xref: /linux/init/main.c (revision 9632e78e82648aa98340df78eab9106f63da151e)
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/srcu.h>
46 #include <linux/moduleparam.h>
47 #include <linux/kallsyms.h>
48 #include <linux/buildid.h>
49 #include <linux/writeback.h>
50 #include <linux/cpu.h>
51 #include <linux/cpuset.h>
52 #include <linux/cgroup.h>
53 #include <linux/efi.h>
54 #include <linux/tick.h>
55 #include <linux/sched/isolation.h>
56 #include <linux/interrupt.h>
57 #include <linux/taskstats_kern.h>
58 #include <linux/delayacct.h>
59 #include <linux/unistd.h>
60 #include <linux/utsname.h>
61 #include <linux/rmap.h>
62 #include <linux/mempolicy.h>
63 #include <linux/key.h>
64 #include <linux/page_ext.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/elevator.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/list.h>
93 #include <linux/integrity.h>
94 #include <linux/proc_ns.h>
95 #include <linux/io.h>
96 #include <linux/cache.h>
97 #include <linux/rodata_test.h>
98 #include <linux/jump_label.h>
99 #include <linux/mem_encrypt.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
103 
104 #include <asm/io.h>
105 #include <asm/bugs.h>
106 #include <asm/setup.h>
107 #include <asm/sections.h>
108 #include <asm/cacheflush.h>
109 
110 #define CREATE_TRACE_POINTS
111 #include <trace/events/initcall.h>
112 
113 #include <kunit/test.h>
114 
115 static int kernel_init(void *);
116 
117 extern void init_IRQ(void);
118 extern void radix_tree_init(void);
119 
120 /*
121  * Debug helper: via this flag we know that we are in 'early bootup code'
122  * where only the boot processor is running with IRQ disabled.  This means
123  * two things - IRQ must not be enabled before the flag is cleared and some
124  * operations which are not allowed with IRQ disabled are allowed while the
125  * flag is set.
126  */
127 bool early_boot_irqs_disabled __read_mostly;
128 
129 enum system_states system_state __read_mostly;
130 EXPORT_SYMBOL(system_state);
131 
132 /*
133  * Boot command-line arguments
134  */
135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
137 
138 extern void time_init(void);
139 /* Default late time init is NULL. archs can override this later. */
140 void (*__initdata late_time_init)(void);
141 
142 /* Untouched command line saved by arch-specific code. */
143 char __initdata boot_command_line[COMMAND_LINE_SIZE];
144 /* Untouched saved command line (eg. for /proc) */
145 char *saved_command_line;
146 /* Command line for parameter parsing */
147 static char *static_command_line;
148 /* Untouched extra command line */
149 static char *extra_command_line;
150 /* Extra init arguments */
151 static char *extra_init_args;
152 
153 #ifdef CONFIG_BOOT_CONFIG
154 /* Is bootconfig on command line? */
155 static bool bootconfig_found;
156 static bool initargs_found;
157 #else
158 # define bootconfig_found false
159 # define initargs_found false
160 #endif
161 
162 static char *execute_command;
163 static char *ramdisk_execute_command = "/init";
164 
165 /*
166  * Used to generate warnings if static_key manipulation functions are used
167  * before jump_label_init is called.
168  */
169 bool static_key_initialized __read_mostly;
170 EXPORT_SYMBOL_GPL(static_key_initialized);
171 
172 /*
173  * If set, this is an indication to the drivers that reset the underlying
174  * device before going ahead with the initialization otherwise driver might
175  * rely on the BIOS and skip the reset operation.
176  *
177  * This is useful if kernel is booting in an unreliable environment.
178  * For ex. kdump situation where previous kernel has crashed, BIOS has been
179  * skipped and devices will be in unknown state.
180  */
181 unsigned int reset_devices;
182 EXPORT_SYMBOL(reset_devices);
183 
184 static int __init set_reset_devices(char *str)
185 {
186 	reset_devices = 1;
187 	return 1;
188 }
189 
190 __setup("reset_devices", set_reset_devices);
191 
192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194 static const char *panic_later, *panic_param;
195 
196 extern const struct obs_kernel_param __setup_start[], __setup_end[];
197 
198 static bool __init obsolete_checksetup(char *line)
199 {
200 	const struct obs_kernel_param *p;
201 	bool had_early_param = false;
202 
203 	p = __setup_start;
204 	do {
205 		int n = strlen(p->str);
206 		if (parameqn(line, p->str, n)) {
207 			if (p->early) {
208 				/* Already done in parse_early_param?
209 				 * (Needs exact match on param part).
210 				 * Keep iterating, as we can have early
211 				 * params and __setups of same names 8( */
212 				if (line[n] == '\0' || line[n] == '=')
213 					had_early_param = true;
214 			} else if (!p->setup_func) {
215 				pr_warn("Parameter %s is obsolete, ignored\n",
216 					p->str);
217 				return true;
218 			} else if (p->setup_func(line + n))
219 				return true;
220 		}
221 		p++;
222 	} while (p < __setup_end);
223 
224 	return had_early_param;
225 }
226 
227 /*
228  * This should be approx 2 Bo*oMips to start (note initial shift), and will
229  * still work even if initially too large, it will just take slightly longer
230  */
231 unsigned long loops_per_jiffy = (1<<12);
232 EXPORT_SYMBOL(loops_per_jiffy);
233 
234 static int __init debug_kernel(char *str)
235 {
236 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
237 	return 0;
238 }
239 
240 static int __init quiet_kernel(char *str)
241 {
242 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
243 	return 0;
244 }
245 
246 early_param("debug", debug_kernel);
247 early_param("quiet", quiet_kernel);
248 
249 static int __init loglevel(char *str)
250 {
251 	int newlevel;
252 
253 	/*
254 	 * Only update loglevel value when a correct setting was passed,
255 	 * to prevent blind crashes (when loglevel being set to 0) that
256 	 * are quite hard to debug
257 	 */
258 	if (get_option(&str, &newlevel)) {
259 		console_loglevel = newlevel;
260 		return 0;
261 	}
262 
263 	return -EINVAL;
264 }
265 
266 early_param("loglevel", loglevel);
267 
268 #ifdef CONFIG_BLK_DEV_INITRD
269 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
270 {
271 	u32 size, csum;
272 	char *data;
273 	u32 *hdr;
274 	int i;
275 
276 	if (!initrd_end)
277 		return NULL;
278 
279 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
280 	/*
281 	 * Since Grub may align the size of initrd to 4, we must
282 	 * check the preceding 3 bytes as well.
283 	 */
284 	for (i = 0; i < 4; i++) {
285 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
286 			goto found;
287 		data--;
288 	}
289 	return NULL;
290 
291 found:
292 	hdr = (u32 *)(data - 8);
293 	size = le32_to_cpu(hdr[0]);
294 	csum = le32_to_cpu(hdr[1]);
295 
296 	data = ((void *)hdr) - size;
297 	if ((unsigned long)data < initrd_start) {
298 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
299 			size, initrd_end - initrd_start);
300 		return NULL;
301 	}
302 
303 	/* Remove bootconfig from initramfs/initrd */
304 	initrd_end = (unsigned long)data;
305 	if (_size)
306 		*_size = size;
307 	if (_csum)
308 		*_csum = csum;
309 
310 	return data;
311 }
312 #else
313 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
314 {
315 	return NULL;
316 }
317 #endif
318 
319 #ifdef CONFIG_BOOT_CONFIG
320 
321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
322 
323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
324 
325 static int __init xbc_snprint_cmdline(char *buf, size_t size,
326 				      struct xbc_node *root)
327 {
328 	struct xbc_node *knode, *vnode;
329 	char *end = buf + size;
330 	const char *val;
331 	int ret;
332 
333 	xbc_node_for_each_key_value(root, knode, val) {
334 		ret = xbc_node_compose_key_after(root, knode,
335 					xbc_namebuf, XBC_KEYLEN_MAX);
336 		if (ret < 0)
337 			return ret;
338 
339 		vnode = xbc_node_get_child(knode);
340 		if (!vnode) {
341 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
342 			if (ret < 0)
343 				return ret;
344 			buf += ret;
345 			continue;
346 		}
347 		xbc_array_for_each_value(vnode, val) {
348 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
349 				       xbc_namebuf, val);
350 			if (ret < 0)
351 				return ret;
352 			buf += ret;
353 		}
354 	}
355 
356 	return buf - (end - size);
357 }
358 #undef rest
359 
360 /* Make an extra command line under given key word */
361 static char * __init xbc_make_cmdline(const char *key)
362 {
363 	struct xbc_node *root;
364 	char *new_cmdline;
365 	int ret, len = 0;
366 
367 	root = xbc_find_node(key);
368 	if (!root)
369 		return NULL;
370 
371 	/* Count required buffer size */
372 	len = xbc_snprint_cmdline(NULL, 0, root);
373 	if (len <= 0)
374 		return NULL;
375 
376 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
377 	if (!new_cmdline) {
378 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
379 		return NULL;
380 	}
381 
382 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
383 	if (ret < 0 || ret > len) {
384 		pr_err("Failed to print extra kernel cmdline.\n");
385 		return NULL;
386 	}
387 
388 	return new_cmdline;
389 }
390 
391 static int __init bootconfig_params(char *param, char *val,
392 				    const char *unused, void *arg)
393 {
394 	if (strcmp(param, "bootconfig") == 0) {
395 		bootconfig_found = true;
396 	}
397 	return 0;
398 }
399 
400 static void __init setup_boot_config(void)
401 {
402 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
403 	const char *msg;
404 	int pos;
405 	u32 size, csum;
406 	char *data, *copy, *err;
407 	int ret;
408 
409 	/* Cut out the bootconfig data even if we have no bootconfig option */
410 	data = get_boot_config_from_initrd(&size, &csum);
411 
412 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
413 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
414 			 bootconfig_params);
415 
416 	if (IS_ERR(err) || !bootconfig_found)
417 		return;
418 
419 	/* parse_args() stops at '--' and returns an address */
420 	if (err)
421 		initargs_found = true;
422 
423 	if (!data) {
424 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
425 		return;
426 	}
427 
428 	if (size >= XBC_DATA_MAX) {
429 		pr_err("bootconfig size %d greater than max size %d\n",
430 			size, XBC_DATA_MAX);
431 		return;
432 	}
433 
434 	if (xbc_calc_checksum(data, size) != csum) {
435 		pr_err("bootconfig checksum failed\n");
436 		return;
437 	}
438 
439 	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
440 	if (!copy) {
441 		pr_err("Failed to allocate memory for bootconfig\n");
442 		return;
443 	}
444 
445 	memcpy(copy, data, size);
446 	copy[size] = '\0';
447 
448 	ret = xbc_init(copy, &msg, &pos);
449 	if (ret < 0) {
450 		if (pos < 0)
451 			pr_err("Failed to init bootconfig: %s.\n", msg);
452 		else
453 			pr_err("Failed to parse bootconfig: %s at %d.\n",
454 				msg, pos);
455 	} else {
456 		pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
457 		/* keys starting with "kernel." are passed via cmdline */
458 		extra_command_line = xbc_make_cmdline("kernel");
459 		/* Also, "init." keys are init arguments */
460 		extra_init_args = xbc_make_cmdline("init");
461 	}
462 	return;
463 }
464 
465 #else
466 
467 static void __init setup_boot_config(void)
468 {
469 	/* Remove bootconfig data from initrd */
470 	get_boot_config_from_initrd(NULL, NULL);
471 }
472 
473 static int __init warn_bootconfig(char *str)
474 {
475 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
476 	return 0;
477 }
478 early_param("bootconfig", warn_bootconfig);
479 
480 #endif
481 
482 /* Change NUL term back to "=", to make "param" the whole string. */
483 static void __init repair_env_string(char *param, char *val)
484 {
485 	if (val) {
486 		/* param=val or param="val"? */
487 		if (val == param+strlen(param)+1)
488 			val[-1] = '=';
489 		else if (val == param+strlen(param)+2) {
490 			val[-2] = '=';
491 			memmove(val-1, val, strlen(val)+1);
492 		} else
493 			BUG();
494 	}
495 }
496 
497 /* Anything after -- gets handed straight to init. */
498 static int __init set_init_arg(char *param, char *val,
499 			       const char *unused, void *arg)
500 {
501 	unsigned int i;
502 
503 	if (panic_later)
504 		return 0;
505 
506 	repair_env_string(param, val);
507 
508 	for (i = 0; argv_init[i]; i++) {
509 		if (i == MAX_INIT_ARGS) {
510 			panic_later = "init";
511 			panic_param = param;
512 			return 0;
513 		}
514 	}
515 	argv_init[i] = param;
516 	return 0;
517 }
518 
519 /*
520  * Unknown boot options get handed to init, unless they look like
521  * unused parameters (modprobe will find them in /proc/cmdline).
522  */
523 static int __init unknown_bootoption(char *param, char *val,
524 				     const char *unused, void *arg)
525 {
526 	size_t len = strlen(param);
527 
528 	repair_env_string(param, val);
529 
530 	/* Handle obsolete-style parameters */
531 	if (obsolete_checksetup(param))
532 		return 0;
533 
534 	/* Unused module parameter. */
535 	if (strnchr(param, len, '.'))
536 		return 0;
537 
538 	if (panic_later)
539 		return 0;
540 
541 	if (val) {
542 		/* Environment option */
543 		unsigned int i;
544 		for (i = 0; envp_init[i]; i++) {
545 			if (i == MAX_INIT_ENVS) {
546 				panic_later = "env";
547 				panic_param = param;
548 			}
549 			if (!strncmp(param, envp_init[i], len+1))
550 				break;
551 		}
552 		envp_init[i] = param;
553 	} else {
554 		/* Command line option */
555 		unsigned int i;
556 		for (i = 0; argv_init[i]; i++) {
557 			if (i == MAX_INIT_ARGS) {
558 				panic_later = "init";
559 				panic_param = param;
560 			}
561 		}
562 		argv_init[i] = param;
563 	}
564 	return 0;
565 }
566 
567 static int __init init_setup(char *str)
568 {
569 	unsigned int i;
570 
571 	execute_command = str;
572 	/*
573 	 * In case LILO is going to boot us with default command line,
574 	 * it prepends "auto" before the whole cmdline which makes
575 	 * the shell think it should execute a script with such name.
576 	 * So we ignore all arguments entered _before_ init=... [MJ]
577 	 */
578 	for (i = 1; i < MAX_INIT_ARGS; i++)
579 		argv_init[i] = NULL;
580 	return 1;
581 }
582 __setup("init=", init_setup);
583 
584 static int __init rdinit_setup(char *str)
585 {
586 	unsigned int i;
587 
588 	ramdisk_execute_command = str;
589 	/* See "auto" comment in init_setup */
590 	for (i = 1; i < MAX_INIT_ARGS; i++)
591 		argv_init[i] = NULL;
592 	return 1;
593 }
594 __setup("rdinit=", rdinit_setup);
595 
596 #ifndef CONFIG_SMP
597 static const unsigned int setup_max_cpus = NR_CPUS;
598 static inline void setup_nr_cpu_ids(void) { }
599 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
600 #endif
601 
602 /*
603  * We need to store the untouched command line for future reference.
604  * We also need to store the touched command line since the parameter
605  * parsing is performed in place, and we should allow a component to
606  * store reference of name/value for future reference.
607  */
608 static void __init setup_command_line(char *command_line)
609 {
610 	size_t len, xlen = 0, ilen = 0;
611 
612 	if (extra_command_line)
613 		xlen = strlen(extra_command_line);
614 	if (extra_init_args)
615 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
616 
617 	len = xlen + strlen(boot_command_line) + 1;
618 
619 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
620 	if (!saved_command_line)
621 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
622 
623 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
624 	if (!static_command_line)
625 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
626 
627 	if (xlen) {
628 		/*
629 		 * We have to put extra_command_line before boot command
630 		 * lines because there could be dashes (separator of init
631 		 * command line) in the command lines.
632 		 */
633 		strcpy(saved_command_line, extra_command_line);
634 		strcpy(static_command_line, extra_command_line);
635 	}
636 	strcpy(saved_command_line + xlen, boot_command_line);
637 	strcpy(static_command_line + xlen, command_line);
638 
639 	if (ilen) {
640 		/*
641 		 * Append supplemental init boot args to saved_command_line
642 		 * so that user can check what command line options passed
643 		 * to init.
644 		 */
645 		len = strlen(saved_command_line);
646 		if (initargs_found) {
647 			saved_command_line[len++] = ' ';
648 		} else {
649 			strcpy(saved_command_line + len, " -- ");
650 			len += 4;
651 		}
652 
653 		strcpy(saved_command_line + len, extra_init_args);
654 	}
655 }
656 
657 /*
658  * We need to finalize in a non-__init function or else race conditions
659  * between the root thread and the init thread may cause start_kernel to
660  * be reaped by free_initmem before the root thread has proceeded to
661  * cpu_idle.
662  *
663  * gcc-3.4 accidentally inlines this function, so use noinline.
664  */
665 
666 static __initdata DECLARE_COMPLETION(kthreadd_done);
667 
668 noinline void __ref rest_init(void)
669 {
670 	struct task_struct *tsk;
671 	int pid;
672 
673 	rcu_scheduler_starting();
674 	/*
675 	 * We need to spawn init first so that it obtains pid 1, however
676 	 * the init task will end up wanting to create kthreads, which, if
677 	 * we schedule it before we create kthreadd, will OOPS.
678 	 */
679 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
680 	/*
681 	 * Pin init on the boot CPU. Task migration is not properly working
682 	 * until sched_init_smp() has been run. It will set the allowed
683 	 * CPUs for init to the non isolated CPUs.
684 	 */
685 	rcu_read_lock();
686 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
687 	tsk->flags |= PF_NO_SETAFFINITY;
688 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
689 	rcu_read_unlock();
690 
691 	numa_default_policy();
692 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
693 	rcu_read_lock();
694 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
695 	rcu_read_unlock();
696 
697 	/*
698 	 * Enable might_sleep() and smp_processor_id() checks.
699 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
700 	 * kernel_thread() would trigger might_sleep() splats. With
701 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
702 	 * already, but it's stuck on the kthreadd_done completion.
703 	 */
704 	system_state = SYSTEM_SCHEDULING;
705 
706 	complete(&kthreadd_done);
707 
708 	/*
709 	 * The boot idle thread must execute schedule()
710 	 * at least once to get things moving:
711 	 */
712 	schedule_preempt_disabled();
713 	/* Call into cpu_idle with preempt disabled */
714 	cpu_startup_entry(CPUHP_ONLINE);
715 }
716 
717 /* Check for early params. */
718 static int __init do_early_param(char *param, char *val,
719 				 const char *unused, void *arg)
720 {
721 	const struct obs_kernel_param *p;
722 
723 	for (p = __setup_start; p < __setup_end; p++) {
724 		if ((p->early && parameq(param, p->str)) ||
725 		    (strcmp(param, "console") == 0 &&
726 		     strcmp(p->str, "earlycon") == 0)
727 		) {
728 			if (p->setup_func(val) != 0)
729 				pr_warn("Malformed early option '%s'\n", param);
730 		}
731 	}
732 	/* We accept everything at this stage. */
733 	return 0;
734 }
735 
736 void __init parse_early_options(char *cmdline)
737 {
738 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
739 		   do_early_param);
740 }
741 
742 /* Arch code calls this early on, or if not, just before other parsing. */
743 void __init parse_early_param(void)
744 {
745 	static int done __initdata;
746 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
747 
748 	if (done)
749 		return;
750 
751 	/* All fall through to do_early_param. */
752 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
753 	parse_early_options(tmp_cmdline);
754 	done = 1;
755 }
756 
757 void __init __weak arch_post_acpi_subsys_init(void) { }
758 
759 void __init __weak smp_setup_processor_id(void)
760 {
761 }
762 
763 # if THREAD_SIZE >= PAGE_SIZE
764 void __init __weak thread_stack_cache_init(void)
765 {
766 }
767 #endif
768 
769 void __init __weak mem_encrypt_init(void) { }
770 
771 void __init __weak poking_init(void) { }
772 
773 void __init __weak pgtable_cache_init(void) { }
774 
775 bool initcall_debug;
776 core_param(initcall_debug, initcall_debug, bool, 0644);
777 
778 #ifdef TRACEPOINTS_ENABLED
779 static void __init initcall_debug_enable(void);
780 #else
781 static inline void initcall_debug_enable(void)
782 {
783 }
784 #endif
785 
786 /* Report memory auto-initialization states for this boot. */
787 static void __init report_meminit(void)
788 {
789 	const char *stack;
790 
791 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
792 		stack = "all(pattern)";
793 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
794 		stack = "all(zero)";
795 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
796 		stack = "byref_all(zero)";
797 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
798 		stack = "byref(zero)";
799 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
800 		stack = "__user(zero)";
801 	else
802 		stack = "off";
803 
804 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
805 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
806 		want_init_on_free() ? "on" : "off");
807 	if (want_init_on_free())
808 		pr_info("mem auto-init: clearing system memory may take some time...\n");
809 }
810 
811 /*
812  * Set up kernel memory allocators
813  */
814 static void __init mm_init(void)
815 {
816 	/*
817 	 * page_ext requires contiguous pages,
818 	 * bigger than MAX_ORDER unless SPARSEMEM.
819 	 */
820 	page_ext_init_flatmem();
821 	init_mem_debugging_and_hardening();
822 	kfence_alloc_pool();
823 	report_meminit();
824 	stack_depot_init();
825 	mem_init();
826 	mem_init_print_info();
827 	/* page_owner must be initialized after buddy is ready */
828 	page_ext_init_flatmem_late();
829 	kmem_cache_init();
830 	kmemleak_init();
831 	pgtable_init();
832 	debug_objects_mem_init();
833 	vmalloc_init();
834 	/* Should be run before the first non-init thread is created */
835 	init_espfix_bsp();
836 	/* Should be run after espfix64 is set up. */
837 	pti_init();
838 }
839 
840 #ifdef CONFIG_HAVE_ARCH_RANDOMIZE_KSTACK_OFFSET
841 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
842 			   randomize_kstack_offset);
843 DEFINE_PER_CPU(u32, kstack_offset);
844 
845 static int __init early_randomize_kstack_offset(char *buf)
846 {
847 	int ret;
848 	bool bool_result;
849 
850 	ret = kstrtobool(buf, &bool_result);
851 	if (ret)
852 		return ret;
853 
854 	if (bool_result)
855 		static_branch_enable(&randomize_kstack_offset);
856 	else
857 		static_branch_disable(&randomize_kstack_offset);
858 	return 0;
859 }
860 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
861 #endif
862 
863 void __init __weak arch_call_rest_init(void)
864 {
865 	rest_init();
866 }
867 
868 static void __init print_unknown_bootoptions(void)
869 {
870 	char *unknown_options;
871 	char *end;
872 	const char *const *p;
873 	size_t len;
874 
875 	if (panic_later || (!argv_init[1] && !envp_init[2]))
876 		return;
877 
878 	/*
879 	 * Determine how many options we have to print out, plus a space
880 	 * before each
881 	 */
882 	len = 1; /* null terminator */
883 	for (p = &argv_init[1]; *p; p++) {
884 		len++;
885 		len += strlen(*p);
886 	}
887 	for (p = &envp_init[2]; *p; p++) {
888 		len++;
889 		len += strlen(*p);
890 	}
891 
892 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
893 	if (!unknown_options) {
894 		pr_err("%s: Failed to allocate %zu bytes\n",
895 			__func__, len);
896 		return;
897 	}
898 	end = unknown_options;
899 
900 	for (p = &argv_init[1]; *p; p++)
901 		end += sprintf(end, " %s", *p);
902 	for (p = &envp_init[2]; *p; p++)
903 		end += sprintf(end, " %s", *p);
904 
905 	pr_notice("Unknown command line parameters:%s\n", unknown_options);
906 	memblock_free(__pa(unknown_options), len);
907 }
908 
909 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
910 {
911 	char *command_line;
912 	char *after_dashes;
913 
914 	set_task_stack_end_magic(&init_task);
915 	smp_setup_processor_id();
916 	debug_objects_early_init();
917 	init_vmlinux_build_id();
918 
919 	cgroup_init_early();
920 
921 	local_irq_disable();
922 	early_boot_irqs_disabled = true;
923 
924 	/*
925 	 * Interrupts are still disabled. Do necessary setups, then
926 	 * enable them.
927 	 */
928 	boot_cpu_init();
929 	page_address_init();
930 	pr_notice("%s", linux_banner);
931 	early_security_init();
932 	setup_arch(&command_line);
933 	setup_boot_config();
934 	setup_command_line(command_line);
935 	setup_nr_cpu_ids();
936 	setup_per_cpu_areas();
937 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
938 	boot_cpu_hotplug_init();
939 
940 	build_all_zonelists(NULL);
941 	page_alloc_init();
942 
943 	pr_notice("Kernel command line: %s\n", saved_command_line);
944 	/* parameters may set static keys */
945 	jump_label_init();
946 	parse_early_param();
947 	after_dashes = parse_args("Booting kernel",
948 				  static_command_line, __start___param,
949 				  __stop___param - __start___param,
950 				  -1, -1, NULL, &unknown_bootoption);
951 	print_unknown_bootoptions();
952 	if (!IS_ERR_OR_NULL(after_dashes))
953 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
954 			   NULL, set_init_arg);
955 	if (extra_init_args)
956 		parse_args("Setting extra init args", extra_init_args,
957 			   NULL, 0, -1, -1, NULL, set_init_arg);
958 
959 	/*
960 	 * These use large bootmem allocations and must precede
961 	 * kmem_cache_init()
962 	 */
963 	setup_log_buf(0);
964 	vfs_caches_init_early();
965 	sort_main_extable();
966 	trap_init();
967 	mm_init();
968 
969 	ftrace_init();
970 
971 	/* trace_printk can be enabled here */
972 	early_trace_init();
973 
974 	/*
975 	 * Set up the scheduler prior starting any interrupts (such as the
976 	 * timer interrupt). Full topology setup happens at smp_init()
977 	 * time - but meanwhile we still have a functioning scheduler.
978 	 */
979 	sched_init();
980 
981 	if (WARN(!irqs_disabled(),
982 		 "Interrupts were enabled *very* early, fixing it\n"))
983 		local_irq_disable();
984 	radix_tree_init();
985 
986 	/*
987 	 * Set up housekeeping before setting up workqueues to allow the unbound
988 	 * workqueue to take non-housekeeping into account.
989 	 */
990 	housekeeping_init();
991 
992 	/*
993 	 * Allow workqueue creation and work item queueing/cancelling
994 	 * early.  Work item execution depends on kthreads and starts after
995 	 * workqueue_init().
996 	 */
997 	workqueue_init_early();
998 
999 	rcu_init();
1000 
1001 	/* Trace events are available after this */
1002 	trace_init();
1003 
1004 	if (initcall_debug)
1005 		initcall_debug_enable();
1006 
1007 	context_tracking_init();
1008 	/* init some links before init_ISA_irqs() */
1009 	early_irq_init();
1010 	init_IRQ();
1011 	tick_init();
1012 	rcu_init_nohz();
1013 	init_timers();
1014 	srcu_init();
1015 	hrtimers_init();
1016 	softirq_init();
1017 	timekeeping_init();
1018 	kfence_init();
1019 
1020 	/*
1021 	 * For best initial stack canary entropy, prepare it after:
1022 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1023 	 * - timekeeping_init() for ktime entropy used in rand_initialize()
1024 	 * - rand_initialize() to get any arch-specific entropy like RDRAND
1025 	 * - add_latent_entropy() to get any latent entropy
1026 	 * - adding command line entropy
1027 	 */
1028 	rand_initialize();
1029 	add_latent_entropy();
1030 	add_device_randomness(command_line, strlen(command_line));
1031 	boot_init_stack_canary();
1032 
1033 	time_init();
1034 	perf_event_init();
1035 	profile_init();
1036 	call_function_init();
1037 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1038 
1039 	early_boot_irqs_disabled = false;
1040 	local_irq_enable();
1041 
1042 	kmem_cache_init_late();
1043 
1044 	/*
1045 	 * HACK ALERT! This is early. We're enabling the console before
1046 	 * we've done PCI setups etc, and console_init() must be aware of
1047 	 * this. But we do want output early, in case something goes wrong.
1048 	 */
1049 	console_init();
1050 	if (panic_later)
1051 		panic("Too many boot %s vars at `%s'", panic_later,
1052 		      panic_param);
1053 
1054 	lockdep_init();
1055 
1056 	/*
1057 	 * Need to run this when irqs are enabled, because it wants
1058 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1059 	 * too:
1060 	 */
1061 	locking_selftest();
1062 
1063 	/*
1064 	 * This needs to be called before any devices perform DMA
1065 	 * operations that might use the SWIOTLB bounce buffers. It will
1066 	 * mark the bounce buffers as decrypted so that their usage will
1067 	 * not cause "plain-text" data to be decrypted when accessed.
1068 	 */
1069 	mem_encrypt_init();
1070 
1071 #ifdef CONFIG_BLK_DEV_INITRD
1072 	if (initrd_start && !initrd_below_start_ok &&
1073 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1074 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1075 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1076 		    min_low_pfn);
1077 		initrd_start = 0;
1078 	}
1079 #endif
1080 	setup_per_cpu_pageset();
1081 	numa_policy_init();
1082 	acpi_early_init();
1083 	if (late_time_init)
1084 		late_time_init();
1085 	sched_clock_init();
1086 	calibrate_delay();
1087 	pid_idr_init();
1088 	anon_vma_init();
1089 #ifdef CONFIG_X86
1090 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1091 		efi_enter_virtual_mode();
1092 #endif
1093 	thread_stack_cache_init();
1094 	cred_init();
1095 	fork_init();
1096 	proc_caches_init();
1097 	uts_ns_init();
1098 	key_init();
1099 	security_init();
1100 	dbg_late_init();
1101 	vfs_caches_init();
1102 	pagecache_init();
1103 	signals_init();
1104 	seq_file_init();
1105 	proc_root_init();
1106 	nsfs_init();
1107 	cpuset_init();
1108 	cgroup_init();
1109 	taskstats_init_early();
1110 	delayacct_init();
1111 
1112 	poking_init();
1113 	check_bugs();
1114 
1115 	acpi_subsystem_init();
1116 	arch_post_acpi_subsys_init();
1117 	kcsan_init();
1118 
1119 	/* Do the rest non-__init'ed, we're now alive */
1120 	arch_call_rest_init();
1121 
1122 	prevent_tail_call_optimization();
1123 }
1124 
1125 /* Call all constructor functions linked into the kernel. */
1126 static void __init do_ctors(void)
1127 {
1128 /*
1129  * For UML, the constructors have already been called by the
1130  * normal setup code as it's just a normal ELF binary, so we
1131  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1132  * even on UML for modules.
1133  */
1134 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1135 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1136 
1137 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1138 		(*fn)();
1139 #endif
1140 }
1141 
1142 #ifdef CONFIG_KALLSYMS
1143 struct blacklist_entry {
1144 	struct list_head next;
1145 	char *buf;
1146 };
1147 
1148 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1149 
1150 static int __init initcall_blacklist(char *str)
1151 {
1152 	char *str_entry;
1153 	struct blacklist_entry *entry;
1154 
1155 	/* str argument is a comma-separated list of functions */
1156 	do {
1157 		str_entry = strsep(&str, ",");
1158 		if (str_entry) {
1159 			pr_debug("blacklisting initcall %s\n", str_entry);
1160 			entry = memblock_alloc(sizeof(*entry),
1161 					       SMP_CACHE_BYTES);
1162 			if (!entry)
1163 				panic("%s: Failed to allocate %zu bytes\n",
1164 				      __func__, sizeof(*entry));
1165 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1166 						    SMP_CACHE_BYTES);
1167 			if (!entry->buf)
1168 				panic("%s: Failed to allocate %zu bytes\n",
1169 				      __func__, strlen(str_entry) + 1);
1170 			strcpy(entry->buf, str_entry);
1171 			list_add(&entry->next, &blacklisted_initcalls);
1172 		}
1173 	} while (str_entry);
1174 
1175 	return 0;
1176 }
1177 
1178 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1179 {
1180 	struct blacklist_entry *entry;
1181 	char fn_name[KSYM_SYMBOL_LEN];
1182 	unsigned long addr;
1183 
1184 	if (list_empty(&blacklisted_initcalls))
1185 		return false;
1186 
1187 	addr = (unsigned long) dereference_function_descriptor(fn);
1188 	sprint_symbol_no_offset(fn_name, addr);
1189 
1190 	/*
1191 	 * fn will be "function_name [module_name]" where [module_name] is not
1192 	 * displayed for built-in init functions.  Strip off the [module_name].
1193 	 */
1194 	strreplace(fn_name, ' ', '\0');
1195 
1196 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1197 		if (!strcmp(fn_name, entry->buf)) {
1198 			pr_debug("initcall %s blacklisted\n", fn_name);
1199 			return true;
1200 		}
1201 	}
1202 
1203 	return false;
1204 }
1205 #else
1206 static int __init initcall_blacklist(char *str)
1207 {
1208 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1209 	return 0;
1210 }
1211 
1212 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1213 {
1214 	return false;
1215 }
1216 #endif
1217 __setup("initcall_blacklist=", initcall_blacklist);
1218 
1219 static __init_or_module void
1220 trace_initcall_start_cb(void *data, initcall_t fn)
1221 {
1222 	ktime_t *calltime = (ktime_t *)data;
1223 
1224 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1225 	*calltime = ktime_get();
1226 }
1227 
1228 static __init_or_module void
1229 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1230 {
1231 	ktime_t *calltime = (ktime_t *)data;
1232 	ktime_t delta, rettime;
1233 	unsigned long long duration;
1234 
1235 	rettime = ktime_get();
1236 	delta = ktime_sub(rettime, *calltime);
1237 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1238 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1239 		 fn, ret, duration);
1240 }
1241 
1242 static ktime_t initcall_calltime;
1243 
1244 #ifdef TRACEPOINTS_ENABLED
1245 static void __init initcall_debug_enable(void)
1246 {
1247 	int ret;
1248 
1249 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1250 					    &initcall_calltime);
1251 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1252 					      &initcall_calltime);
1253 	WARN(ret, "Failed to register initcall tracepoints\n");
1254 }
1255 # define do_trace_initcall_start	trace_initcall_start
1256 # define do_trace_initcall_finish	trace_initcall_finish
1257 #else
1258 static inline void do_trace_initcall_start(initcall_t fn)
1259 {
1260 	if (!initcall_debug)
1261 		return;
1262 	trace_initcall_start_cb(&initcall_calltime, fn);
1263 }
1264 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1265 {
1266 	if (!initcall_debug)
1267 		return;
1268 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1269 }
1270 #endif /* !TRACEPOINTS_ENABLED */
1271 
1272 int __init_or_module do_one_initcall(initcall_t fn)
1273 {
1274 	int count = preempt_count();
1275 	char msgbuf[64];
1276 	int ret;
1277 
1278 	if (initcall_blacklisted(fn))
1279 		return -EPERM;
1280 
1281 	do_trace_initcall_start(fn);
1282 	ret = fn();
1283 	do_trace_initcall_finish(fn, ret);
1284 
1285 	msgbuf[0] = 0;
1286 
1287 	if (preempt_count() != count) {
1288 		sprintf(msgbuf, "preemption imbalance ");
1289 		preempt_count_set(count);
1290 	}
1291 	if (irqs_disabled()) {
1292 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1293 		local_irq_enable();
1294 	}
1295 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1296 
1297 	add_latent_entropy();
1298 	return ret;
1299 }
1300 
1301 
1302 extern initcall_entry_t __initcall_start[];
1303 extern initcall_entry_t __initcall0_start[];
1304 extern initcall_entry_t __initcall1_start[];
1305 extern initcall_entry_t __initcall2_start[];
1306 extern initcall_entry_t __initcall3_start[];
1307 extern initcall_entry_t __initcall4_start[];
1308 extern initcall_entry_t __initcall5_start[];
1309 extern initcall_entry_t __initcall6_start[];
1310 extern initcall_entry_t __initcall7_start[];
1311 extern initcall_entry_t __initcall_end[];
1312 
1313 static initcall_entry_t *initcall_levels[] __initdata = {
1314 	__initcall0_start,
1315 	__initcall1_start,
1316 	__initcall2_start,
1317 	__initcall3_start,
1318 	__initcall4_start,
1319 	__initcall5_start,
1320 	__initcall6_start,
1321 	__initcall7_start,
1322 	__initcall_end,
1323 };
1324 
1325 /* Keep these in sync with initcalls in include/linux/init.h */
1326 static const char *initcall_level_names[] __initdata = {
1327 	"pure",
1328 	"core",
1329 	"postcore",
1330 	"arch",
1331 	"subsys",
1332 	"fs",
1333 	"device",
1334 	"late",
1335 };
1336 
1337 static int __init ignore_unknown_bootoption(char *param, char *val,
1338 			       const char *unused, void *arg)
1339 {
1340 	return 0;
1341 }
1342 
1343 static void __init do_initcall_level(int level, char *command_line)
1344 {
1345 	initcall_entry_t *fn;
1346 
1347 	parse_args(initcall_level_names[level],
1348 		   command_line, __start___param,
1349 		   __stop___param - __start___param,
1350 		   level, level,
1351 		   NULL, ignore_unknown_bootoption);
1352 
1353 	trace_initcall_level(initcall_level_names[level]);
1354 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1355 		do_one_initcall(initcall_from_entry(fn));
1356 }
1357 
1358 static void __init do_initcalls(void)
1359 {
1360 	int level;
1361 	size_t len = strlen(saved_command_line) + 1;
1362 	char *command_line;
1363 
1364 	command_line = kzalloc(len, GFP_KERNEL);
1365 	if (!command_line)
1366 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1367 
1368 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1369 		/* Parser modifies command_line, restore it each time */
1370 		strcpy(command_line, saved_command_line);
1371 		do_initcall_level(level, command_line);
1372 	}
1373 
1374 	kfree(command_line);
1375 }
1376 
1377 /*
1378  * Ok, the machine is now initialized. None of the devices
1379  * have been touched yet, but the CPU subsystem is up and
1380  * running, and memory and process management works.
1381  *
1382  * Now we can finally start doing some real work..
1383  */
1384 static void __init do_basic_setup(void)
1385 {
1386 	cpuset_init_smp();
1387 	driver_init();
1388 	init_irq_proc();
1389 	do_ctors();
1390 	usermodehelper_enable();
1391 	do_initcalls();
1392 }
1393 
1394 static void __init do_pre_smp_initcalls(void)
1395 {
1396 	initcall_entry_t *fn;
1397 
1398 	trace_initcall_level("early");
1399 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1400 		do_one_initcall(initcall_from_entry(fn));
1401 }
1402 
1403 static int run_init_process(const char *init_filename)
1404 {
1405 	const char *const *p;
1406 
1407 	argv_init[0] = init_filename;
1408 	pr_info("Run %s as init process\n", init_filename);
1409 	pr_debug("  with arguments:\n");
1410 	for (p = argv_init; *p; p++)
1411 		pr_debug("    %s\n", *p);
1412 	pr_debug("  with environment:\n");
1413 	for (p = envp_init; *p; p++)
1414 		pr_debug("    %s\n", *p);
1415 	return kernel_execve(init_filename, argv_init, envp_init);
1416 }
1417 
1418 static int try_to_run_init_process(const char *init_filename)
1419 {
1420 	int ret;
1421 
1422 	ret = run_init_process(init_filename);
1423 
1424 	if (ret && ret != -ENOENT) {
1425 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1426 		       init_filename, ret);
1427 	}
1428 
1429 	return ret;
1430 }
1431 
1432 static noinline void __init kernel_init_freeable(void);
1433 
1434 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1435 bool rodata_enabled __ro_after_init = true;
1436 static int __init set_debug_rodata(char *str)
1437 {
1438 	return strtobool(str, &rodata_enabled);
1439 }
1440 __setup("rodata=", set_debug_rodata);
1441 #endif
1442 
1443 #ifdef CONFIG_STRICT_KERNEL_RWX
1444 static void mark_readonly(void)
1445 {
1446 	if (rodata_enabled) {
1447 		/*
1448 		 * load_module() results in W+X mappings, which are cleaned
1449 		 * up with call_rcu().  Let's make sure that queued work is
1450 		 * flushed so that we don't hit false positives looking for
1451 		 * insecure pages which are W+X.
1452 		 */
1453 		rcu_barrier();
1454 		mark_rodata_ro();
1455 		rodata_test();
1456 	} else
1457 		pr_info("Kernel memory protection disabled.\n");
1458 }
1459 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1460 static inline void mark_readonly(void)
1461 {
1462 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1463 }
1464 #else
1465 static inline void mark_readonly(void)
1466 {
1467 	pr_warn("This architecture does not have kernel memory protection.\n");
1468 }
1469 #endif
1470 
1471 void __weak free_initmem(void)
1472 {
1473 	free_initmem_default(POISON_FREE_INITMEM);
1474 }
1475 
1476 static int __ref kernel_init(void *unused)
1477 {
1478 	int ret;
1479 
1480 	/*
1481 	 * Wait until kthreadd is all set-up.
1482 	 */
1483 	wait_for_completion(&kthreadd_done);
1484 
1485 	kernel_init_freeable();
1486 	/* need to finish all async __init code before freeing the memory */
1487 	async_synchronize_full();
1488 	kprobe_free_init_mem();
1489 	ftrace_free_init_mem();
1490 	kgdb_free_init_mem();
1491 	free_initmem();
1492 	mark_readonly();
1493 
1494 	/*
1495 	 * Kernel mappings are now finalized - update the userspace page-table
1496 	 * to finalize PTI.
1497 	 */
1498 	pti_finalize();
1499 
1500 	system_state = SYSTEM_RUNNING;
1501 	numa_default_policy();
1502 
1503 	rcu_end_inkernel_boot();
1504 
1505 	do_sysctl_args();
1506 
1507 	if (ramdisk_execute_command) {
1508 		ret = run_init_process(ramdisk_execute_command);
1509 		if (!ret)
1510 			return 0;
1511 		pr_err("Failed to execute %s (error %d)\n",
1512 		       ramdisk_execute_command, ret);
1513 	}
1514 
1515 	/*
1516 	 * We try each of these until one succeeds.
1517 	 *
1518 	 * The Bourne shell can be used instead of init if we are
1519 	 * trying to recover a really broken machine.
1520 	 */
1521 	if (execute_command) {
1522 		ret = run_init_process(execute_command);
1523 		if (!ret)
1524 			return 0;
1525 		panic("Requested init %s failed (error %d).",
1526 		      execute_command, ret);
1527 	}
1528 
1529 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1530 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1531 		if (ret)
1532 			pr_err("Default init %s failed (error %d)\n",
1533 			       CONFIG_DEFAULT_INIT, ret);
1534 		else
1535 			return 0;
1536 	}
1537 
1538 	if (!try_to_run_init_process("/sbin/init") ||
1539 	    !try_to_run_init_process("/etc/init") ||
1540 	    !try_to_run_init_process("/bin/init") ||
1541 	    !try_to_run_init_process("/bin/sh"))
1542 		return 0;
1543 
1544 	panic("No working init found.  Try passing init= option to kernel. "
1545 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1546 }
1547 
1548 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1549 void __init console_on_rootfs(void)
1550 {
1551 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1552 
1553 	if (IS_ERR(file)) {
1554 		pr_err("Warning: unable to open an initial console.\n");
1555 		return;
1556 	}
1557 	init_dup(file);
1558 	init_dup(file);
1559 	init_dup(file);
1560 	fput(file);
1561 }
1562 
1563 static noinline void __init kernel_init_freeable(void)
1564 {
1565 	/* Now the scheduler is fully set up and can do blocking allocations */
1566 	gfp_allowed_mask = __GFP_BITS_MASK;
1567 
1568 	/*
1569 	 * init can allocate pages on any node
1570 	 */
1571 	set_mems_allowed(node_states[N_MEMORY]);
1572 
1573 	cad_pid = get_pid(task_pid(current));
1574 
1575 	smp_prepare_cpus(setup_max_cpus);
1576 
1577 	workqueue_init();
1578 
1579 	init_mm_internals();
1580 
1581 	rcu_init_tasks_generic();
1582 	do_pre_smp_initcalls();
1583 	lockup_detector_init();
1584 
1585 	smp_init();
1586 	sched_init_smp();
1587 
1588 	padata_init();
1589 	page_alloc_init_late();
1590 	/* Initialize page ext after all struct pages are initialized. */
1591 	page_ext_init();
1592 
1593 	do_basic_setup();
1594 
1595 	kunit_run_all_tests();
1596 
1597 	wait_for_initramfs();
1598 	console_on_rootfs();
1599 
1600 	/*
1601 	 * check if there is an early userspace init.  If yes, let it do all
1602 	 * the work
1603 	 */
1604 	if (init_eaccess(ramdisk_execute_command) != 0) {
1605 		ramdisk_execute_command = NULL;
1606 		prepare_namespace();
1607 	}
1608 
1609 	/*
1610 	 * Ok, we have completed the initial bootup, and
1611 	 * we're essentially up and running. Get rid of the
1612 	 * initmem segments and start the user-mode stuff..
1613 	 *
1614 	 * rootfs is available now, try loading the public keys
1615 	 * and default modules
1616 	 */
1617 
1618 	integrity_load_keys();
1619 }
1620