xref: /linux/init/main.c (revision 8eecf1c9929aef24e9e75280a39ed1ba3c64fb71)
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/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/list.h>
92 #include <linux/integrity.h>
93 #include <linux/proc_ns.h>
94 #include <linux/io.h>
95 #include <linux/cache.h>
96 #include <linux/rodata_test.h>
97 #include <linux/jump_label.h>
98 #include <linux/mem_encrypt.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #include <linux/stackdepot.h>
102 #include <net/net_namespace.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 size_t initargs_offs;
157 #else
158 # define bootconfig_found false
159 # define initargs_offs 0
160 #endif
161 
162 static char *execute_command;
163 static char *ramdisk_execute_command = "/init";
164 
165 /*
166  * Used to generate warnings if static_key manipulation functions are used
167  * before jump_label_init is called.
168  */
169 bool static_key_initialized __read_mostly;
170 EXPORT_SYMBOL_GPL(static_key_initialized);
171 
172 /*
173  * If set, this is an indication to the drivers that reset the underlying
174  * device before going ahead with the initialization otherwise driver might
175  * rely on the BIOS and skip the reset operation.
176  *
177  * This is useful if kernel is booting in an unreliable environment.
178  * For ex. kdump situation where previous kernel has crashed, BIOS has been
179  * skipped and devices will be in unknown state.
180  */
181 unsigned int reset_devices;
182 EXPORT_SYMBOL(reset_devices);
183 
184 static int __init set_reset_devices(char *str)
185 {
186 	reset_devices = 1;
187 	return 1;
188 }
189 
190 __setup("reset_devices", set_reset_devices);
191 
192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194 static const char *panic_later, *panic_param;
195 
196 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 		memblock_free(new_cmdline, len + 1);
386 		return NULL;
387 	}
388 
389 	return new_cmdline;
390 }
391 
392 static int __init bootconfig_params(char *param, char *val,
393 				    const char *unused, void *arg)
394 {
395 	if (strcmp(param, "bootconfig") == 0) {
396 		bootconfig_found = true;
397 	}
398 	return 0;
399 }
400 
401 static int __init warn_bootconfig(char *str)
402 {
403 	/* The 'bootconfig' has been handled by bootconfig_params(). */
404 	return 0;
405 }
406 
407 static void __init setup_boot_config(void)
408 {
409 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
410 	const char *msg;
411 	int pos;
412 	u32 size, csum;
413 	char *data, *err;
414 	int ret;
415 
416 	/* Cut out the bootconfig data even if we have no bootconfig option */
417 	data = get_boot_config_from_initrd(&size, &csum);
418 
419 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
420 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
421 			 bootconfig_params);
422 
423 	if (IS_ERR(err) || !bootconfig_found)
424 		return;
425 
426 	/* parse_args() stops at the next param of '--' and returns an address */
427 	if (err)
428 		initargs_offs = err - tmp_cmdline;
429 
430 	if (!data) {
431 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
432 		return;
433 	}
434 
435 	if (size >= XBC_DATA_MAX) {
436 		pr_err("bootconfig size %d greater than max size %d\n",
437 			size, XBC_DATA_MAX);
438 		return;
439 	}
440 
441 	if (xbc_calc_checksum(data, size) != csum) {
442 		pr_err("bootconfig checksum failed\n");
443 		return;
444 	}
445 
446 	ret = xbc_init(data, size, &msg, &pos);
447 	if (ret < 0) {
448 		if (pos < 0)
449 			pr_err("Failed to init bootconfig: %s.\n", msg);
450 		else
451 			pr_err("Failed to parse bootconfig: %s at %d.\n",
452 				msg, pos);
453 	} else {
454 		xbc_get_info(&ret, NULL);
455 		pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
456 		/* keys starting with "kernel." are passed via cmdline */
457 		extra_command_line = xbc_make_cmdline("kernel");
458 		/* Also, "init." keys are init arguments */
459 		extra_init_args = xbc_make_cmdline("init");
460 	}
461 	return;
462 }
463 
464 static void __init exit_boot_config(void)
465 {
466 	xbc_exit();
467 }
468 
469 #else	/* !CONFIG_BOOT_CONFIG */
470 
471 static void __init setup_boot_config(void)
472 {
473 	/* Remove bootconfig data from initrd */
474 	get_boot_config_from_initrd(NULL, NULL);
475 }
476 
477 static int __init warn_bootconfig(char *str)
478 {
479 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
480 	return 0;
481 }
482 
483 #define exit_boot_config()	do {} while (0)
484 
485 #endif	/* CONFIG_BOOT_CONFIG */
486 
487 early_param("bootconfig", warn_bootconfig);
488 
489 /* Change NUL term back to "=", to make "param" the whole string. */
490 static void __init repair_env_string(char *param, char *val)
491 {
492 	if (val) {
493 		/* param=val or param="val"? */
494 		if (val == param+strlen(param)+1)
495 			val[-1] = '=';
496 		else if (val == param+strlen(param)+2) {
497 			val[-2] = '=';
498 			memmove(val-1, val, strlen(val)+1);
499 		} else
500 			BUG();
501 	}
502 }
503 
504 /* Anything after -- gets handed straight to init. */
505 static int __init set_init_arg(char *param, char *val,
506 			       const char *unused, void *arg)
507 {
508 	unsigned int i;
509 
510 	if (panic_later)
511 		return 0;
512 
513 	repair_env_string(param, val);
514 
515 	for (i = 0; argv_init[i]; i++) {
516 		if (i == MAX_INIT_ARGS) {
517 			panic_later = "init";
518 			panic_param = param;
519 			return 0;
520 		}
521 	}
522 	argv_init[i] = param;
523 	return 0;
524 }
525 
526 /*
527  * Unknown boot options get handed to init, unless they look like
528  * unused parameters (modprobe will find them in /proc/cmdline).
529  */
530 static int __init unknown_bootoption(char *param, char *val,
531 				     const char *unused, void *arg)
532 {
533 	size_t len = strlen(param);
534 
535 	repair_env_string(param, val);
536 
537 	/* Handle obsolete-style parameters */
538 	if (obsolete_checksetup(param))
539 		return 0;
540 
541 	/* Unused module parameter. */
542 	if (strnchr(param, len, '.'))
543 		return 0;
544 
545 	if (panic_later)
546 		return 0;
547 
548 	if (val) {
549 		/* Environment option */
550 		unsigned int i;
551 		for (i = 0; envp_init[i]; i++) {
552 			if (i == MAX_INIT_ENVS) {
553 				panic_later = "env";
554 				panic_param = param;
555 			}
556 			if (!strncmp(param, envp_init[i], len+1))
557 				break;
558 		}
559 		envp_init[i] = param;
560 	} else {
561 		/* Command line option */
562 		unsigned int i;
563 		for (i = 0; argv_init[i]; i++) {
564 			if (i == MAX_INIT_ARGS) {
565 				panic_later = "init";
566 				panic_param = param;
567 			}
568 		}
569 		argv_init[i] = param;
570 	}
571 	return 0;
572 }
573 
574 static int __init init_setup(char *str)
575 {
576 	unsigned int i;
577 
578 	execute_command = str;
579 	/*
580 	 * In case LILO is going to boot us with default command line,
581 	 * it prepends "auto" before the whole cmdline which makes
582 	 * the shell think it should execute a script with such name.
583 	 * So we ignore all arguments entered _before_ init=... [MJ]
584 	 */
585 	for (i = 1; i < MAX_INIT_ARGS; i++)
586 		argv_init[i] = NULL;
587 	return 1;
588 }
589 __setup("init=", init_setup);
590 
591 static int __init rdinit_setup(char *str)
592 {
593 	unsigned int i;
594 
595 	ramdisk_execute_command = str;
596 	/* See "auto" comment in init_setup */
597 	for (i = 1; i < MAX_INIT_ARGS; i++)
598 		argv_init[i] = NULL;
599 	return 1;
600 }
601 __setup("rdinit=", rdinit_setup);
602 
603 #ifndef CONFIG_SMP
604 static const unsigned int setup_max_cpus = NR_CPUS;
605 static inline void setup_nr_cpu_ids(void) { }
606 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
607 #endif
608 
609 /*
610  * We need to store the untouched command line for future reference.
611  * We also need to store the touched command line since the parameter
612  * parsing is performed in place, and we should allow a component to
613  * store reference of name/value for future reference.
614  */
615 static void __init setup_command_line(char *command_line)
616 {
617 	size_t len, xlen = 0, ilen = 0;
618 
619 	if (extra_command_line)
620 		xlen = strlen(extra_command_line);
621 	if (extra_init_args)
622 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
623 
624 	len = xlen + strlen(boot_command_line) + 1;
625 
626 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
627 	if (!saved_command_line)
628 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
629 
630 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
631 	if (!static_command_line)
632 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
633 
634 	if (xlen) {
635 		/*
636 		 * We have to put extra_command_line before boot command
637 		 * lines because there could be dashes (separator of init
638 		 * command line) in the command lines.
639 		 */
640 		strcpy(saved_command_line, extra_command_line);
641 		strcpy(static_command_line, extra_command_line);
642 	}
643 	strcpy(saved_command_line + xlen, boot_command_line);
644 	strcpy(static_command_line + xlen, command_line);
645 
646 	if (ilen) {
647 		/*
648 		 * Append supplemental init boot args to saved_command_line
649 		 * so that user can check what command line options passed
650 		 * to init.
651 		 * The order should always be
652 		 * " -- "[bootconfig init-param][cmdline init-param]
653 		 */
654 		if (initargs_offs) {
655 			len = xlen + initargs_offs;
656 			strcpy(saved_command_line + len, extra_init_args);
657 			len += ilen - 4;	/* strlen(extra_init_args) */
658 			strcpy(saved_command_line + len,
659 				boot_command_line + initargs_offs - 1);
660 		} else {
661 			len = strlen(saved_command_line);
662 			strcpy(saved_command_line + len, " -- ");
663 			len += 4;
664 			strcpy(saved_command_line + len, extra_init_args);
665 		}
666 	}
667 }
668 
669 /*
670  * We need to finalize in a non-__init function or else race conditions
671  * between the root thread and the init thread may cause start_kernel to
672  * be reaped by free_initmem before the root thread has proceeded to
673  * cpu_idle.
674  *
675  * gcc-3.4 accidentally inlines this function, so use noinline.
676  */
677 
678 static __initdata DECLARE_COMPLETION(kthreadd_done);
679 
680 noinline void __ref rest_init(void)
681 {
682 	struct task_struct *tsk;
683 	int pid;
684 
685 	rcu_scheduler_starting();
686 	/*
687 	 * We need to spawn init first so that it obtains pid 1, however
688 	 * the init task will end up wanting to create kthreads, which, if
689 	 * we schedule it before we create kthreadd, will OOPS.
690 	 */
691 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
692 	/*
693 	 * Pin init on the boot CPU. Task migration is not properly working
694 	 * until sched_init_smp() has been run. It will set the allowed
695 	 * CPUs for init to the non isolated CPUs.
696 	 */
697 	rcu_read_lock();
698 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
699 	tsk->flags |= PF_NO_SETAFFINITY;
700 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
701 	rcu_read_unlock();
702 
703 	numa_default_policy();
704 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
705 	rcu_read_lock();
706 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
707 	rcu_read_unlock();
708 
709 	/*
710 	 * Enable might_sleep() and smp_processor_id() checks.
711 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
712 	 * kernel_thread() would trigger might_sleep() splats. With
713 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
714 	 * already, but it's stuck on the kthreadd_done completion.
715 	 */
716 	system_state = SYSTEM_SCHEDULING;
717 
718 	complete(&kthreadd_done);
719 
720 	/*
721 	 * The boot idle thread must execute schedule()
722 	 * at least once to get things moving:
723 	 */
724 	schedule_preempt_disabled();
725 	/* Call into cpu_idle with preempt disabled */
726 	cpu_startup_entry(CPUHP_ONLINE);
727 }
728 
729 /* Check for early params. */
730 static int __init do_early_param(char *param, char *val,
731 				 const char *unused, void *arg)
732 {
733 	const struct obs_kernel_param *p;
734 
735 	for (p = __setup_start; p < __setup_end; p++) {
736 		if ((p->early && parameq(param, p->str)) ||
737 		    (strcmp(param, "console") == 0 &&
738 		     strcmp(p->str, "earlycon") == 0)
739 		) {
740 			if (p->setup_func(val) != 0)
741 				pr_warn("Malformed early option '%s'\n", param);
742 		}
743 	}
744 	/* We accept everything at this stage. */
745 	return 0;
746 }
747 
748 void __init parse_early_options(char *cmdline)
749 {
750 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
751 		   do_early_param);
752 }
753 
754 /* Arch code calls this early on, or if not, just before other parsing. */
755 void __init parse_early_param(void)
756 {
757 	static int done __initdata;
758 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
759 
760 	if (done)
761 		return;
762 
763 	/* All fall through to do_early_param. */
764 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
765 	parse_early_options(tmp_cmdline);
766 	done = 1;
767 }
768 
769 void __init __weak arch_post_acpi_subsys_init(void) { }
770 
771 void __init __weak smp_setup_processor_id(void)
772 {
773 }
774 
775 # if THREAD_SIZE >= PAGE_SIZE
776 void __init __weak thread_stack_cache_init(void)
777 {
778 }
779 #endif
780 
781 void __init __weak mem_encrypt_init(void) { }
782 
783 void __init __weak poking_init(void) { }
784 
785 void __init __weak pgtable_cache_init(void) { }
786 
787 void __init __weak trap_init(void) { }
788 
789 bool initcall_debug;
790 core_param(initcall_debug, initcall_debug, bool, 0644);
791 
792 #ifdef TRACEPOINTS_ENABLED
793 static void __init initcall_debug_enable(void);
794 #else
795 static inline void initcall_debug_enable(void)
796 {
797 }
798 #endif
799 
800 /* Report memory auto-initialization states for this boot. */
801 static void __init report_meminit(void)
802 {
803 	const char *stack;
804 
805 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
806 		stack = "all(pattern)";
807 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
808 		stack = "all(zero)";
809 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
810 		stack = "byref_all(zero)";
811 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
812 		stack = "byref(zero)";
813 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
814 		stack = "__user(zero)";
815 	else
816 		stack = "off";
817 
818 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
819 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
820 		want_init_on_free() ? "on" : "off");
821 	if (want_init_on_free())
822 		pr_info("mem auto-init: clearing system memory may take some time...\n");
823 }
824 
825 /*
826  * Set up kernel memory allocators
827  */
828 static void __init mm_init(void)
829 {
830 	/*
831 	 * page_ext requires contiguous pages,
832 	 * bigger than MAX_ORDER unless SPARSEMEM.
833 	 */
834 	page_ext_init_flatmem();
835 	init_mem_debugging_and_hardening();
836 	kfence_alloc_pool();
837 	report_meminit();
838 	stack_depot_early_init();
839 	mem_init();
840 	mem_init_print_info();
841 	kmem_cache_init();
842 	/*
843 	 * page_owner must be initialized after buddy is ready, and also after
844 	 * slab is ready so that stack_depot_init() works properly
845 	 */
846 	page_ext_init_flatmem_late();
847 	kmemleak_init();
848 	pgtable_init();
849 	debug_objects_mem_init();
850 	vmalloc_init();
851 	/* Should be run before the first non-init thread is created */
852 	init_espfix_bsp();
853 	/* Should be run after espfix64 is set up. */
854 	pti_init();
855 }
856 
857 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
858 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
859 			   randomize_kstack_offset);
860 DEFINE_PER_CPU(u32, kstack_offset);
861 
862 static int __init early_randomize_kstack_offset(char *buf)
863 {
864 	int ret;
865 	bool bool_result;
866 
867 	ret = kstrtobool(buf, &bool_result);
868 	if (ret)
869 		return ret;
870 
871 	if (bool_result)
872 		static_branch_enable(&randomize_kstack_offset);
873 	else
874 		static_branch_disable(&randomize_kstack_offset);
875 	return 0;
876 }
877 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
878 #endif
879 
880 void __init __weak arch_call_rest_init(void)
881 {
882 	rest_init();
883 }
884 
885 static void __init print_unknown_bootoptions(void)
886 {
887 	char *unknown_options;
888 	char *end;
889 	const char *const *p;
890 	size_t len;
891 
892 	if (panic_later || (!argv_init[1] && !envp_init[2]))
893 		return;
894 
895 	/*
896 	 * Determine how many options we have to print out, plus a space
897 	 * before each
898 	 */
899 	len = 1; /* null terminator */
900 	for (p = &argv_init[1]; *p; p++) {
901 		len++;
902 		len += strlen(*p);
903 	}
904 	for (p = &envp_init[2]; *p; p++) {
905 		len++;
906 		len += strlen(*p);
907 	}
908 
909 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
910 	if (!unknown_options) {
911 		pr_err("%s: Failed to allocate %zu bytes\n",
912 			__func__, len);
913 		return;
914 	}
915 	end = unknown_options;
916 
917 	for (p = &argv_init[1]; *p; p++)
918 		end += sprintf(end, " %s", *p);
919 	for (p = &envp_init[2]; *p; p++)
920 		end += sprintf(end, " %s", *p);
921 
922 	/* Start at unknown_options[1] to skip the initial space */
923 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
924 		&unknown_options[1]);
925 	memblock_free(unknown_options, len);
926 }
927 
928 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
929 {
930 	char *command_line;
931 	char *after_dashes;
932 
933 	set_task_stack_end_magic(&init_task);
934 	smp_setup_processor_id();
935 	debug_objects_early_init();
936 	init_vmlinux_build_id();
937 
938 	cgroup_init_early();
939 
940 	local_irq_disable();
941 	early_boot_irqs_disabled = true;
942 
943 	/*
944 	 * Interrupts are still disabled. Do necessary setups, then
945 	 * enable them.
946 	 */
947 	boot_cpu_init();
948 	page_address_init();
949 	pr_notice("%s", linux_banner);
950 	early_security_init();
951 	setup_arch(&command_line);
952 	setup_boot_config();
953 	setup_command_line(command_line);
954 	setup_nr_cpu_ids();
955 	setup_per_cpu_areas();
956 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
957 	boot_cpu_hotplug_init();
958 
959 	build_all_zonelists(NULL);
960 	page_alloc_init();
961 
962 	pr_notice("Kernel command line: %s\n", saved_command_line);
963 	/* parameters may set static keys */
964 	jump_label_init();
965 	parse_early_param();
966 	after_dashes = parse_args("Booting kernel",
967 				  static_command_line, __start___param,
968 				  __stop___param - __start___param,
969 				  -1, -1, NULL, &unknown_bootoption);
970 	print_unknown_bootoptions();
971 	if (!IS_ERR_OR_NULL(after_dashes))
972 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
973 			   NULL, set_init_arg);
974 	if (extra_init_args)
975 		parse_args("Setting extra init args", extra_init_args,
976 			   NULL, 0, -1, -1, NULL, set_init_arg);
977 
978 	/*
979 	 * These use large bootmem allocations and must precede
980 	 * kmem_cache_init()
981 	 */
982 	setup_log_buf(0);
983 	vfs_caches_init_early();
984 	sort_main_extable();
985 	trap_init();
986 	mm_init();
987 
988 	ftrace_init();
989 
990 	/* trace_printk can be enabled here */
991 	early_trace_init();
992 
993 	/*
994 	 * Set up the scheduler prior starting any interrupts (such as the
995 	 * timer interrupt). Full topology setup happens at smp_init()
996 	 * time - but meanwhile we still have a functioning scheduler.
997 	 */
998 	sched_init();
999 
1000 	if (WARN(!irqs_disabled(),
1001 		 "Interrupts were enabled *very* early, fixing it\n"))
1002 		local_irq_disable();
1003 	radix_tree_init();
1004 
1005 	/*
1006 	 * Set up housekeeping before setting up workqueues to allow the unbound
1007 	 * workqueue to take non-housekeeping into account.
1008 	 */
1009 	housekeeping_init();
1010 
1011 	/*
1012 	 * Allow workqueue creation and work item queueing/cancelling
1013 	 * early.  Work item execution depends on kthreads and starts after
1014 	 * workqueue_init().
1015 	 */
1016 	workqueue_init_early();
1017 
1018 	rcu_init();
1019 
1020 	/* Trace events are available after this */
1021 	trace_init();
1022 
1023 	if (initcall_debug)
1024 		initcall_debug_enable();
1025 
1026 	context_tracking_init();
1027 	/* init some links before init_ISA_irqs() */
1028 	early_irq_init();
1029 	init_IRQ();
1030 	tick_init();
1031 	rcu_init_nohz();
1032 	init_timers();
1033 	srcu_init();
1034 	hrtimers_init();
1035 	softirq_init();
1036 	timekeeping_init();
1037 	kfence_init();
1038 	time_init();
1039 
1040 	/*
1041 	 * For best initial stack canary entropy, prepare it after:
1042 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
1043 	 * - timekeeping_init() for ktime entropy used in random_init()
1044 	 * - time_init() for making random_get_entropy() work on some platforms
1045 	 * - random_init() to initialize the RNG from from early entropy sources
1046 	 */
1047 	random_init(command_line);
1048 	boot_init_stack_canary();
1049 
1050 	perf_event_init();
1051 	profile_init();
1052 	call_function_init();
1053 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1054 
1055 	early_boot_irqs_disabled = false;
1056 	local_irq_enable();
1057 
1058 	kmem_cache_init_late();
1059 
1060 	/*
1061 	 * HACK ALERT! This is early. We're enabling the console before
1062 	 * we've done PCI setups etc, and console_init() must be aware of
1063 	 * this. But we do want output early, in case something goes wrong.
1064 	 */
1065 	console_init();
1066 	if (panic_later)
1067 		panic("Too many boot %s vars at `%s'", panic_later,
1068 		      panic_param);
1069 
1070 	lockdep_init();
1071 
1072 	/*
1073 	 * Need to run this when irqs are enabled, because it wants
1074 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1075 	 * too:
1076 	 */
1077 	locking_selftest();
1078 
1079 	/*
1080 	 * This needs to be called before any devices perform DMA
1081 	 * operations that might use the SWIOTLB bounce buffers. It will
1082 	 * mark the bounce buffers as decrypted so that their usage will
1083 	 * not cause "plain-text" data to be decrypted when accessed.
1084 	 */
1085 	mem_encrypt_init();
1086 
1087 #ifdef CONFIG_BLK_DEV_INITRD
1088 	if (initrd_start && !initrd_below_start_ok &&
1089 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1090 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1091 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1092 		    min_low_pfn);
1093 		initrd_start = 0;
1094 	}
1095 #endif
1096 	setup_per_cpu_pageset();
1097 	numa_policy_init();
1098 	acpi_early_init();
1099 	if (late_time_init)
1100 		late_time_init();
1101 	sched_clock_init();
1102 	calibrate_delay();
1103 	pid_idr_init();
1104 	anon_vma_init();
1105 #ifdef CONFIG_X86
1106 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1107 		efi_enter_virtual_mode();
1108 #endif
1109 	thread_stack_cache_init();
1110 	cred_init();
1111 	fork_init();
1112 	proc_caches_init();
1113 	uts_ns_init();
1114 	key_init();
1115 	security_init();
1116 	dbg_late_init();
1117 	net_ns_init();
1118 	vfs_caches_init();
1119 	pagecache_init();
1120 	signals_init();
1121 	seq_file_init();
1122 	proc_root_init();
1123 	nsfs_init();
1124 	cpuset_init();
1125 	cgroup_init();
1126 	taskstats_init_early();
1127 	delayacct_init();
1128 
1129 	poking_init();
1130 	check_bugs();
1131 
1132 	acpi_subsystem_init();
1133 	arch_post_acpi_subsys_init();
1134 	kcsan_init();
1135 
1136 	/* Do the rest non-__init'ed, we're now alive */
1137 	arch_call_rest_init();
1138 
1139 	prevent_tail_call_optimization();
1140 }
1141 
1142 /* Call all constructor functions linked into the kernel. */
1143 static void __init do_ctors(void)
1144 {
1145 /*
1146  * For UML, the constructors have already been called by the
1147  * normal setup code as it's just a normal ELF binary, so we
1148  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1149  * even on UML for modules.
1150  */
1151 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1152 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1153 
1154 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1155 		(*fn)();
1156 #endif
1157 }
1158 
1159 #ifdef CONFIG_KALLSYMS
1160 struct blacklist_entry {
1161 	struct list_head next;
1162 	char *buf;
1163 };
1164 
1165 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1166 
1167 static int __init initcall_blacklist(char *str)
1168 {
1169 	char *str_entry;
1170 	struct blacklist_entry *entry;
1171 
1172 	/* str argument is a comma-separated list of functions */
1173 	do {
1174 		str_entry = strsep(&str, ",");
1175 		if (str_entry) {
1176 			pr_debug("blacklisting initcall %s\n", str_entry);
1177 			entry = memblock_alloc(sizeof(*entry),
1178 					       SMP_CACHE_BYTES);
1179 			if (!entry)
1180 				panic("%s: Failed to allocate %zu bytes\n",
1181 				      __func__, sizeof(*entry));
1182 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1183 						    SMP_CACHE_BYTES);
1184 			if (!entry->buf)
1185 				panic("%s: Failed to allocate %zu bytes\n",
1186 				      __func__, strlen(str_entry) + 1);
1187 			strcpy(entry->buf, str_entry);
1188 			list_add(&entry->next, &blacklisted_initcalls);
1189 		}
1190 	} while (str_entry);
1191 
1192 	return 1;
1193 }
1194 
1195 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1196 {
1197 	struct blacklist_entry *entry;
1198 	char fn_name[KSYM_SYMBOL_LEN];
1199 	unsigned long addr;
1200 
1201 	if (list_empty(&blacklisted_initcalls))
1202 		return false;
1203 
1204 	addr = (unsigned long) dereference_function_descriptor(fn);
1205 	sprint_symbol_no_offset(fn_name, addr);
1206 
1207 	/*
1208 	 * fn will be "function_name [module_name]" where [module_name] is not
1209 	 * displayed for built-in init functions.  Strip off the [module_name].
1210 	 */
1211 	strreplace(fn_name, ' ', '\0');
1212 
1213 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1214 		if (!strcmp(fn_name, entry->buf)) {
1215 			pr_debug("initcall %s blacklisted\n", fn_name);
1216 			return true;
1217 		}
1218 	}
1219 
1220 	return false;
1221 }
1222 #else
1223 static int __init initcall_blacklist(char *str)
1224 {
1225 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1226 	return 0;
1227 }
1228 
1229 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1230 {
1231 	return false;
1232 }
1233 #endif
1234 __setup("initcall_blacklist=", initcall_blacklist);
1235 
1236 static __init_or_module void
1237 trace_initcall_start_cb(void *data, initcall_t fn)
1238 {
1239 	ktime_t *calltime = (ktime_t *)data;
1240 
1241 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1242 	*calltime = ktime_get();
1243 }
1244 
1245 static __init_or_module void
1246 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1247 {
1248 	ktime_t rettime, *calltime = (ktime_t *)data;
1249 
1250 	rettime = ktime_get();
1251 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1252 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1253 }
1254 
1255 static ktime_t initcall_calltime;
1256 
1257 #ifdef TRACEPOINTS_ENABLED
1258 static void __init initcall_debug_enable(void)
1259 {
1260 	int ret;
1261 
1262 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1263 					    &initcall_calltime);
1264 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1265 					      &initcall_calltime);
1266 	WARN(ret, "Failed to register initcall tracepoints\n");
1267 }
1268 # define do_trace_initcall_start	trace_initcall_start
1269 # define do_trace_initcall_finish	trace_initcall_finish
1270 #else
1271 static inline void do_trace_initcall_start(initcall_t fn)
1272 {
1273 	if (!initcall_debug)
1274 		return;
1275 	trace_initcall_start_cb(&initcall_calltime, fn);
1276 }
1277 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1278 {
1279 	if (!initcall_debug)
1280 		return;
1281 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1282 }
1283 #endif /* !TRACEPOINTS_ENABLED */
1284 
1285 int __init_or_module do_one_initcall(initcall_t fn)
1286 {
1287 	int count = preempt_count();
1288 	char msgbuf[64];
1289 	int ret;
1290 
1291 	if (initcall_blacklisted(fn))
1292 		return -EPERM;
1293 
1294 	do_trace_initcall_start(fn);
1295 	ret = fn();
1296 	do_trace_initcall_finish(fn, ret);
1297 
1298 	msgbuf[0] = 0;
1299 
1300 	if (preempt_count() != count) {
1301 		sprintf(msgbuf, "preemption imbalance ");
1302 		preempt_count_set(count);
1303 	}
1304 	if (irqs_disabled()) {
1305 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1306 		local_irq_enable();
1307 	}
1308 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1309 
1310 	add_latent_entropy();
1311 	return ret;
1312 }
1313 
1314 
1315 extern initcall_entry_t __initcall_start[];
1316 extern initcall_entry_t __initcall0_start[];
1317 extern initcall_entry_t __initcall1_start[];
1318 extern initcall_entry_t __initcall2_start[];
1319 extern initcall_entry_t __initcall3_start[];
1320 extern initcall_entry_t __initcall4_start[];
1321 extern initcall_entry_t __initcall5_start[];
1322 extern initcall_entry_t __initcall6_start[];
1323 extern initcall_entry_t __initcall7_start[];
1324 extern initcall_entry_t __initcall_end[];
1325 
1326 static initcall_entry_t *initcall_levels[] __initdata = {
1327 	__initcall0_start,
1328 	__initcall1_start,
1329 	__initcall2_start,
1330 	__initcall3_start,
1331 	__initcall4_start,
1332 	__initcall5_start,
1333 	__initcall6_start,
1334 	__initcall7_start,
1335 	__initcall_end,
1336 };
1337 
1338 /* Keep these in sync with initcalls in include/linux/init.h */
1339 static const char *initcall_level_names[] __initdata = {
1340 	"pure",
1341 	"core",
1342 	"postcore",
1343 	"arch",
1344 	"subsys",
1345 	"fs",
1346 	"device",
1347 	"late",
1348 };
1349 
1350 static int __init ignore_unknown_bootoption(char *param, char *val,
1351 			       const char *unused, void *arg)
1352 {
1353 	return 0;
1354 }
1355 
1356 static void __init do_initcall_level(int level, char *command_line)
1357 {
1358 	initcall_entry_t *fn;
1359 
1360 	parse_args(initcall_level_names[level],
1361 		   command_line, __start___param,
1362 		   __stop___param - __start___param,
1363 		   level, level,
1364 		   NULL, ignore_unknown_bootoption);
1365 
1366 	trace_initcall_level(initcall_level_names[level]);
1367 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1368 		do_one_initcall(initcall_from_entry(fn));
1369 }
1370 
1371 static void __init do_initcalls(void)
1372 {
1373 	int level;
1374 	size_t len = strlen(saved_command_line) + 1;
1375 	char *command_line;
1376 
1377 	command_line = kzalloc(len, GFP_KERNEL);
1378 	if (!command_line)
1379 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1380 
1381 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1382 		/* Parser modifies command_line, restore it each time */
1383 		strcpy(command_line, saved_command_line);
1384 		do_initcall_level(level, command_line);
1385 	}
1386 
1387 	kfree(command_line);
1388 }
1389 
1390 /*
1391  * Ok, the machine is now initialized. None of the devices
1392  * have been touched yet, but the CPU subsystem is up and
1393  * running, and memory and process management works.
1394  *
1395  * Now we can finally start doing some real work..
1396  */
1397 static void __init do_basic_setup(void)
1398 {
1399 	cpuset_init_smp();
1400 	driver_init();
1401 	init_irq_proc();
1402 	do_ctors();
1403 	do_initcalls();
1404 }
1405 
1406 static void __init do_pre_smp_initcalls(void)
1407 {
1408 	initcall_entry_t *fn;
1409 
1410 	trace_initcall_level("early");
1411 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1412 		do_one_initcall(initcall_from_entry(fn));
1413 }
1414 
1415 static int run_init_process(const char *init_filename)
1416 {
1417 	const char *const *p;
1418 
1419 	argv_init[0] = init_filename;
1420 	pr_info("Run %s as init process\n", init_filename);
1421 	pr_debug("  with arguments:\n");
1422 	for (p = argv_init; *p; p++)
1423 		pr_debug("    %s\n", *p);
1424 	pr_debug("  with environment:\n");
1425 	for (p = envp_init; *p; p++)
1426 		pr_debug("    %s\n", *p);
1427 	return kernel_execve(init_filename, argv_init, envp_init);
1428 }
1429 
1430 static int try_to_run_init_process(const char *init_filename)
1431 {
1432 	int ret;
1433 
1434 	ret = run_init_process(init_filename);
1435 
1436 	if (ret && ret != -ENOENT) {
1437 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1438 		       init_filename, ret);
1439 	}
1440 
1441 	return ret;
1442 }
1443 
1444 static noinline void __init kernel_init_freeable(void);
1445 
1446 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1447 bool rodata_enabled __ro_after_init = true;
1448 static int __init set_debug_rodata(char *str)
1449 {
1450 	if (strtobool(str, &rodata_enabled))
1451 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1452 	return 1;
1453 }
1454 __setup("rodata=", set_debug_rodata);
1455 #endif
1456 
1457 #ifdef CONFIG_STRICT_KERNEL_RWX
1458 static void mark_readonly(void)
1459 {
1460 	if (rodata_enabled) {
1461 		/*
1462 		 * load_module() results in W+X mappings, which are cleaned
1463 		 * up with call_rcu().  Let's make sure that queued work is
1464 		 * flushed so that we don't hit false positives looking for
1465 		 * insecure pages which are W+X.
1466 		 */
1467 		rcu_barrier();
1468 		mark_rodata_ro();
1469 		rodata_test();
1470 	} else
1471 		pr_info("Kernel memory protection disabled.\n");
1472 }
1473 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1474 static inline void mark_readonly(void)
1475 {
1476 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1477 }
1478 #else
1479 static inline void mark_readonly(void)
1480 {
1481 	pr_warn("This architecture does not have kernel memory protection.\n");
1482 }
1483 #endif
1484 
1485 void __weak free_initmem(void)
1486 {
1487 	free_initmem_default(POISON_FREE_INITMEM);
1488 }
1489 
1490 static int __ref kernel_init(void *unused)
1491 {
1492 	int ret;
1493 
1494 	/*
1495 	 * Wait until kthreadd is all set-up.
1496 	 */
1497 	wait_for_completion(&kthreadd_done);
1498 
1499 	kernel_init_freeable();
1500 	/* need to finish all async __init code before freeing the memory */
1501 	async_synchronize_full();
1502 
1503 	system_state = SYSTEM_FREEING_INITMEM;
1504 	kprobe_free_init_mem();
1505 	ftrace_free_init_mem();
1506 	kgdb_free_init_mem();
1507 	exit_boot_config();
1508 	free_initmem();
1509 	mark_readonly();
1510 
1511 	/*
1512 	 * Kernel mappings are now finalized - update the userspace page-table
1513 	 * to finalize PTI.
1514 	 */
1515 	pti_finalize();
1516 
1517 	system_state = SYSTEM_RUNNING;
1518 	numa_default_policy();
1519 
1520 	rcu_end_inkernel_boot();
1521 
1522 	do_sysctl_args();
1523 
1524 	if (ramdisk_execute_command) {
1525 		ret = run_init_process(ramdisk_execute_command);
1526 		if (!ret)
1527 			return 0;
1528 		pr_err("Failed to execute %s (error %d)\n",
1529 		       ramdisk_execute_command, ret);
1530 	}
1531 
1532 	/*
1533 	 * We try each of these until one succeeds.
1534 	 *
1535 	 * The Bourne shell can be used instead of init if we are
1536 	 * trying to recover a really broken machine.
1537 	 */
1538 	if (execute_command) {
1539 		ret = run_init_process(execute_command);
1540 		if (!ret)
1541 			return 0;
1542 		panic("Requested init %s failed (error %d).",
1543 		      execute_command, ret);
1544 	}
1545 
1546 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1547 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1548 		if (ret)
1549 			pr_err("Default init %s failed (error %d)\n",
1550 			       CONFIG_DEFAULT_INIT, ret);
1551 		else
1552 			return 0;
1553 	}
1554 
1555 	if (!try_to_run_init_process("/sbin/init") ||
1556 	    !try_to_run_init_process("/etc/init") ||
1557 	    !try_to_run_init_process("/bin/init") ||
1558 	    !try_to_run_init_process("/bin/sh"))
1559 		return 0;
1560 
1561 	panic("No working init found.  Try passing init= option to kernel. "
1562 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1563 }
1564 
1565 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1566 void __init console_on_rootfs(void)
1567 {
1568 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1569 
1570 	if (IS_ERR(file)) {
1571 		pr_err("Warning: unable to open an initial console.\n");
1572 		return;
1573 	}
1574 	init_dup(file);
1575 	init_dup(file);
1576 	init_dup(file);
1577 	fput(file);
1578 }
1579 
1580 static noinline void __init kernel_init_freeable(void)
1581 {
1582 	/* Now the scheduler is fully set up and can do blocking allocations */
1583 	gfp_allowed_mask = __GFP_BITS_MASK;
1584 
1585 	/*
1586 	 * init can allocate pages on any node
1587 	 */
1588 	set_mems_allowed(node_states[N_MEMORY]);
1589 
1590 	cad_pid = get_pid(task_pid(current));
1591 
1592 	smp_prepare_cpus(setup_max_cpus);
1593 
1594 	workqueue_init();
1595 
1596 	init_mm_internals();
1597 
1598 	rcu_init_tasks_generic();
1599 	do_pre_smp_initcalls();
1600 	lockup_detector_init();
1601 
1602 	smp_init();
1603 	sched_init_smp();
1604 
1605 	padata_init();
1606 	page_alloc_init_late();
1607 	/* Initialize page ext after all struct pages are initialized. */
1608 	page_ext_init();
1609 
1610 	do_basic_setup();
1611 
1612 	kunit_run_all_tests();
1613 
1614 	wait_for_initramfs();
1615 	console_on_rootfs();
1616 
1617 	/*
1618 	 * check if there is an early userspace init.  If yes, let it do all
1619 	 * the work
1620 	 */
1621 	if (init_eaccess(ramdisk_execute_command) != 0) {
1622 		ramdisk_execute_command = NULL;
1623 		prepare_namespace();
1624 	}
1625 
1626 	/*
1627 	 * Ok, we have completed the initial bootup, and
1628 	 * we're essentially up and running. Get rid of the
1629 	 * initmem segments and start the user-mode stuff..
1630 	 *
1631 	 * rootfs is available now, try loading the public keys
1632 	 * and default modules
1633 	 */
1634 
1635 	integrity_load_keys();
1636 }
1637