xref: /openbmc/linux/init/main.c (revision dd1431e5)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/init/main.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
8  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
11  */
12 
13 #define DEBUG		/* Enable initcall_debug */
14 
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/vmalloc.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/start_kernel.h>
40 #include <linux/security.h>
41 #include <linux/smp.h>
42 #include <linux/profile.h>
43 #include <linux/kfence.h>
44 #include <linux/rcupdate.h>
45 #include <linux/moduleparam.h>
46 #include <linux/kallsyms.h>
47 #include <linux/writeback.h>
48 #include <linux/cpu.h>
49 #include <linux/cpuset.h>
50 #include <linux/cgroup.h>
51 #include <linux/efi.h>
52 #include <linux/tick.h>
53 #include <linux/sched/isolation.h>
54 #include <linux/interrupt.h>
55 #include <linux/taskstats_kern.h>
56 #include <linux/delayacct.h>
57 #include <linux/unistd.h>
58 #include <linux/utsname.h>
59 #include <linux/rmap.h>
60 #include <linux/mempolicy.h>
61 #include <linux/key.h>
62 #include <linux/page_ext.h>
63 #include <linux/debug_locks.h>
64 #include <linux/debugobjects.h>
65 #include <linux/lockdep.h>
66 #include <linux/kmemleak.h>
67 #include <linux/padata.h>
68 #include <linux/pid_namespace.h>
69 #include <linux/device/driver.h>
70 #include <linux/kthread.h>
71 #include <linux/sched.h>
72 #include <linux/sched/init.h>
73 #include <linux/signal.h>
74 #include <linux/idr.h>
75 #include <linux/kgdb.h>
76 #include <linux/ftrace.h>
77 #include <linux/async.h>
78 #include <linux/shmem_fs.h>
79 #include <linux/slab.h>
80 #include <linux/perf_event.h>
81 #include <linux/ptrace.h>
82 #include <linux/pti.h>
83 #include <linux/blkdev.h>
84 #include <linux/elevator.h>
85 #include <linux/sched/clock.h>
86 #include <linux/sched/task.h>
87 #include <linux/sched/task_stack.h>
88 #include <linux/context_tracking.h>
89 #include <linux/random.h>
90 #include <linux/list.h>
91 #include <linux/integrity.h>
92 #include <linux/proc_ns.h>
93 #include <linux/io.h>
94 #include <linux/cache.h>
95 #include <linux/rodata_test.h>
96 #include <linux/jump_label.h>
97 #include <linux/mem_encrypt.h>
98 #include <linux/kcsan.h>
99 #include <linux/init_syscalls.h>
100 #include <linux/stackdepot.h>
101 
102 #include <asm/io.h>
103 #include <asm/bugs.h>
104 #include <asm/setup.h>
105 #include <asm/sections.h>
106 #include <asm/cacheflush.h>
107 
108 #define CREATE_TRACE_POINTS
109 #include <trace/events/initcall.h>
110 
111 #include <kunit/test.h>
112 
113 static int kernel_init(void *);
114 
115 extern void init_IRQ(void);
116 extern void radix_tree_init(void);
117 
118 /*
119  * Debug helper: via this flag we know that we are in 'early bootup code'
120  * where only the boot processor is running with IRQ disabled.  This means
121  * two things - IRQ must not be enabled before the flag is cleared and some
122  * operations which are not allowed with IRQ disabled are allowed while the
123  * flag is set.
124  */
125 bool early_boot_irqs_disabled __read_mostly;
126 
127 enum system_states system_state __read_mostly;
128 EXPORT_SYMBOL(system_state);
129 
130 /*
131  * Boot command-line arguments
132  */
133 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
134 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
135 
136 extern void time_init(void);
137 /* Default late time init is NULL. archs can override this later. */
138 void (*__initdata late_time_init)(void);
139 
140 /* Untouched command line saved by arch-specific code. */
141 char __initdata boot_command_line[COMMAND_LINE_SIZE];
142 /* Untouched saved command line (eg. for /proc) */
143 char *saved_command_line;
144 /* Command line for parameter parsing */
145 static char *static_command_line;
146 /* Untouched extra command line */
147 static char *extra_command_line;
148 /* Extra init arguments */
149 static char *extra_init_args;
150 
151 #ifdef CONFIG_BOOT_CONFIG
152 /* Is bootconfig on command line? */
153 static bool bootconfig_found;
154 static bool initargs_found;
155 #else
156 # define bootconfig_found false
157 # define initargs_found false
158 #endif
159 
160 static char *execute_command;
161 static char *ramdisk_execute_command = "/init";
162 
163 /*
164  * Used to generate warnings if static_key manipulation functions are used
165  * before jump_label_init is called.
166  */
167 bool static_key_initialized __read_mostly;
168 EXPORT_SYMBOL_GPL(static_key_initialized);
169 
170 /*
171  * If set, this is an indication to the drivers that reset the underlying
172  * device before going ahead with the initialization otherwise driver might
173  * rely on the BIOS and skip the reset operation.
174  *
175  * This is useful if kernel is booting in an unreliable environment.
176  * For ex. kdump situation where previous kernel has crashed, BIOS has been
177  * skipped and devices will be in unknown state.
178  */
179 unsigned int reset_devices;
180 EXPORT_SYMBOL(reset_devices);
181 
182 static int __init set_reset_devices(char *str)
183 {
184 	reset_devices = 1;
185 	return 1;
186 }
187 
188 __setup("reset_devices", set_reset_devices);
189 
190 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
191 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
192 static const char *panic_later, *panic_param;
193 
194 extern const struct obs_kernel_param __setup_start[], __setup_end[];
195 
196 static bool __init obsolete_checksetup(char *line)
197 {
198 	const struct obs_kernel_param *p;
199 	bool had_early_param = false;
200 
201 	p = __setup_start;
202 	do {
203 		int n = strlen(p->str);
204 		if (parameqn(line, p->str, n)) {
205 			if (p->early) {
206 				/* Already done in parse_early_param?
207 				 * (Needs exact match on param part).
208 				 * Keep iterating, as we can have early
209 				 * params and __setups of same names 8( */
210 				if (line[n] == '\0' || line[n] == '=')
211 					had_early_param = true;
212 			} else if (!p->setup_func) {
213 				pr_warn("Parameter %s is obsolete, ignored\n",
214 					p->str);
215 				return true;
216 			} else if (p->setup_func(line + n))
217 				return true;
218 		}
219 		p++;
220 	} while (p < __setup_end);
221 
222 	return had_early_param;
223 }
224 
225 /*
226  * This should be approx 2 Bo*oMips to start (note initial shift), and will
227  * still work even if initially too large, it will just take slightly longer
228  */
229 unsigned long loops_per_jiffy = (1<<12);
230 EXPORT_SYMBOL(loops_per_jiffy);
231 
232 static int __init debug_kernel(char *str)
233 {
234 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
235 	return 0;
236 }
237 
238 static int __init quiet_kernel(char *str)
239 {
240 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
241 	return 0;
242 }
243 
244 early_param("debug", debug_kernel);
245 early_param("quiet", quiet_kernel);
246 
247 static int __init loglevel(char *str)
248 {
249 	int newlevel;
250 
251 	/*
252 	 * Only update loglevel value when a correct setting was passed,
253 	 * to prevent blind crashes (when loglevel being set to 0) that
254 	 * are quite hard to debug
255 	 */
256 	if (get_option(&str, &newlevel)) {
257 		console_loglevel = newlevel;
258 		return 0;
259 	}
260 
261 	return -EINVAL;
262 }
263 
264 early_param("loglevel", loglevel);
265 
266 #ifdef CONFIG_BLK_DEV_INITRD
267 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
268 {
269 	u32 size, csum;
270 	char *data;
271 	u32 *hdr;
272 	int i;
273 
274 	if (!initrd_end)
275 		return NULL;
276 
277 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
278 	/*
279 	 * Since Grub may align the size of initrd to 4, we must
280 	 * check the preceding 3 bytes as well.
281 	 */
282 	for (i = 0; i < 4; i++) {
283 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
284 			goto found;
285 		data--;
286 	}
287 	return NULL;
288 
289 found:
290 	hdr = (u32 *)(data - 8);
291 	size = le32_to_cpu(hdr[0]);
292 	csum = le32_to_cpu(hdr[1]);
293 
294 	data = ((void *)hdr) - size;
295 	if ((unsigned long)data < initrd_start) {
296 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
297 			size, initrd_end - initrd_start);
298 		return NULL;
299 	}
300 
301 	/* Remove bootconfig from initramfs/initrd */
302 	initrd_end = (unsigned long)data;
303 	if (_size)
304 		*_size = size;
305 	if (_csum)
306 		*_csum = csum;
307 
308 	return data;
309 }
310 #else
311 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
312 {
313 	return NULL;
314 }
315 #endif
316 
317 #ifdef CONFIG_BOOT_CONFIG
318 
319 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
320 
321 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
322 
323 static int __init xbc_snprint_cmdline(char *buf, size_t size,
324 				      struct xbc_node *root)
325 {
326 	struct xbc_node *knode, *vnode;
327 	char *end = buf + size;
328 	const char *val;
329 	int ret;
330 
331 	xbc_node_for_each_key_value(root, knode, val) {
332 		ret = xbc_node_compose_key_after(root, knode,
333 					xbc_namebuf, XBC_KEYLEN_MAX);
334 		if (ret < 0)
335 			return ret;
336 
337 		vnode = xbc_node_get_child(knode);
338 		if (!vnode) {
339 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
340 			if (ret < 0)
341 				return ret;
342 			buf += ret;
343 			continue;
344 		}
345 		xbc_array_for_each_value(vnode, val) {
346 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
347 				       xbc_namebuf, val);
348 			if (ret < 0)
349 				return ret;
350 			buf += ret;
351 		}
352 	}
353 
354 	return buf - (end - size);
355 }
356 #undef rest
357 
358 /* Make an extra command line under given key word */
359 static char * __init xbc_make_cmdline(const char *key)
360 {
361 	struct xbc_node *root;
362 	char *new_cmdline;
363 	int ret, len = 0;
364 
365 	root = xbc_find_node(key);
366 	if (!root)
367 		return NULL;
368 
369 	/* Count required buffer size */
370 	len = xbc_snprint_cmdline(NULL, 0, root);
371 	if (len <= 0)
372 		return NULL;
373 
374 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
375 	if (!new_cmdline) {
376 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
377 		return NULL;
378 	}
379 
380 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
381 	if (ret < 0 || ret > len) {
382 		pr_err("Failed to print extra kernel cmdline.\n");
383 		return NULL;
384 	}
385 
386 	return new_cmdline;
387 }
388 
389 static u32 boot_config_checksum(unsigned char *p, u32 size)
390 {
391 	u32 ret = 0;
392 
393 	while (size--)
394 		ret += *p++;
395 
396 	return ret;
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 void __init setup_boot_config(void)
409 {
410 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
411 	const char *msg;
412 	int pos;
413 	u32 size, csum;
414 	char *data, *copy, *err;
415 	int ret;
416 
417 	/* Cut out the bootconfig data even if we have no bootconfig option */
418 	data = get_boot_config_from_initrd(&size, &csum);
419 
420 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
421 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
422 			 bootconfig_params);
423 
424 	if (IS_ERR(err) || !bootconfig_found)
425 		return;
426 
427 	/* parse_args() stops at '--' and returns an address */
428 	if (err)
429 		initargs_found = true;
430 
431 	if (!data) {
432 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
433 		return;
434 	}
435 
436 	if (size >= XBC_DATA_MAX) {
437 		pr_err("bootconfig size %d greater than max size %d\n",
438 			size, XBC_DATA_MAX);
439 		return;
440 	}
441 
442 	if (boot_config_checksum((unsigned char *)data, size) != csum) {
443 		pr_err("bootconfig checksum failed\n");
444 		return;
445 	}
446 
447 	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
448 	if (!copy) {
449 		pr_err("Failed to allocate memory for bootconfig\n");
450 		return;
451 	}
452 
453 	memcpy(copy, data, size);
454 	copy[size] = '\0';
455 
456 	ret = xbc_init(copy, &msg, &pos);
457 	if (ret < 0) {
458 		if (pos < 0)
459 			pr_err("Failed to init bootconfig: %s.\n", msg);
460 		else
461 			pr_err("Failed to parse bootconfig: %s at %d.\n",
462 				msg, pos);
463 	} else {
464 		pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
465 		/* keys starting with "kernel." are passed via cmdline */
466 		extra_command_line = xbc_make_cmdline("kernel");
467 		/* Also, "init." keys are init arguments */
468 		extra_init_args = xbc_make_cmdline("init");
469 	}
470 	return;
471 }
472 
473 #else
474 
475 static void __init setup_boot_config(void)
476 {
477 	/* Remove bootconfig data from initrd */
478 	get_boot_config_from_initrd(NULL, NULL);
479 }
480 
481 static int __init warn_bootconfig(char *str)
482 {
483 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
484 	return 0;
485 }
486 early_param("bootconfig", warn_bootconfig);
487 
488 #endif
489 
490 /* Change NUL term back to "=", to make "param" the whole string. */
491 static void __init repair_env_string(char *param, char *val)
492 {
493 	if (val) {
494 		/* param=val or param="val"? */
495 		if (val == param+strlen(param)+1)
496 			val[-1] = '=';
497 		else if (val == param+strlen(param)+2) {
498 			val[-2] = '=';
499 			memmove(val-1, val, strlen(val)+1);
500 		} else
501 			BUG();
502 	}
503 }
504 
505 /* Anything after -- gets handed straight to init. */
506 static int __init set_init_arg(char *param, char *val,
507 			       const char *unused, void *arg)
508 {
509 	unsigned int i;
510 
511 	if (panic_later)
512 		return 0;
513 
514 	repair_env_string(param, val);
515 
516 	for (i = 0; argv_init[i]; i++) {
517 		if (i == MAX_INIT_ARGS) {
518 			panic_later = "init";
519 			panic_param = param;
520 			return 0;
521 		}
522 	}
523 	argv_init[i] = param;
524 	return 0;
525 }
526 
527 /*
528  * Unknown boot options get handed to init, unless they look like
529  * unused parameters (modprobe will find them in /proc/cmdline).
530  */
531 static int __init unknown_bootoption(char *param, char *val,
532 				     const char *unused, void *arg)
533 {
534 	size_t len = strlen(param);
535 
536 	repair_env_string(param, val);
537 
538 	/* Handle obsolete-style parameters */
539 	if (obsolete_checksetup(param))
540 		return 0;
541 
542 	/* Unused module parameter. */
543 	if (strnchr(param, len, '.'))
544 		return 0;
545 
546 	if (panic_later)
547 		return 0;
548 
549 	if (val) {
550 		/* Environment option */
551 		unsigned int i;
552 		for (i = 0; envp_init[i]; i++) {
553 			if (i == MAX_INIT_ENVS) {
554 				panic_later = "env";
555 				panic_param = param;
556 			}
557 			if (!strncmp(param, envp_init[i], len+1))
558 				break;
559 		}
560 		envp_init[i] = param;
561 	} else {
562 		/* Command line option */
563 		unsigned int i;
564 		for (i = 0; argv_init[i]; i++) {
565 			if (i == MAX_INIT_ARGS) {
566 				panic_later = "init";
567 				panic_param = param;
568 			}
569 		}
570 		argv_init[i] = param;
571 	}
572 	return 0;
573 }
574 
575 static int __init init_setup(char *str)
576 {
577 	unsigned int i;
578 
579 	execute_command = str;
580 	/*
581 	 * In case LILO is going to boot us with default command line,
582 	 * it prepends "auto" before the whole cmdline which makes
583 	 * the shell think it should execute a script with such name.
584 	 * So we ignore all arguments entered _before_ init=... [MJ]
585 	 */
586 	for (i = 1; i < MAX_INIT_ARGS; i++)
587 		argv_init[i] = NULL;
588 	return 1;
589 }
590 __setup("init=", init_setup);
591 
592 static int __init rdinit_setup(char *str)
593 {
594 	unsigned int i;
595 
596 	ramdisk_execute_command = str;
597 	/* See "auto" comment in init_setup */
598 	for (i = 1; i < MAX_INIT_ARGS; i++)
599 		argv_init[i] = NULL;
600 	return 1;
601 }
602 __setup("rdinit=", rdinit_setup);
603 
604 #ifndef CONFIG_SMP
605 static const unsigned int setup_max_cpus = NR_CPUS;
606 static inline void setup_nr_cpu_ids(void) { }
607 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
608 #endif
609 
610 /*
611  * We need to store the untouched command line for future reference.
612  * We also need to store the touched command line since the parameter
613  * parsing is performed in place, and we should allow a component to
614  * store reference of name/value for future reference.
615  */
616 static void __init setup_command_line(char *command_line)
617 {
618 	size_t len, xlen = 0, ilen = 0;
619 
620 	if (extra_command_line)
621 		xlen = strlen(extra_command_line);
622 	if (extra_init_args)
623 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
624 
625 	len = xlen + strlen(boot_command_line) + 1;
626 
627 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
628 	if (!saved_command_line)
629 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
630 
631 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
632 	if (!static_command_line)
633 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
634 
635 	if (xlen) {
636 		/*
637 		 * We have to put extra_command_line before boot command
638 		 * lines because there could be dashes (separator of init
639 		 * command line) in the command lines.
640 		 */
641 		strcpy(saved_command_line, extra_command_line);
642 		strcpy(static_command_line, extra_command_line);
643 	}
644 	strcpy(saved_command_line + xlen, boot_command_line);
645 	strcpy(static_command_line + xlen, command_line);
646 
647 	if (ilen) {
648 		/*
649 		 * Append supplemental init boot args to saved_command_line
650 		 * so that user can check what command line options passed
651 		 * to init.
652 		 */
653 		len = strlen(saved_command_line);
654 		if (initargs_found) {
655 			saved_command_line[len++] = ' ';
656 		} else {
657 			strcpy(saved_command_line + len, " -- ");
658 			len += 4;
659 		}
660 
661 		strcpy(saved_command_line + len, extra_init_args);
662 	}
663 }
664 
665 /*
666  * We need to finalize in a non-__init function or else race conditions
667  * between the root thread and the init thread may cause start_kernel to
668  * be reaped by free_initmem before the root thread has proceeded to
669  * cpu_idle.
670  *
671  * gcc-3.4 accidentally inlines this function, so use noinline.
672  */
673 
674 static __initdata DECLARE_COMPLETION(kthreadd_done);
675 
676 noinline void __ref rest_init(void)
677 {
678 	struct task_struct *tsk;
679 	int pid;
680 
681 	rcu_scheduler_starting();
682 	/*
683 	 * We need to spawn init first so that it obtains pid 1, however
684 	 * the init task will end up wanting to create kthreads, which, if
685 	 * we schedule it before we create kthreadd, will OOPS.
686 	 */
687 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
688 	/*
689 	 * Pin init on the boot CPU. Task migration is not properly working
690 	 * until sched_init_smp() has been run. It will set the allowed
691 	 * CPUs for init to the non isolated CPUs.
692 	 */
693 	rcu_read_lock();
694 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
695 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
696 	rcu_read_unlock();
697 
698 	numa_default_policy();
699 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
700 	rcu_read_lock();
701 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
702 	rcu_read_unlock();
703 
704 	/*
705 	 * Enable might_sleep() and smp_processor_id() checks.
706 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
707 	 * kernel_thread() would trigger might_sleep() splats. With
708 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
709 	 * already, but it's stuck on the kthreadd_done completion.
710 	 */
711 	system_state = SYSTEM_SCHEDULING;
712 
713 	complete(&kthreadd_done);
714 
715 	/*
716 	 * The boot idle thread must execute schedule()
717 	 * at least once to get things moving:
718 	 */
719 	schedule_preempt_disabled();
720 	/* Call into cpu_idle with preempt disabled */
721 	cpu_startup_entry(CPUHP_ONLINE);
722 }
723 
724 /* Check for early params. */
725 static int __init do_early_param(char *param, char *val,
726 				 const char *unused, void *arg)
727 {
728 	const struct obs_kernel_param *p;
729 
730 	for (p = __setup_start; p < __setup_end; p++) {
731 		if ((p->early && parameq(param, p->str)) ||
732 		    (strcmp(param, "console") == 0 &&
733 		     strcmp(p->str, "earlycon") == 0)
734 		) {
735 			if (p->setup_func(val) != 0)
736 				pr_warn("Malformed early option '%s'\n", param);
737 		}
738 	}
739 	/* We accept everything at this stage. */
740 	return 0;
741 }
742 
743 void __init parse_early_options(char *cmdline)
744 {
745 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
746 		   do_early_param);
747 }
748 
749 /* Arch code calls this early on, or if not, just before other parsing. */
750 void __init parse_early_param(void)
751 {
752 	static int done __initdata;
753 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
754 
755 	if (done)
756 		return;
757 
758 	/* All fall through to do_early_param. */
759 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
760 	parse_early_options(tmp_cmdline);
761 	done = 1;
762 }
763 
764 void __init __weak arch_post_acpi_subsys_init(void) { }
765 
766 void __init __weak smp_setup_processor_id(void)
767 {
768 }
769 
770 # if THREAD_SIZE >= PAGE_SIZE
771 void __init __weak thread_stack_cache_init(void)
772 {
773 }
774 #endif
775 
776 void __init __weak mem_encrypt_init(void) { }
777 
778 void __init __weak poking_init(void) { }
779 
780 void __init __weak pgtable_cache_init(void) { }
781 
782 bool initcall_debug;
783 core_param(initcall_debug, initcall_debug, bool, 0644);
784 
785 #ifdef TRACEPOINTS_ENABLED
786 static void __init initcall_debug_enable(void);
787 #else
788 static inline void initcall_debug_enable(void)
789 {
790 }
791 #endif
792 
793 /* Report memory auto-initialization states for this boot. */
794 static void __init report_meminit(void)
795 {
796 	const char *stack;
797 
798 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
799 		stack = "all(pattern)";
800 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
801 		stack = "all(zero)";
802 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
803 		stack = "byref_all(zero)";
804 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
805 		stack = "byref(zero)";
806 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
807 		stack = "__user(zero)";
808 	else
809 		stack = "off";
810 
811 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
812 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
813 		want_init_on_free() ? "on" : "off");
814 	if (want_init_on_free())
815 		pr_info("mem auto-init: clearing system memory may take some time...\n");
816 }
817 
818 /*
819  * Set up kernel memory allocators
820  */
821 static void __init mm_init(void)
822 {
823 	/*
824 	 * page_ext requires contiguous pages,
825 	 * bigger than MAX_ORDER unless SPARSEMEM.
826 	 */
827 	page_ext_init_flatmem();
828 	init_mem_debugging_and_hardening();
829 	kfence_alloc_pool();
830 	report_meminit();
831 	stack_depot_init();
832 	mem_init();
833 	mem_init_print_info();
834 	/* page_owner must be initialized after buddy is ready */
835 	page_ext_init_flatmem_late();
836 	kmem_cache_init();
837 	kmemleak_init();
838 	pgtable_init();
839 	debug_objects_mem_init();
840 	vmalloc_init();
841 	/* Should be run before the first non-init thread is created */
842 	init_espfix_bsp();
843 	/* Should be run after espfix64 is set up. */
844 	pti_init();
845 }
846 
847 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
848 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
849 			   randomize_kstack_offset);
850 DEFINE_PER_CPU(u32, kstack_offset);
851 
852 static int __init early_randomize_kstack_offset(char *buf)
853 {
854 	int ret;
855 	bool bool_result;
856 
857 	ret = kstrtobool(buf, &bool_result);
858 	if (ret)
859 		return ret;
860 
861 	if (bool_result)
862 		static_branch_enable(&randomize_kstack_offset);
863 	else
864 		static_branch_disable(&randomize_kstack_offset);
865 	return 0;
866 }
867 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
868 #endif
869 
870 void __init __weak arch_call_rest_init(void)
871 {
872 	rest_init();
873 }
874 
875 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
876 {
877 	char *command_line;
878 	char *after_dashes;
879 
880 	set_task_stack_end_magic(&init_task);
881 	smp_setup_processor_id();
882 	debug_objects_early_init();
883 
884 	cgroup_init_early();
885 
886 	local_irq_disable();
887 	early_boot_irqs_disabled = true;
888 
889 	/*
890 	 * Interrupts are still disabled. Do necessary setups, then
891 	 * enable them.
892 	 */
893 	boot_cpu_init();
894 	page_address_init();
895 	pr_notice("%s", linux_banner);
896 	early_security_init();
897 	setup_arch(&command_line);
898 	setup_boot_config();
899 	setup_command_line(command_line);
900 	setup_nr_cpu_ids();
901 	setup_per_cpu_areas();
902 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
903 	boot_cpu_hotplug_init();
904 
905 	build_all_zonelists(NULL);
906 	page_alloc_init();
907 
908 	pr_notice("Kernel command line: %s\n", saved_command_line);
909 	/* parameters may set static keys */
910 	jump_label_init();
911 	parse_early_param();
912 	after_dashes = parse_args("Booting kernel",
913 				  static_command_line, __start___param,
914 				  __stop___param - __start___param,
915 				  -1, -1, NULL, &unknown_bootoption);
916 	if (!IS_ERR_OR_NULL(after_dashes))
917 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
918 			   NULL, set_init_arg);
919 	if (extra_init_args)
920 		parse_args("Setting extra init args", extra_init_args,
921 			   NULL, 0, -1, -1, NULL, set_init_arg);
922 
923 	/*
924 	 * These use large bootmem allocations and must precede
925 	 * kmem_cache_init()
926 	 */
927 	setup_log_buf(0);
928 	vfs_caches_init_early();
929 	sort_main_extable();
930 	trap_init();
931 	mm_init();
932 
933 	ftrace_init();
934 
935 	/* trace_printk can be enabled here */
936 	early_trace_init();
937 
938 	/*
939 	 * Set up the scheduler prior starting any interrupts (such as the
940 	 * timer interrupt). Full topology setup happens at smp_init()
941 	 * time - but meanwhile we still have a functioning scheduler.
942 	 */
943 	sched_init();
944 	/*
945 	 * Disable preemption - early bootup scheduling is extremely
946 	 * fragile until we cpu_idle() for the first time.
947 	 */
948 	preempt_disable();
949 	if (WARN(!irqs_disabled(),
950 		 "Interrupts were enabled *very* early, fixing it\n"))
951 		local_irq_disable();
952 	radix_tree_init();
953 
954 	/*
955 	 * Set up housekeeping before setting up workqueues to allow the unbound
956 	 * workqueue to take non-housekeeping into account.
957 	 */
958 	housekeeping_init();
959 
960 	/*
961 	 * Allow workqueue creation and work item queueing/cancelling
962 	 * early.  Work item execution depends on kthreads and starts after
963 	 * workqueue_init().
964 	 */
965 	workqueue_init_early();
966 
967 	rcu_init();
968 
969 	/* Trace events are available after this */
970 	trace_init();
971 
972 	if (initcall_debug)
973 		initcall_debug_enable();
974 
975 	context_tracking_init();
976 	/* init some links before init_ISA_irqs() */
977 	early_irq_init();
978 	init_IRQ();
979 	tick_init();
980 	rcu_init_nohz();
981 	init_timers();
982 	hrtimers_init();
983 	softirq_init();
984 	timekeeping_init();
985 	kfence_init();
986 
987 	/*
988 	 * For best initial stack canary entropy, prepare it after:
989 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
990 	 * - timekeeping_init() for ktime entropy used in rand_initialize()
991 	 * - rand_initialize() to get any arch-specific entropy like RDRAND
992 	 * - add_latent_entropy() to get any latent entropy
993 	 * - adding command line entropy
994 	 */
995 	rand_initialize();
996 	add_latent_entropy();
997 	add_device_randomness(command_line, strlen(command_line));
998 	boot_init_stack_canary();
999 
1000 	time_init();
1001 	perf_event_init();
1002 	profile_init();
1003 	call_function_init();
1004 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1005 
1006 	early_boot_irqs_disabled = false;
1007 	local_irq_enable();
1008 
1009 	kmem_cache_init_late();
1010 
1011 	/*
1012 	 * HACK ALERT! This is early. We're enabling the console before
1013 	 * we've done PCI setups etc, and console_init() must be aware of
1014 	 * this. But we do want output early, in case something goes wrong.
1015 	 */
1016 	console_init();
1017 	if (panic_later)
1018 		panic("Too many boot %s vars at `%s'", panic_later,
1019 		      panic_param);
1020 
1021 	lockdep_init();
1022 
1023 	/*
1024 	 * Need to run this when irqs are enabled, because it wants
1025 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1026 	 * too:
1027 	 */
1028 	locking_selftest();
1029 
1030 	/*
1031 	 * This needs to be called before any devices perform DMA
1032 	 * operations that might use the SWIOTLB bounce buffers. It will
1033 	 * mark the bounce buffers as decrypted so that their usage will
1034 	 * not cause "plain-text" data to be decrypted when accessed.
1035 	 */
1036 	mem_encrypt_init();
1037 
1038 #ifdef CONFIG_BLK_DEV_INITRD
1039 	if (initrd_start && !initrd_below_start_ok &&
1040 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1041 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1042 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1043 		    min_low_pfn);
1044 		initrd_start = 0;
1045 	}
1046 #endif
1047 	setup_per_cpu_pageset();
1048 	numa_policy_init();
1049 	acpi_early_init();
1050 	if (late_time_init)
1051 		late_time_init();
1052 	sched_clock_init();
1053 	calibrate_delay();
1054 	pid_idr_init();
1055 	anon_vma_init();
1056 #ifdef CONFIG_X86
1057 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1058 		efi_enter_virtual_mode();
1059 #endif
1060 	thread_stack_cache_init();
1061 	cred_init();
1062 	fork_init();
1063 	proc_caches_init();
1064 	uts_ns_init();
1065 	key_init();
1066 	security_init();
1067 	dbg_late_init();
1068 	vfs_caches_init();
1069 	pagecache_init();
1070 	signals_init();
1071 	seq_file_init();
1072 	proc_root_init();
1073 	nsfs_init();
1074 	cpuset_init();
1075 	cgroup_init();
1076 	taskstats_init_early();
1077 	delayacct_init();
1078 
1079 	poking_init();
1080 	check_bugs();
1081 
1082 	acpi_subsystem_init();
1083 	arch_post_acpi_subsys_init();
1084 	kcsan_init();
1085 
1086 	/* Do the rest non-__init'ed, we're now alive */
1087 	arch_call_rest_init();
1088 
1089 	prevent_tail_call_optimization();
1090 }
1091 
1092 /* Call all constructor functions linked into the kernel. */
1093 static void __init do_ctors(void)
1094 {
1095 /*
1096  * For UML, the constructors have already been called by the
1097  * normal setup code as it's just a normal ELF binary, so we
1098  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1099  * even on UML for modules.
1100  */
1101 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1102 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1103 
1104 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1105 		(*fn)();
1106 #endif
1107 }
1108 
1109 #ifdef CONFIG_KALLSYMS
1110 struct blacklist_entry {
1111 	struct list_head next;
1112 	char *buf;
1113 };
1114 
1115 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1116 
1117 static int __init initcall_blacklist(char *str)
1118 {
1119 	char *str_entry;
1120 	struct blacklist_entry *entry;
1121 
1122 	/* str argument is a comma-separated list of functions */
1123 	do {
1124 		str_entry = strsep(&str, ",");
1125 		if (str_entry) {
1126 			pr_debug("blacklisting initcall %s\n", str_entry);
1127 			entry = memblock_alloc(sizeof(*entry),
1128 					       SMP_CACHE_BYTES);
1129 			if (!entry)
1130 				panic("%s: Failed to allocate %zu bytes\n",
1131 				      __func__, sizeof(*entry));
1132 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1133 						    SMP_CACHE_BYTES);
1134 			if (!entry->buf)
1135 				panic("%s: Failed to allocate %zu bytes\n",
1136 				      __func__, strlen(str_entry) + 1);
1137 			strcpy(entry->buf, str_entry);
1138 			list_add(&entry->next, &blacklisted_initcalls);
1139 		}
1140 	} while (str_entry);
1141 
1142 	return 0;
1143 }
1144 
1145 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1146 {
1147 	struct blacklist_entry *entry;
1148 	char fn_name[KSYM_SYMBOL_LEN];
1149 	unsigned long addr;
1150 
1151 	if (list_empty(&blacklisted_initcalls))
1152 		return false;
1153 
1154 	addr = (unsigned long) dereference_function_descriptor(fn);
1155 	sprint_symbol_no_offset(fn_name, addr);
1156 
1157 	/*
1158 	 * fn will be "function_name [module_name]" where [module_name] is not
1159 	 * displayed for built-in init functions.  Strip off the [module_name].
1160 	 */
1161 	strreplace(fn_name, ' ', '\0');
1162 
1163 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1164 		if (!strcmp(fn_name, entry->buf)) {
1165 			pr_debug("initcall %s blacklisted\n", fn_name);
1166 			return true;
1167 		}
1168 	}
1169 
1170 	return false;
1171 }
1172 #else
1173 static int __init initcall_blacklist(char *str)
1174 {
1175 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1176 	return 0;
1177 }
1178 
1179 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1180 {
1181 	return false;
1182 }
1183 #endif
1184 __setup("initcall_blacklist=", initcall_blacklist);
1185 
1186 static __init_or_module void
1187 trace_initcall_start_cb(void *data, initcall_t fn)
1188 {
1189 	ktime_t *calltime = (ktime_t *)data;
1190 
1191 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1192 	*calltime = ktime_get();
1193 }
1194 
1195 static __init_or_module void
1196 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1197 {
1198 	ktime_t *calltime = (ktime_t *)data;
1199 	ktime_t delta, rettime;
1200 	unsigned long long duration;
1201 
1202 	rettime = ktime_get();
1203 	delta = ktime_sub(rettime, *calltime);
1204 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1205 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1206 		 fn, ret, duration);
1207 }
1208 
1209 static ktime_t initcall_calltime;
1210 
1211 #ifdef TRACEPOINTS_ENABLED
1212 static void __init initcall_debug_enable(void)
1213 {
1214 	int ret;
1215 
1216 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1217 					    &initcall_calltime);
1218 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1219 					      &initcall_calltime);
1220 	WARN(ret, "Failed to register initcall tracepoints\n");
1221 }
1222 # define do_trace_initcall_start	trace_initcall_start
1223 # define do_trace_initcall_finish	trace_initcall_finish
1224 #else
1225 static inline void do_trace_initcall_start(initcall_t fn)
1226 {
1227 	if (!initcall_debug)
1228 		return;
1229 	trace_initcall_start_cb(&initcall_calltime, fn);
1230 }
1231 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1232 {
1233 	if (!initcall_debug)
1234 		return;
1235 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1236 }
1237 #endif /* !TRACEPOINTS_ENABLED */
1238 
1239 int __init_or_module do_one_initcall(initcall_t fn)
1240 {
1241 	int count = preempt_count();
1242 	char msgbuf[64];
1243 	int ret;
1244 
1245 	if (initcall_blacklisted(fn))
1246 		return -EPERM;
1247 
1248 	do_trace_initcall_start(fn);
1249 	ret = fn();
1250 	do_trace_initcall_finish(fn, ret);
1251 
1252 	msgbuf[0] = 0;
1253 
1254 	if (preempt_count() != count) {
1255 		sprintf(msgbuf, "preemption imbalance ");
1256 		preempt_count_set(count);
1257 	}
1258 	if (irqs_disabled()) {
1259 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1260 		local_irq_enable();
1261 	}
1262 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1263 
1264 	add_latent_entropy();
1265 	return ret;
1266 }
1267 
1268 
1269 extern initcall_entry_t __initcall_start[];
1270 extern initcall_entry_t __initcall0_start[];
1271 extern initcall_entry_t __initcall1_start[];
1272 extern initcall_entry_t __initcall2_start[];
1273 extern initcall_entry_t __initcall3_start[];
1274 extern initcall_entry_t __initcall4_start[];
1275 extern initcall_entry_t __initcall5_start[];
1276 extern initcall_entry_t __initcall6_start[];
1277 extern initcall_entry_t __initcall7_start[];
1278 extern initcall_entry_t __initcall_end[];
1279 
1280 static initcall_entry_t *initcall_levels[] __initdata = {
1281 	__initcall0_start,
1282 	__initcall1_start,
1283 	__initcall2_start,
1284 	__initcall3_start,
1285 	__initcall4_start,
1286 	__initcall5_start,
1287 	__initcall6_start,
1288 	__initcall7_start,
1289 	__initcall_end,
1290 };
1291 
1292 /* Keep these in sync with initcalls in include/linux/init.h */
1293 static const char *initcall_level_names[] __initdata = {
1294 	"pure",
1295 	"core",
1296 	"postcore",
1297 	"arch",
1298 	"subsys",
1299 	"fs",
1300 	"device",
1301 	"late",
1302 };
1303 
1304 static int __init ignore_unknown_bootoption(char *param, char *val,
1305 			       const char *unused, void *arg)
1306 {
1307 	return 0;
1308 }
1309 
1310 static void __init do_initcall_level(int level, char *command_line)
1311 {
1312 	initcall_entry_t *fn;
1313 
1314 	parse_args(initcall_level_names[level],
1315 		   command_line, __start___param,
1316 		   __stop___param - __start___param,
1317 		   level, level,
1318 		   NULL, ignore_unknown_bootoption);
1319 
1320 	trace_initcall_level(initcall_level_names[level]);
1321 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1322 		do_one_initcall(initcall_from_entry(fn));
1323 }
1324 
1325 static void __init do_initcalls(void)
1326 {
1327 	int level;
1328 	size_t len = strlen(saved_command_line) + 1;
1329 	char *command_line;
1330 
1331 	command_line = kzalloc(len, GFP_KERNEL);
1332 	if (!command_line)
1333 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1334 
1335 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1336 		/* Parser modifies command_line, restore it each time */
1337 		strcpy(command_line, saved_command_line);
1338 		do_initcall_level(level, command_line);
1339 	}
1340 
1341 	kfree(command_line);
1342 }
1343 
1344 /*
1345  * Ok, the machine is now initialized. None of the devices
1346  * have been touched yet, but the CPU subsystem is up and
1347  * running, and memory and process management works.
1348  *
1349  * Now we can finally start doing some real work..
1350  */
1351 static void __init do_basic_setup(void)
1352 {
1353 	cpuset_init_smp();
1354 	driver_init();
1355 	init_irq_proc();
1356 	do_ctors();
1357 	usermodehelper_enable();
1358 	do_initcalls();
1359 }
1360 
1361 static void __init do_pre_smp_initcalls(void)
1362 {
1363 	initcall_entry_t *fn;
1364 
1365 	trace_initcall_level("early");
1366 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1367 		do_one_initcall(initcall_from_entry(fn));
1368 }
1369 
1370 static int run_init_process(const char *init_filename)
1371 {
1372 	const char *const *p;
1373 
1374 	argv_init[0] = init_filename;
1375 	pr_info("Run %s as init process\n", init_filename);
1376 	pr_debug("  with arguments:\n");
1377 	for (p = argv_init; *p; p++)
1378 		pr_debug("    %s\n", *p);
1379 	pr_debug("  with environment:\n");
1380 	for (p = envp_init; *p; p++)
1381 		pr_debug("    %s\n", *p);
1382 	return kernel_execve(init_filename, argv_init, envp_init);
1383 }
1384 
1385 static int try_to_run_init_process(const char *init_filename)
1386 {
1387 	int ret;
1388 
1389 	ret = run_init_process(init_filename);
1390 
1391 	if (ret && ret != -ENOENT) {
1392 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1393 		       init_filename, ret);
1394 	}
1395 
1396 	return ret;
1397 }
1398 
1399 static noinline void __init kernel_init_freeable(void);
1400 
1401 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1402 bool rodata_enabled __ro_after_init = true;
1403 static int __init set_debug_rodata(char *str)
1404 {
1405 	return strtobool(str, &rodata_enabled);
1406 }
1407 __setup("rodata=", set_debug_rodata);
1408 #endif
1409 
1410 #ifdef CONFIG_STRICT_KERNEL_RWX
1411 static void mark_readonly(void)
1412 {
1413 	if (rodata_enabled) {
1414 		/*
1415 		 * load_module() results in W+X mappings, which are cleaned
1416 		 * up with call_rcu().  Let's make sure that queued work is
1417 		 * flushed so that we don't hit false positives looking for
1418 		 * insecure pages which are W+X.
1419 		 */
1420 		rcu_barrier();
1421 		mark_rodata_ro();
1422 		rodata_test();
1423 	} else
1424 		pr_info("Kernel memory protection disabled.\n");
1425 }
1426 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1427 static inline void mark_readonly(void)
1428 {
1429 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1430 }
1431 #else
1432 static inline void mark_readonly(void)
1433 {
1434 	pr_warn("This architecture does not have kernel memory protection.\n");
1435 }
1436 #endif
1437 
1438 void __weak free_initmem(void)
1439 {
1440 	free_initmem_default(POISON_FREE_INITMEM);
1441 }
1442 
1443 static int __ref kernel_init(void *unused)
1444 {
1445 	int ret;
1446 
1447 	kernel_init_freeable();
1448 	/* need to finish all async __init code before freeing the memory */
1449 	async_synchronize_full();
1450 	kprobe_free_init_mem();
1451 	ftrace_free_init_mem();
1452 	kgdb_free_init_mem();
1453 	free_initmem();
1454 	mark_readonly();
1455 
1456 	/*
1457 	 * Kernel mappings are now finalized - update the userspace page-table
1458 	 * to finalize PTI.
1459 	 */
1460 	pti_finalize();
1461 
1462 	system_state = SYSTEM_RUNNING;
1463 	numa_default_policy();
1464 
1465 	rcu_end_inkernel_boot();
1466 
1467 	do_sysctl_args();
1468 
1469 	if (ramdisk_execute_command) {
1470 		ret = run_init_process(ramdisk_execute_command);
1471 		if (!ret)
1472 			return 0;
1473 		pr_err("Failed to execute %s (error %d)\n",
1474 		       ramdisk_execute_command, ret);
1475 	}
1476 
1477 	/*
1478 	 * We try each of these until one succeeds.
1479 	 *
1480 	 * The Bourne shell can be used instead of init if we are
1481 	 * trying to recover a really broken machine.
1482 	 */
1483 	if (execute_command) {
1484 		ret = run_init_process(execute_command);
1485 		if (!ret)
1486 			return 0;
1487 		panic("Requested init %s failed (error %d).",
1488 		      execute_command, ret);
1489 	}
1490 
1491 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1492 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1493 		if (ret)
1494 			pr_err("Default init %s failed (error %d)\n",
1495 			       CONFIG_DEFAULT_INIT, ret);
1496 		else
1497 			return 0;
1498 	}
1499 
1500 	if (!try_to_run_init_process("/sbin/init") ||
1501 	    !try_to_run_init_process("/etc/init") ||
1502 	    !try_to_run_init_process("/bin/init") ||
1503 	    !try_to_run_init_process("/bin/sh"))
1504 		return 0;
1505 
1506 	panic("No working init found.  Try passing init= option to kernel. "
1507 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1508 }
1509 
1510 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1511 void __init console_on_rootfs(void)
1512 {
1513 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1514 
1515 	if (IS_ERR(file)) {
1516 		pr_err("Warning: unable to open an initial console.\n");
1517 		return;
1518 	}
1519 	init_dup(file);
1520 	init_dup(file);
1521 	init_dup(file);
1522 	fput(file);
1523 }
1524 
1525 static noinline void __init kernel_init_freeable(void)
1526 {
1527 	/*
1528 	 * Wait until kthreadd is all set-up.
1529 	 */
1530 	wait_for_completion(&kthreadd_done);
1531 
1532 	/* Now the scheduler is fully set up and can do blocking allocations */
1533 	gfp_allowed_mask = __GFP_BITS_MASK;
1534 
1535 	/*
1536 	 * init can allocate pages on any node
1537 	 */
1538 	set_mems_allowed(node_states[N_MEMORY]);
1539 
1540 	cad_pid = task_pid(current);
1541 
1542 	smp_prepare_cpus(setup_max_cpus);
1543 
1544 	workqueue_init();
1545 
1546 	init_mm_internals();
1547 
1548 	rcu_init_tasks_generic();
1549 	do_pre_smp_initcalls();
1550 	lockup_detector_init();
1551 
1552 	smp_init();
1553 	sched_init_smp();
1554 
1555 	padata_init();
1556 	page_alloc_init_late();
1557 	/* Initialize page ext after all struct pages are initialized. */
1558 	page_ext_init();
1559 
1560 	do_basic_setup();
1561 
1562 	kunit_run_all_tests();
1563 
1564 	wait_for_initramfs();
1565 	console_on_rootfs();
1566 
1567 	/*
1568 	 * check if there is an early userspace init.  If yes, let it do all
1569 	 * the work
1570 	 */
1571 	if (init_eaccess(ramdisk_execute_command) != 0) {
1572 		ramdisk_execute_command = NULL;
1573 		prepare_namespace();
1574 	}
1575 
1576 	/*
1577 	 * Ok, we have completed the initial bootup, and
1578 	 * we're essentially up and running. Get rid of the
1579 	 * initmem segments and start the user-mode stuff..
1580 	 *
1581 	 * rootfs is available now, try loading the public keys
1582 	 * and default modules
1583 	 */
1584 
1585 	integrity_load_keys();
1586 }
1587