall_debug;
799 core_param(initcall_debug, initcall_debug, bool, 0644);
800 
801 #ifdef TRACEPOINTS_ENABLED
802 static void __init initcall_debug_enable(void);
803 #else
804 static inline void initcall_debug_enable(void)
805 {
806 }
807 #endif
808 
809 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
810 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
811 			   randomize_kstack_offset);
812 DEFINE_PER_CPU(u32, kstack_offset);
813 
814 static int __init early_randomize_kstack_offset(char *buf)
815 {
816 	int ret;
817 	bool bool_result;
818 
819 	ret = kstrtobool(buf, &bool_result);
820 	if (ret)
821 		return ret;
822 
823 	if (bool_result)
824 		static_branch_enable(&randomize_kstack_offset);
825 	else
826 		static_branch_disable(&randomize_kstack_offset);
827 	return 0;
828 }
829 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
830 #endif
831 
832 void __init __weak arch_call_rest_init(void)
833 {
834 	rest_init();
835 }
836 
837 static void __init print_unknown_bootoptions(void)
838 {
839 	char *unknown_options;
840 	char *end;
841 	const char *const *p;
842 	size_t len;
843 
844 	if (panic_later || (!argv_init[1] && !envp_init[2]))
845 		return;
846 
847 	/*
848 	 * Determine how many options we have to print out, plus a space
849 	 * before each
850 	 */
851 	len = 1; /* null terminator */
852 	for (p = &argv_init[1]; *p; p++) {
853 		len++;
854 		len += strlen(*p);
855 	}
856 	for (p = &envp_init[2]; *p; p++) {
857 		len++;
858 		len += strlen(*p);
859 	}
860 
861 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
862 	if (!unknown_options) {
863 		pr_err("%s: Failed to allocate %zu bytes\n",
864 			__func__, len);
865 		return;
866 	}
867 	end = unknown_options;
868 
869 	for (p = &argv_init[1]; *p; p++)
870 		end += sprintf(end, " %s", *p);
871 	for (p = &envp_init[2]; *p; p++)
872 		end += sprintf(end, " %s", *p);
873 
874 	/* Start at unknown_options[1] to skip the initial space */
875 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
876 		&unknown_options[1]);
877 	memblock_free(unknown_options, len);
878 }
879 
880 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
881 {
882 	char *command_line;
883 	char *after_dashes;
884 
885 	set_task_stack_end_magic(&init_task);
886 	smp_setup_processor_id();
887 	debug_objects_early_init();
888 	init_vmlinux_build_id();
889 
890 	cgroup_init_early();
891 
892 	local_irq_disable();
893 	early_boot_irqs_disabled = true;
894 
895 	/*
896 	 * Interrupts are still disabled. Do necessary setups, then
897 	 * enable them.
898 	 */
899 	boot_cpu_init();
900 	page_address_init();
901 	pr_notice("%s", linux_banner);
902 	early_security_init();
903 	setup_arch(&command_line);
904 	setup_boot_config();
905 	setup_command_line(command_line);
906 	setup_nr_cpu_ids();
907 	setup_per_cpu_areas();
908 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
909 	boot_cpu_hotplug_init();
910 
911 	pr_notice("Kernel command line: %s\n", saved_command_line);
912 	/* parameters may set static keys */
913 	jump_label_init();
914 	parse_early_param();
915 	after_dashes = parse_args("Booting kernel",
916 				  static_command_line, __start___param,
917 				  __stop___param - __start___param,
918 				  -1, -1, NULL, &unknown_bootoption);
919 	print_unknown_bootoptions();
920 	if (!IS_ERR_OR_NULL(after_dashes))
921 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
922 			   NULL, set_init_arg);
923 	if (extra_init_args)
924 		parse_args("Setting extra init args", extra_init_args,
925 			   NULL, 0, -1, -1, NULL, set_init_arg);
926 
927 	/* Architectural and non-timekeeping rng init, before allocator init */
928 	random_init_early(command_line);
929 
930 	/*
931 	 * These use large bootmem allocations and must precede
932 	 * initalization of page allocator
933 	 */
934 	setup_log_buf(0);
935 	vfs_caches_init_early();
936 	sort_main_extable();
937 	trap_init();
938 	mm_core_init();
939 	poking_init();
940 	ftrace_init();
941 
942 	/* trace_printk can be enabled here */
943 	early_trace_init();
944 
945 	/*
946 	 * Set up the scheduler prior starting any interrupts (such as the
947 	 * timer interrupt). Full topology setup happens at smp_init()
948 	 * time - but meanwhile we still have a functioning scheduler.
949 	 */
950 	sched_init();
951 
952 	if (WARN(!irqs_disabled(),
953 		 "Interrupts were enabled *very* early, fixing it\n"))
954 		local_irq_disable();
955 	radix_tree_init();
956 	maple_tree_init();
957 
958 	/*
959 	 * Set up housekeeping before setting up workqueues to allow the unbound
960 	 * workqueue to take non-housekeeping into account.
961 	 */
962 	housekeeping_init();
963 
964 	/*
965 	 * Allow workqueue creation and work item queueing/cancelling
966 	 * early.  Work item execution depends on kthreads and starts after
967 	 * workqueue_init().
968 	 */
969 	workqueue_init_early();
970 
971 	rcu_init();
972 
973 	/* Trace events are available after this */
974 	trace_init();
975 
976 	if (initcall_debug)
977 		initcall_debug_enable();
978 
979 	context_tracking_init();
980 	/* init some links before init_ISA_irqs() */
981 	early_irq_init();
982 	init_IRQ();
983 	tick_init();
984 	rcu_init_nohz();
985 	init_timers();
986 	srcu_init();
987 	hrtimers_init();
988 	softirq_init();
989 	timekeeping_init();
990 	time_init();
991 
992 	/* This must be after timekeeping is initialized */
993 	random_init();
994 
995 	/* These make use of the fully initialized rng */
996 	kfence_init();
997 	boot_init_stack_canary();
998 
999 	perf_event_init();
1000 	profile_init();
1001 	call_function_init();
1002 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1003 
1004 	early_boot_irqs_disabled = false;
1005 	local_irq_enable();
1006 
1007 	kmem_cache_init_late();
1008 
1009 	/*
1010 	 * HACK ALERT! This is early. We're enabling the console before
1011 	 * we've done PCI setups etc, and console_init() must be aware of
1012 	 * this. But we do want output early, in case something goes wrong.
1013 	 */
1014 	console_init();
1015 	if (panic_later)
1016 		panic("Too many boot %s vars at `%s'", panic_later,
1017 		      panic_param);
1018 
1019 	lockdep_init();
1020 
1021 	/*
1022 	 * Need to run this when irqs are enabled, because it wants
1023 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1024 	 * too:
1025 	 */
1026 	locking_selftest();
1027 
1028 	/*
1029 	 * This needs to be called before any devices perform DMA
1030 	 * operations that might use the SWIOTLB bounce buffers. It will
1031 	 * mark the bounce buffers as decrypted so that their usage will
1032 	 * not cause "plain-text" data to be decrypted when accessed.
1033 	 */
1034 	mem_encrypt_init();
1035 
1036 #ifdef CONFIG_BLK_DEV_INITRD
1037 	if (initrd_start && !initrd_below_start_ok &&
1038 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1039 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1040 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1041 		    min_low_pfn);
1042 		initrd_start = 0;
1043 	}
1044 #endif
1045 	setup_per_cpu_pageset();
1046 	numa_policy_init();
1047 	acpi_early_init();
1048 	if (late_time_init)
1049 		late_time_init();
1050 	sched_clock_init();
1051 	calibrate_delay();
1052 	pid_idr_init();
1053 	anon_vma_init();
1054 #ifdef CONFIG_X86
1055 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1056 		efi_enter_virtual_mode();
1057 #endif
1058 	thread_stack_cache_init();
1059 	cred_init();
1060 	fork_init();
1061 	proc_caches_init();
1062 	uts_ns_init();
1063 	key_init();
1064 	security_init();
1065 	dbg_late_init();
1066 	net_ns_init();
1067 	vfs_caches_init();
1068 	pagecache_init();
1069 	signals_init();
1070 	seq_file_init();
1071 	proc_root_init();
1072 	nsfs_init();
1073 	cpuset_init();
1074 	cgroup_init();
1075 	taskstats_init_early();
1076 	delayacct_init();
1077 
1078 	check_bugs();
1079 
1080 	acpi_subsystem_init();
1081 	arch_post_acpi_subsys_init();
1082 	kcsan_init();
1083 
1084 	/* Do the rest non-__init'ed, we're now alive */
1085 	arch_call_rest_init();
1086 
1087 	prevent_tail_call_optimization();
1088 }
1089 
1090 /* Call all constructor functions linked into the kernel. */
1091 static void __init do_ctors(void)
1092 {
1093 /*
1094  * For UML, the constructors have already been called by the
1095  * normal setup code as it's just a normal ELF binary, so we
1096  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1097  * even on UML for modules.
1098  */
1099 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1100 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1101 
1102 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1103 		(*fn)();
1104 #endif
1105 }
1106 
1107 #ifdef CONFIG_KALLSYMS
1108 struct blacklist_entry {
1109 	struct list_head next;
1110 	char *buf;
1111 };
1112 
1113 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1114 
1115 static int __init initcall_blacklist(char *str)
1116 {
1117 	char *str_entry;
1118 	struct blacklist_entry *entry;
1119 
1120 	/* str argument is a comma-separated list of functions */
1121 	do {
1122 		str_entry = strsep(&str, ",");
1123 		if (str_entry) {
1124 			pr_debug("blacklisting initcall %s\n", str_entry);
1125 			entry = memblock_alloc(sizeof(*entry),
1126 					       SMP_CACHE_BYTES);
1127 			if (!entry)
1128 				panic("%s: Failed to allocate %zu bytes\n",
1129 				      __func__, sizeof(*entry));
1130 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1131 						    SMP_CACHE_BYTES);
1132 			if (!entry->buf)
1133 				panic("%s: Failed to allocate %zu bytes\n",
1134 				      __func__, strlen(str_entry) + 1);
1135 			strcpy(entry->buf, str_entry);
1136 			list_add(&entry->next, &blacklisted_initcalls);
1137 		}
1138 	} while (str_entry);
1139 
1140 	return 1;
1141 }
1142 
1143 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1144 {
1145 	struct blacklist_entry *entry;
1146 	char fn_name[KSYM_SYMBOL_LEN];
1147 	unsigned long addr;
1148 
1149 	if (list_empty(&blacklisted_initcalls))
1150 		return false;
1151 
1152 	addr = (unsigned long) dereference_function_descriptor(fn);
1153 	sprint_symbol_no_offset(fn_name, addr);
1154 
1155 	/*
1156 	 * fn will be "function_name [module_name]" where [module_name] is not
1157 	 * displayed for built-in init functions.  Strip off the [module_name].
1158 	 */
1159 	strreplace(fn_name, ' ', '\0');
1160 
1161 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1162 		if (!strcmp(fn_name, entry->buf)) {
1163 			pr_debug("initcall %s blacklisted\n", fn_name);
1164 			return true;
1165 		}
1166 	}
1167 
1168 	return false;
1169 }
1170 #else
1171 static int __init initcall_blacklist(char *str)
1172 {
1173 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1174 	return 0;
1175 }
1176 
1177 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1178 {
1179 	return false;
1180 }
1181 #endif
1182 __setup("initcall_blacklist=", initcall_blacklist);
1183 
1184 static __init_or_module void
1185 trace_initcall_start_cb(void *data, initcall_t fn)
1186 {
1187 	ktime_t *calltime = data;
1188 
1189 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1190 	*calltime = ktime_get();
1191 }
1192 
1193 static __init_or_module void
1194 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1195 {
1196 	ktime_t rettime, *calltime = data;
1197 
1198 	rettime = ktime_get();
1199 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1200 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1201 }
1202 
1203 static ktime_t initcall_calltime;
1204 
1205 #ifdef TRACEPOINTS_ENABLED
1206 static void __init initcall_debug_enable(void)
1207 {
1208 	int ret;
1209 
1210 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1211 					    &initcall_calltime);
1212 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1213 					      &initcall_calltime);
1214 	WARN(ret, "Failed to register initcall tracepoints\n");
1215 }
1216 # define do_trace_initcall_start	trace_initcall_start
1217 # define do_trace_initcall_finish	trace_initcall_finish
1218 #else
1219 static inline void do_trace_initcall_start(initcall_t fn)
1220 {
1221 	if (!initcall_debug)
1222 		return;
1223 	trace_initcall_start_cb(&initcall_calltime, fn);
1224 }
1225 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1226 {
1227 	if (!initcall_debug)
1228 		return;
1229 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1230 }
1231 #endif /* !TRACEPOINTS_ENABLED */
1232 
1233 int __init_or_module do_one_initcall(initcall_t fn)
1234 {
1235 	int count = preempt_count();
1236 	char msgbuf[64];
1237 	int ret;
1238 
1239 	if (initcall_blacklisted(fn))
1240 		return -EPERM;
1241 
1242 	do_trace_initcall_start(fn);
1243 	ret = fn();
1244 	do_trace_initcall_finish(fn, ret);
1245 
1246 	msgbuf[0] = 0;
1247 
1248 	if (preempt_count() != count) {
1249 		sprintf(msgbuf, "preemption imbalance ");
1250 		preempt_count_set(count);
1251 	}
1252 	if (irqs_disabled()) {
1253 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1254 		local_irq_enable();
1255 	}
1256 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1257 
1258 	add_latent_entropy();
1259 	return ret;
1260 }
1261 
1262 
1263 extern initcall_entry_t __initcall_start[];
1264 extern initcall_entry_t __initcall0_start[];
1265 extern initcall_entry_t __initcall1_start[];
1266 extern initcall_entry_t __initcall2_start[];
1267 extern initcall_entry_t __initcall3_start[];
1268 extern initcall_entry_t __initcall4_start[];
1269 extern initcall_entry_t __initcall5_start[];
1270 extern initcall_entry_t __initcall6_start[];
1271 extern initcall_entry_t __initcall7_start[];
1272 extern initcall_entry_t __initcall_end[];
1273 
1274 static initcall_entry_t *initcall_levels[] __initdata = {
1275 	__initcall0_start,
1276 	__initcall1_start,
1277 	__initcall2_start,
1278 	__initcall3_start,
1279 	__initcall4_start,
1280 	__initcall5_start,
1281 	__initcall6_start,
1282 	__initcall7_start,
1283 	__initcall_end,
1284 };
1285 
1286 /* Keep these in sync with initcalls in include/linux/init.h */
1287 static const char *initcall_level_names[] __initdata = {
1288 	"pure",
1289 	"core",
1290 	"postcore",
1291 	"arch",
1292 	"subsys",
1293 	"fs",
1294 	"device",
1295 	"late",
1296 };
1297 
1298 static int __init ignore_unknown_bootoption(char *param, char *val,
1299 			       const char *unused, void *arg)
1300 {
1301 	return 0;
1302 }
1303 
1304 static void __init do_initcall_level(int level, char *command_line)
1305 {
1306 	initcall_entry_t *fn;
1307 
1308 	parse_args(initcall_level_names[level],
1309 		   command_line, __start___param,
1310 		   __stop___param - __start___param,
1311 		   level, level,
1312 		   NULL, ignore_unknown_bootoption);
1313 
1314 	trace_initcall_level(initcall_level_names[level]);
1315 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1316 		do_one_initcall(initcall_from_entry(fn));
1317 }
1318 
1319 static void __init do_initcalls(void)
1320 {
1321 	int level;
1322 	size_t len = saved_command_line_len + 1;
1323 	char *command_line;
1324 
1325 	command_line = kzalloc(len, GFP_KERNEL);
1326 	if (!command_line)
1327 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1328 
1329 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1330 		/* Parser modifies command_line, restore it each time */
1331 		strcpy(command_line, saved_command_line);
1332 		do_initcall_level(level, command_line);
1333 	}
1334 
1335 	kfree(command_line);
1336 }
1337 
1338 /*
1339  * Ok, the machine is now initialized. None of the devices
1340  * have been touched yet, but the CPU subsystem is up and
1341  * running, and memory and process management works.
1342  *
1343  * Now we can finally start doing some real work..
1344  */
1345 static void __init do_basic_setup(void)
1346 {
1347 	cpuset_init_smp();
1348 	driver_init();
1349 	init_irq_proc();
1350 	do_ctors();
1351 	do_initcalls();
1352 }
1353 
1354 static void __init do_pre_smp_initcalls(void)
1355 {
1356 	initcall_entry_t *fn;
1357 
1358 	trace_initcall_level("early");
1359 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1360 		do_one_initcall(initcall_from_entry(fn));
1361 }
1362 
1363 static int run_init_process(const char *init_filename)
1364 {
1365 	const char *const *p;
1366 
1367 	argv_init[0] = init_filename;
1368 	pr_info("Run %s as init process\n", init_filename);
1369 	pr_debug("  with arguments:\n");
1370 	for (p = argv_init; *p; p++)
1371 		pr_debug("    %s\n", *p);
1372 	pr_debug("  with environment:\n");
1373 	for (p = envp_init; *p; p++)
1374 		pr_debug("    %s\n", *p);
1375 	return kernel_execve(init_filename, argv_init, envp_init);
1376 }
1377 
1378 static int try_to_run_init_process(const char *init_filename)
1379 {
1380 	int ret;
1381 
1382 	ret = run_init_process(init_filename);
1383 
1384 	if (ret && ret != -ENOENT) {
1385 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1386 		       init_filename, ret);
1387 	}
1388 
1389 	return ret;
1390 }
1391 
1392 static noinline void __init kernel_init_freeable(void);
1393 
1394 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1395 bool rodata_enabled __ro_after_init = true;
1396 
1397 #ifndef arch_parse_debug_rodata
1398 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1399 #endif
1400 
1401 static int __init set_debug_rodata(char *str)
1402 {
1403 	if (arch_parse_debug_rodata(str))
1404 		return 0;
1405 
1406 	if (str && !strcmp(str, "on"))
1407 		rodata_enabled = true;
1408 	else if (str && !strcmp(str, "off"))
1409 		rodata_enabled = false;
1410 	else
1411 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1412 	return 0;
1413 }
1414 early_param("rodata", set_debug_rodata);
1415 #endif
1416 
1417 #ifdef CONFIG_STRICT_KERNEL_RWX
1418 static void mark_readonly(void)
1419 {
1420 	if (rodata_enabled) {
1421 		/*
1422 		 * load_module() results in W+X mappings, which are cleaned
1423 		 * up with call_rcu().  Let's make sure that queued work is
1424 		 * flushed so that we don't hit false positives looking for
1425 		 * insecure pages which are W+X.
1426 		 */
1427 		rcu_barrier();
1428 		mark_rodata_ro();
1429 		rodata_test();
1430 	} else
1431 		pr_info("Kernel memory protection disabled.\n");
1432 }
1433 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1434 static inline void mark_readonly(void)
1435 {
1436 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1437 }
1438 #else
1439 static inline void mark_readonly(void)
1440 {
1441 	pr_warn("This architecture does not have kernel memory protection.\n");
1442 }
1443 #endif
1444 
1445 void __weak free_initmem(void)
1446 {
1447 	free_initmem_default(POISON_FREE_INITMEM);
1448 }
1449 
1450 static int __ref kernel_init(void *unused)
1451 {
1452 	int ret;
1453 
1454 	/*
1455 	 * Wait until kthreadd is all set-up.
1456 	 */
1457 	wait_for_completion(&kthreadd_done);
1458 
1459 	kernel_init_freeable();
1460 	/* need to finish all async __init code before freeing the memory */
1461 	async_synchronize_full();
1462 
1463 	system_state = SYSTEM_FREEING_INITMEM;
1464 	kprobe_free_init_mem();
1465 	ftrace_free_init_mem();
1466 	kgdb_free_init_mem();
1467 	exit_boot_config();
1468 	free_initmem();
1469 	mark_readonly();
1470 
1471 	/*
1472 	 * Kernel mappings are now finalized - update the userspace page-table
1473 	 * to finalize PTI.
1474 	 */
1475 	pti_finalize();
1476 
1477 	system_state = SYSTEM_RUNNING;
1478 	numa_default_policy();
1479 
1480 	rcu_end_inkernel_boot();
1481 
1482 	do_sysctl_args();
1483 
1484 	if (ramdisk_execute_command) {
1485 		ret = run_init_process(ramdisk_execute_command);
1486 		if (!ret)
1487 			return 0;
1488 		pr_err("Failed to execute %s (error %d)\n",
1489 		       ramdisk_execute_command, ret);
1490 	}
1491 
1492 	/*
1493 	 * We try each of these until one succeeds.
1494 	 *
1495 	 * The Bourne shell can be used instead of init if we are
1496 	 * trying to recover a really broken machine.
1497 	 */
1498 	if (execute_command) {
1499 		ret = run_init_process(execute_command);
1500 		if (!ret)
1501 			return 0;
1502 		panic("Requested init %s failed (error %d).",
1503 		      execute_command, ret);
1504 	}
1505 
1506 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1507 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1508 		if (ret)
1509 			pr_err("Default init %s failed (error %d)\n",
1510 			       CONFIG_DEFAULT_INIT, ret);
1511 		else
1512 			return 0;
1513 	}
1514 
1515 	if (!try_to_run_init_process("/sbin/init") ||
1516 	    !try_to_run_init_process("/etc/init") ||
1517 	    !try_to_run_init_process("/bin/init") ||
1518 	    !try_to_run_init_process("/bin/sh"))
1519 		return 0;
1520 
1521 	panic("No working init found.  Try passing init= option to kernel. "
1522 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1523 }
1524 
1525 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1526 void __init console_on_rootfs(void)
1527 {
1528 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1529 
1530 	if (IS_ERR(file)) {
1531 		pr_err("Warning: unable to open an initial console.\n");
1532 		return;
1533 	}
1534 	init_dup(file);
1535 	init_dup(file);
1536 	init_dup(file);
1537 	fput(file);
1538 }
1539 
1540 static noinline void __init kernel_init_freeable(void)
1541 {
1542 	/* Now the scheduler is fully set up and can do blocking allocations */
1543 	gfp_allowed_mask = __GFP_BITS_MASK;
1544 
1545 	/*
1546 	 * init can allocate pages on any node
1547 	 */
1548 	set_mems_allowed(node_states[N_MEMORY]);
1549 
1550 	cad_pid = get_pid(task_pid(current));
1551 
1552 	smp_prepare_cpus(setup_max_cpus);
1553 
1554 	workqueue_init();
1555 
1556 	init_mm_internals();
1557 
1558 	rcu_init_tasks_generic();
1559 	do_pre_smp_initcalls();
1560 	lockup_detector_init();
1561 
1562 	smp_init();
1563 	sched_init_smp();
1564 
1565 	padata_init();
1566 	page_alloc_init_late();
1567 
1568 	do_basic_setup();
1569 
1570 	kunit_run_all_tests();
1571 
1572 	wait_for_initramfs();
1573 	console_on_rootfs();
1574 
1575 	/*
1576 	 * check if there is an early userspace init.  If yes, let it do all
1577 	 * the work
1578 	 */
1579 	if (init_eaccess(ramdisk_execute_command) != 0) {
1580 		ramdisk_execute_command = NULL;
1581 		prepare_namespace();
1582 	}
1583 
1584 	/*
1585 	 * Ok, we have completed the initial bootup, and
1586 	 * we're essentially up and running. Get rid of the
1587 	 * initmem segments and start the user-mode stuff..
1588 	 *
1589 	 * rootfs is available now, try loading the public keys
1590 	 * and default modules
1591 	 */
1592 
1593 	integrity_load_keys();
1594 }
1595