xref: /openbmc/linux/init/main.c (revision 37f0ab14)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/init/main.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
8  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
11  */
12 
13 #define DEBUG		/* Enable initcall_debug */
14 
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/cgroup.h>
54 #include <linux/efi.h>
55 #include <linux/tick.h>
56 #include <linux/sched/isolation.h>
57 #include <linux/interrupt.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/delayacct.h>
60 #include <linux/unistd.h>
61 #include <linux/utsname.h>
62 #include <linux/rmap.h>
63 #include <linux/mempolicy.h>
64 #include <linux/key.h>
65 #include <linux/page_ext.h>
66 #include <linux/debug_locks.h>
67 #include <linux/debugobjects.h>
68 #include <linux/lockdep.h>
69 #include <linux/kmemleak.h>
70 #include <linux/padata.h>
71 #include <linux/pid_namespace.h>
72 #include <linux/device/driver.h>
73 #include <linux/kthread.h>
74 #include <linux/sched.h>
75 #include <linux/sched/init.h>
76 #include <linux/signal.h>
77 #include <linux/idr.h>
78 #include <linux/kgdb.h>
79 #include <linux/ftrace.h>
80 #include <linux/async.h>
81 #include <linux/shmem_fs.h>
82 #include <linux/slab.h>
83 #include <linux/perf_event.h>
84 #include <linux/ptrace.h>
85 #include <linux/pti.h>
86 #include <linux/blkdev.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/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 #include <linux/randomize_kstack.h>
104 #include <net/net_namespace.h>
105 
106 #include <asm/io.h>
107 #include <asm/bugs.h>
108 #include <asm/setup.h>
109 #include <asm/sections.h>
110 #include <asm/cacheflush.h>
111 
112 #define CREATE_TRACE_POINTS
113 #include <trace/events/initcall.h>
114 
115 #include <kunit/test.h>
116 
117 static int kernel_init(void *);
118 
119 extern void init_IRQ(void);
120 extern void radix_tree_init(void);
121 extern void maple_tree_init(void);
122 
123 /*
124  * Debug helper: via this flag we know that we are in 'early bootup code'
125  * where only the boot processor is running with IRQ disabled.  This means
126  * two things - IRQ must not be enabled before the flag is cleared and some
127  * operations which are not allowed with IRQ disabled are allowed while the
128  * flag is set.
129  */
130 bool early_boot_irqs_disabled __read_mostly;
131 
132 enum system_states system_state __read_mostly;
133 EXPORT_SYMBOL(system_state);
134 
135 /*
136  * Boot command-line arguments
137  */
138 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
139 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
140 
141 extern void time_init(void);
142 /* Default late time init is NULL. archs can override this later. */
143 void (*__initdata late_time_init)(void);
144 
145 /* Untouched command line saved by arch-specific code. */
146 char __initdata boot_command_line[COMMAND_LINE_SIZE];
147 /* Untouched saved command line (eg. for /proc) */
148 char *saved_command_line;
149 /* Command line for parameter parsing */
150 static char *static_command_line;
151 /* Untouched extra command line */
152 static char *extra_command_line;
153 /* Extra init arguments */
154 static char *extra_init_args;
155 
156 #ifdef CONFIG_BOOT_CONFIG
157 /* Is bootconfig on command line? */
158 static bool bootconfig_found;
159 static size_t initargs_offs;
160 #else
161 # define bootconfig_found false
162 # define initargs_offs 0
163 #endif
164 
165 static char *execute_command;
166 static char *ramdisk_execute_command = "/init";
167 
168 /*
169  * Used to generate warnings if static_key manipulation functions are used
170  * before jump_label_init is called.
171  */
172 bool static_key_initialized __read_mostly;
173 EXPORT_SYMBOL_GPL(static_key_initialized);
174 
175 /*
176  * If set, this is an indication to the drivers that reset the underlying
177  * device before going ahead with the initialization otherwise driver might
178  * rely on the BIOS and skip the reset operation.
179  *
180  * This is useful if kernel is booting in an unreliable environment.
181  * For ex. kdump situation where previous kernel has crashed, BIOS has been
182  * skipped and devices will be in unknown state.
183  */
184 unsigned int reset_devices;
185 EXPORT_SYMBOL(reset_devices);
186 
187 static int __init set_reset_devices(char *str)
188 {
189 	reset_devices = 1;
190 	return 1;
191 }
192 
193 __setup("reset_devices", set_reset_devices);
194 
195 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
196 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
197 static const char *panic_later, *panic_param;
198 
199 extern const struct obs_kernel_param __setup_start[], __setup_end[];
200 
201 static bool __init obsolete_checksetup(char *line)
202 {
203 	const struct obs_kernel_param *p;
204 	bool had_early_param = false;
205 
206 	p = __setup_start;
207 	do {
208 		int n = strlen(p->str);
209 		if (parameqn(line, p->str, n)) {
210 			if (p->early) {
211 				/* Already done in parse_early_param?
212 				 * (Needs exact match on param part).
213 				 * Keep iterating, as we can have early
214 				 * params and __setups of same names 8( */
215 				if (line[n] == '\0' || line[n] == '=')
216 					had_early_param = true;
217 			} else if (!p->setup_func) {
218 				pr_warn("Parameter %s is obsolete, ignored\n",
219 					p->str);
220 				return true;
221 			} else if (p->setup_func(line + n))
222 				return true;
223 		}
224 		p++;
225 	} while (p < __setup_end);
226 
227 	return had_early_param;
228 }
229 
230 /*
231  * This should be approx 2 Bo*oMips to start (note initial shift), and will
232  * still work even if initially too large, it will just take slightly longer
233  */
234 unsigned long loops_per_jiffy = (1<<12);
235 EXPORT_SYMBOL(loops_per_jiffy);
236 
237 static int __init debug_kernel(char *str)
238 {
239 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
240 	return 0;
241 }
242 
243 static int __init quiet_kernel(char *str)
244 {
245 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
246 	return 0;
247 }
248 
249 early_param("debug", debug_kernel);
250 early_param("quiet", quiet_kernel);
251 
252 static int __init loglevel(char *str)
253 {
254 	int newlevel;
255 
256 	/*
257 	 * Only update loglevel value when a correct setting was passed,
258 	 * to prevent blind crashes (when loglevel being set to 0) that
259 	 * are quite hard to debug
260 	 */
261 	if (get_option(&str, &newlevel)) {
262 		console_loglevel = newlevel;
263 		return 0;
264 	}
265 
266 	return -EINVAL;
267 }
268 
269 early_param("loglevel", loglevel);
270 
271 #ifdef CONFIG_BLK_DEV_INITRD
272 static void * __init get_boot_config_from_initrd(size_t *_size)
273 {
274 	u32 size, csum;
275 	char *data;
276 	u32 *hdr;
277 	int i;
278 
279 	if (!initrd_end)
280 		return NULL;
281 
282 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
283 	/*
284 	 * Since Grub may align the size of initrd to 4, we must
285 	 * check the preceding 3 bytes as well.
286 	 */
287 	for (i = 0; i < 4; i++) {
288 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
289 			goto found;
290 		data--;
291 	}
292 	return NULL;
293 
294 found:
295 	hdr = (u32 *)(data - 8);
296 	size = le32_to_cpu(hdr[0]);
297 	csum = le32_to_cpu(hdr[1]);
298 
299 	data = ((void *)hdr) - size;
300 	if ((unsigned long)data < initrd_start) {
301 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
302 			size, initrd_end - initrd_start);
303 		return NULL;
304 	}
305 
306 	if (xbc_calc_checksum(data, size) != csum) {
307 		pr_err("bootconfig checksum failed\n");
308 		return NULL;
309 	}
310 
311 	/* Remove bootconfig from initramfs/initrd */
312 	initrd_end = (unsigned long)data;
313 	if (_size)
314 		*_size = size;
315 
316 	return data;
317 }
318 #else
319 static void * __init get_boot_config_from_initrd(size_t *_size)
320 {
321 	return NULL;
322 }
323 #endif
324 
325 #ifdef CONFIG_BOOT_CONFIG
326 
327 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
328 
329 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
330 
331 static int __init xbc_snprint_cmdline(char *buf, size_t size,
332 				      struct xbc_node *root)
333 {
334 	struct xbc_node *knode, *vnode;
335 	char *end = buf + size;
336 	const char *val;
337 	int ret;
338 
339 	xbc_node_for_each_key_value(root, knode, val) {
340 		ret = xbc_node_compose_key_after(root, knode,
341 					xbc_namebuf, XBC_KEYLEN_MAX);
342 		if (ret < 0)
343 			return ret;
344 
345 		vnode = xbc_node_get_child(knode);
346 		if (!vnode) {
347 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
348 			if (ret < 0)
349 				return ret;
350 			buf += ret;
351 			continue;
352 		}
353 		xbc_array_for_each_value(vnode, val) {
354 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
355 				       xbc_namebuf, val);
356 			if (ret < 0)
357 				return ret;
358 			buf += ret;
359 		}
360 	}
361 
362 	return buf - (end - size);
363 }
364 #undef rest
365 
366 /* Make an extra command line under given key word */
367 static char * __init xbc_make_cmdline(const char *key)
368 {
369 	struct xbc_node *root;
370 	char *new_cmdline;
371 	int ret, len = 0;
372 
373 	root = xbc_find_node(key);
374 	if (!root)
375 		return NULL;
376 
377 	/* Count required buffer size */
378 	len = xbc_snprint_cmdline(NULL, 0, root);
379 	if (len <= 0)
380 		return NULL;
381 
382 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
383 	if (!new_cmdline) {
384 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
385 		return NULL;
386 	}
387 
388 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
389 	if (ret < 0 || ret > len) {
390 		pr_err("Failed to print extra kernel cmdline.\n");
391 		memblock_free(new_cmdline, len + 1);
392 		return NULL;
393 	}
394 
395 	return new_cmdline;
396 }
397 
398 static int __init bootconfig_params(char *param, char *val,
399 				    const char *unused, void *arg)
400 {
401 	if (strcmp(param, "bootconfig") == 0) {
402 		bootconfig_found = true;
403 	}
404 	return 0;
405 }
406 
407 static int __init warn_bootconfig(char *str)
408 {
409 	/* The 'bootconfig' has been handled by bootconfig_params(). */
410 	return 0;
411 }
412 
413 static void __init setup_boot_config(void)
414 {
415 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
416 	const char *msg, *data;
417 	int pos, ret;
418 	size_t size;
419 	char *err;
420 
421 	/* Cut out the bootconfig data even if we have no bootconfig option */
422 	data = get_boot_config_from_initrd(&size);
423 	/* If there is no bootconfig in initrd, try embedded one. */
424 	if (!data)
425 		data = xbc_get_embedded_bootconfig(&size);
426 
427 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
428 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
429 			 bootconfig_params);
430 
431 	if (IS_ERR(err) || !bootconfig_found)
432 		return;
433 
434 	/* parse_args() stops at the next param of '--' and returns an address */
435 	if (err)
436 		initargs_offs = err - tmp_cmdline;
437 
438 	if (!data) {
439 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
440 		return;
441 	}
442 
443 	if (size >= XBC_DATA_MAX) {
444 		pr_err("bootconfig size %ld greater than max size %d\n",
445 			(long)size, XBC_DATA_MAX);
446 		return;
447 	}
448 
449 	ret = xbc_init(data, size, &msg, &pos);
450 	if (ret < 0) {
451 		if (pos < 0)
452 			pr_err("Failed to init bootconfig: %s.\n", msg);
453 		else
454 			pr_err("Failed to parse bootconfig: %s at %d.\n",
455 				msg, pos);
456 	} else {
457 		xbc_get_info(&ret, NULL);
458 		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
459 		/* keys starting with "kernel." are passed via cmdline */
460 		extra_command_line = xbc_make_cmdline("kernel");
461 		/* Also, "init." keys are init arguments */
462 		extra_init_args = xbc_make_cmdline("init");
463 	}
464 	return;
465 }
466 
467 static void __init exit_boot_config(void)
468 {
469 	xbc_exit();
470 }
471 
472 #else	/* !CONFIG_BOOT_CONFIG */
473 
474 static void __init setup_boot_config(void)
475 {
476 	/* Remove bootconfig data from initrd */
477 	get_boot_config_from_initrd(NULL);
478 }
479 
480 static int __init warn_bootconfig(char *str)
481 {
482 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
483 	return 0;
484 }
485 
486 #define exit_boot_config()	do {} while (0)
487 
488 #endif	/* CONFIG_BOOT_CONFIG */
489 
490 early_param("bootconfig", warn_bootconfig);
491 
492 /* Change NUL term back to "=", to make "param" the whole string. */
493 static void __init repair_env_string(char *param, char *val)
494 {
495 	if (val) {
496 		/* param=val or param="val"? */
497 		if (val == param+strlen(param)+1)
498 			val[-1] = '=';
499 		else if (val == param+strlen(param)+2) {
500 			val[-2] = '=';
501 			memmove(val-1, val, strlen(val)+1);
502 		} else
503 			BUG();
504 	}
505 }
506 
507 /* Anything after -- gets handed straight to init. */
508 static int __init set_init_arg(char *param, char *val,
509 			       const char *unused, void *arg)
510 {
511 	unsigned int i;
512 
513 	if (panic_later)
514 		return 0;
515 
516 	repair_env_string(param, val);
517 
518 	for (i = 0; argv_init[i]; i++) {
519 		if (i == MAX_INIT_ARGS) {
520 			panic_later = "init";
521 			panic_param = param;
522 			return 0;
523 		}
524 	}
525 	argv_init[i] = param;
526 	return 0;
527 }
528 
529 /*
530  * Unknown boot options get handed to init, unless they look like
531  * unused parameters (modprobe will find them in /proc/cmdline).
532  */
533 static int __init unknown_bootoption(char *param, char *val,
534 				     const char *unused, void *arg)
535 {
536 	size_t len = strlen(param);
537 
538 	repair_env_string(param, val);
539 
540 	/* Handle obsolete-style parameters */
541 	if (obsolete_checksetup(param))
542 		return 0;
543 
544 	/* Unused module parameter. */
545 	if (strnchr(param, len, '.'))
546 		return 0;
547 
548 	if (panic_later)
549 		return 0;
550 
551 	if (val) {
552 		/* Environment option */
553 		unsigned int i;
554 		for (i = 0; envp_init[i]; i++) {
555 			if (i == MAX_INIT_ENVS) {
556 				panic_later = "env";
557 				panic_param = param;
558 			}
559 			if (!strncmp(param, envp_init[i], len+1))
560 				break;
561 		}
562 		envp_init[i] = param;
563 	} else {
564 		/* Command line option */
565 		unsigned int i;
566 		for (i = 0; argv_init[i]; i++) {
567 			if (i == MAX_INIT_ARGS) {
568 				panic_later = "init";
569 				panic_param = param;
570 			}
571 		}
572 		argv_init[i] = param;
573 	}
574 	return 0;
575 }
576 
577 static int __init init_setup(char *str)
578 {
579 	unsigned int i;
580 
581 	execute_command = str;
582 	/*
583 	 * In case LILO is going to boot us with default command line,
584 	 * it prepends "auto" before the whole cmdline which makes
585 	 * the shell think it should execute a script with such name.
586 	 * So we ignore all arguments entered _before_ init=... [MJ]
587 	 */
588 	for (i = 1; i < MAX_INIT_ARGS; i++)
589 		argv_init[i] = NULL;
590 	return 1;
591 }
592 __setup("init=", init_setup);
593 
594 static int __init rdinit_setup(char *str)
595 {
596 	unsigned int i;
597 
598 	ramdisk_execute_command = str;
599 	/* See "auto" comment in init_setup */
600 	for (i = 1; i < MAX_INIT_ARGS; i++)
601 		argv_init[i] = NULL;
602 	return 1;
603 }
604 __setup("rdinit=", rdinit_setup);
605 
606 #ifndef CONFIG_SMP
607 static const unsigned int setup_max_cpus = NR_CPUS;
608 static inline void setup_nr_cpu_ids(void) { }
609 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
610 #endif
611 
612 /*
613  * We need to store the untouched command line for future reference.
614  * We also need to store the touched command line since the parameter
615  * parsing is performed in place, and we should allow a component to
616  * store reference of name/value for future reference.
617  */
618 static void __init setup_command_line(char *command_line)
619 {
620 	size_t len, xlen = 0, ilen = 0;
621 
622 	if (extra_command_line)
623 		xlen = strlen(extra_command_line);
624 	if (extra_init_args)
625 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
626 
627 	len = xlen + strlen(boot_command_line) + 1;
628 
629 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
630 	if (!saved_command_line)
631 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
632 
633 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
634 	if (!static_command_line)
635 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
636 
637 	if (xlen) {
638 		/*
639 		 * We have to put extra_command_line before boot command
640 		 * lines because there could be dashes (separator of init
641 		 * command line) in the command lines.
642 		 */
643 		strcpy(saved_command_line, extra_command_line);
644 		strcpy(static_command_line, extra_command_line);
645 	}
646 	strcpy(saved_command_line + xlen, boot_command_line);
647 	strcpy(static_command_line + xlen, command_line);
648 
649 	if (ilen) {
650 		/*
651 		 * Append supplemental init boot args to saved_command_line
652 		 * so that user can check what command line options passed
653 		 * to init.
654 		 * The order should always be
655 		 * " -- "[bootconfig init-param][cmdline init-param]
656 		 */
657 		if (initargs_offs) {
658 			len = xlen + initargs_offs;
659 			strcpy(saved_command_line + len, extra_init_args);
660 			len += ilen - 4;	/* strlen(extra_init_args) */
661 			strcpy(saved_command_line + len,
662 				boot_command_line + initargs_offs - 1);
663 		} else {
664 			len = strlen(saved_command_line);
665 			strcpy(saved_command_line + len, " -- ");
666 			len += 4;
667 			strcpy(saved_command_line + len, extra_init_args);
668 		}
669 	}
670 }
671 
672 /*
673  * We need to finalize in a non-__init function or else race conditions
674  * between the root thread and the init thread may cause start_kernel to
675  * be reaped by free_initmem before the root thread has proceeded to
676  * cpu_idle.
677  *
678  * gcc-3.4 accidentally inlines this function, so use noinline.
679  */
680 
681 static __initdata DECLARE_COMPLETION(kthreadd_done);
682 
683 noinline void __ref rest_init(void)
684 {
685 	struct task_struct *tsk;
686 	int pid;
687 
688 	rcu_scheduler_starting();
689 	/*
690 	 * We need to spawn init first so that it obtains pid 1, however
691 	 * the init task will end up wanting to create kthreads, which, if
692 	 * we schedule it before we create kthreadd, will OOPS.
693 	 */
694 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
695 	/*
696 	 * Pin init on the boot CPU. Task migration is not properly working
697 	 * until sched_init_smp() has been run. It will set the allowed
698 	 * CPUs for init to the non isolated CPUs.
699 	 */
700 	rcu_read_lock();
701 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
702 	tsk->flags |= PF_NO_SETAFFINITY;
703 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
704 	rcu_read_unlock();
705 
706 	numa_default_policy();
707 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
708 	rcu_read_lock();
709 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
710 	rcu_read_unlock();
711 
712 	/*
713 	 * Enable might_sleep() and smp_processor_id() checks.
714 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
715 	 * kernel_thread() would trigger might_sleep() splats. With
716 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
717 	 * already, but it's stuck on the kthreadd_done completion.
718 	 */
719 	system_state = SYSTEM_SCHEDULING;
720 
721 	complete(&kthreadd_done);
722 
723 	/*
724 	 * The boot idle thread must execute schedule()
725 	 * at least once to get things moving:
726 	 */
727 	schedule_preempt_disabled();
728 	/* Call into cpu_idle with preempt disabled */
729 	cpu_startup_entry(CPUHP_ONLINE);
730 }
731 
732 /* Check for early params. */
733 static int __init do_early_param(char *param, char *val,
734 				 const char *unused, void *arg)
735 {
736 	const struct obs_kernel_param *p;
737 
738 	for (p = __setup_start; p < __setup_end; p++) {
739 		if ((p->early && parameq(param, p->str)) ||
740 		    (strcmp(param, "console") == 0 &&
741 		     strcmp(p->str, "earlycon") == 0)
742 		) {
743 			if (p->setup_func(val) != 0)
744 				pr_warn("Malformed early option '%s'\n", param);
745 		}
746 	}
747 	/* We accept everything at this stage. */
748 	return 0;
749 }
750 
751 void __init parse_early_options(char *cmdline)
752 {
753 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
754 		   do_early_param);
755 }
756 
757 /* Arch code calls this early on, or if not, just before other parsing. */
758 void __init parse_early_param(void)
759 {
760 	static int done __initdata;
761 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
762 
763 	if (done)
764 		return;
765 
766 	/* All fall through to do_early_param. */
767 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
768 	parse_early_options(tmp_cmdline);
769 	done = 1;
770 }
771 
772 void __init __weak arch_post_acpi_subsys_init(void) { }
773 
774 void __init __weak smp_setup_processor_id(void)
775 {
776 }
777 
778 # if THREAD_SIZE >= PAGE_SIZE
779 void __init __weak thread_stack_cache_init(void)
780 {
781 }
782 #endif
783 
784 void __init __weak mem_encrypt_init(void) { }
785 
786 void __init __weak poking_init(void) { }
787 
788 void __init __weak pgtable_cache_init(void) { }
789 
790 void __init __weak trap_init(void) { }
791 
792 bool initcall_debug;
793 core_param(initcall_debug, initcall_debug, bool, 0644);
794 
795 #ifdef TRACEPOINTS_ENABLED
796 static void __init initcall_debug_enable(void);
797 #else
798 static inline void initcall_debug_enable(void)
799 {
800 }
801 #endif
802 
803 /* Report memory auto-initialization states for this boot. */
804 static void __init report_meminit(void)
805 {
806 	const char *stack;
807 
808 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
809 		stack = "all(pattern)";
810 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
811 		stack = "all(zero)";
812 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
813 		stack = "byref_all(zero)";
814 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
815 		stack = "byref(zero)";
816 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
817 		stack = "__user(zero)";
818 	else
819 		stack = "off";
820 
821 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
822 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
823 		want_init_on_free() ? "on" : "off");
824 	if (want_init_on_free())
825 		pr_info("mem auto-init: clearing system memory may take some time...\n");
826 }
827 
828 /*
829  * Set up kernel memory allocators
830  */
831 static void __init mm_init(void)
832 {
833 	/*
834 	 * page_ext requires contiguous pages,
835 	 * bigger than MAX_ORDER unless SPARSEMEM.
836 	 */
837 	page_ext_init_flatmem();
838 	init_mem_debugging_and_hardening();
839 	kfence_alloc_pool();
840 	report_meminit();
841 	kmsan_init_shadow();
842 	stack_depot_early_init();
843 	mem_init();
844 	mem_init_print_info();
845 	kmem_cache_init();
846 	/*
847 	 * page_owner must be initialized after buddy is ready, and also after
848 	 * slab is ready so that stack_depot_init() works properly
849 	 */
850 	page_ext_init_flatmem_late();
851 	kmemleak_init();
852 	pgtable_init();
853 	debug_objects_mem_init();
854 	vmalloc_init();
855 	/* Should be run after vmap initialization */
856 	if (early_page_ext_enabled())
857 		page_ext_init();
858 	/* Should be run before the first non-init thread is created */
859 	init_espfix_bsp();
860 	/* Should be run after espfix64 is set up. */
861 	pti_init();
862 	kmsan_init_runtime();
863 }
864 
865 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
866 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
867 			   randomize_kstack_offset);
868 DEFINE_PER_CPU(u32, kstack_offset);
869 
870 static int __init early_randomize_kstack_offset(char *buf)
871 {
872 	int ret;
873 	bool bool_result;
874 
875 	ret = kstrtobool(buf, &bool_result);
876 	if (ret)
877 		return ret;
878 
879 	if (bool_result)
880 		static_branch_enable(&randomize_kstack_offset);
881 	else
882 		static_branch_disable(&randomize_kstack_offset);
883 	return 0;
884 }
885 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
886 #endif
887 
888 void __init __weak arch_call_rest_init(void)
889 {
890 	rest_init();
891 }
892 
893 static void __init print_unknown_bootoptions(void)
894 {
895 	char *unknown_options;
896 	char *end;
897 	const char *const *p;
898 	size_t len;
899 
900 	if (panic_later || (!argv_init[1] && !envp_init[2]))
901 		return;
902 
903 	/*
904 	 * Determine how many options we have to print out, plus a space
905 	 * before each
906 	 */
907 	len = 1; /* null terminator */
908 	for (p = &argv_init[1]; *p; p++) {
909 		len++;
910 		len += strlen(*p);
911 	}
912 	for (p = &envp_init[2]; *p; p++) {
913 		len++;
914 		len += strlen(*p);
915 	}
916 
917 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
918 	if (!unknown_options) {
919 		pr_err("%s: Failed to allocate %zu bytes\n",
920 			__func__, len);
921 		return;
922 	}
923 	end = unknown_options;
924 
925 	for (p = &argv_init[1]; *p; p++)
926 		end += sprintf(end, " %s", *p);
927 	for (p = &envp_init[2]; *p; p++)
928 		end += sprintf(end, " %s", *p);
929 
930 	/* Start at unknown_options[1] to skip the initial space */
931 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
932 		&unknown_options[1]);
933 	memblock_free(unknown_options, len);
934 }
935 
936 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
937 {
938 	char *command_line;
939 	char *after_dashes;
940 
941 	set_task_stack_end_magic(&init_task);
942 	smp_setup_processor_id();
943 	debug_objects_early_init();
944 	init_vmlinux_build_id();
945 
946 	cgroup_init_early();
947 
948 	local_irq_disable();
949 	early_boot_irqs_disabled = true;
950 
951 	/*
952 	 * Interrupts are still disabled. Do necessary setups, then
953 	 * enable them.
954 	 */
955 	boot_cpu_init();
956 	page_address_init();
957 	pr_notice("%s", linux_banner);
958 	early_security_init();
959 	setup_arch(&command_line);
960 	setup_boot_config();
961 	setup_command_line(command_line);
962 	setup_nr_cpu_ids();
963 	setup_per_cpu_areas();
964 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
965 	boot_cpu_hotplug_init();
966 
967 	build_all_zonelists(NULL);
968 	page_alloc_init();
969 
970 	pr_notice("Kernel command line: %s\n", saved_command_line);
971 	/* parameters may set static keys */
972 	jump_label_init();
973 	parse_early_param();
974 	after_dashes = parse_args("Booting kernel",
975 				  static_command_line, __start___param,
976 				  __stop___param - __start___param,
977 				  -1, -1, NULL, &unknown_bootoption);
978 	print_unknown_bootoptions();
979 	if (!IS_ERR_OR_NULL(after_dashes))
980 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
981 			   NULL, set_init_arg);
982 	if (extra_init_args)
983 		parse_args("Setting extra init args", extra_init_args,
984 			   NULL, 0, -1, -1, NULL, set_init_arg);
985 
986 	/* Architectural and non-timekeeping rng init, before allocator init */
987 	random_init_early(command_line);
988 
989 	/*
990 	 * These use large bootmem allocations and must precede
991 	 * kmem_cache_init()
992 	 */
993 	setup_log_buf(0);
994 	vfs_caches_init_early();
995 	sort_main_extable();
996 	trap_init();
997 	mm_init();
998 
999 	ftrace_init();
1000 
1001 	/* trace_printk can be enabled here */
1002 	early_trace_init();
1003 
1004 	/*
1005 	 * Set up the scheduler prior starting any interrupts (such as the
1006 	 * timer interrupt). Full topology setup happens at smp_init()
1007 	 * time - but meanwhile we still have a functioning scheduler.
1008 	 */
1009 	sched_init();
1010 
1011 	if (WARN(!irqs_disabled(),
1012 		 "Interrupts were enabled *very* early, fixing it\n"))
1013 		local_irq_disable();
1014 	radix_tree_init();
1015 	maple_tree_init();
1016 
1017 	/*
1018 	 * Set up housekeeping before setting up workqueues to allow the unbound
1019 	 * workqueue to take non-housekeeping into account.
1020 	 */
1021 	housekeeping_init();
1022 
1023 	/*
1024 	 * Allow workqueue creation and work item queueing/cancelling
1025 	 * early.  Work item execution depends on kthreads and starts after
1026 	 * workqueue_init().
1027 	 */
1028 	workqueue_init_early();
1029 
1030 	rcu_init();
1031 
1032 	/* Trace events are available after this */
1033 	trace_init();
1034 
1035 	if (initcall_debug)
1036 		initcall_debug_enable();
1037 
1038 	context_tracking_init();
1039 	/* init some links before init_ISA_irqs() */
1040 	early_irq_init();
1041 	init_IRQ();
1042 	tick_init();
1043 	rcu_init_nohz();
1044 	init_timers();
1045 	srcu_init();
1046 	hrtimers_init();
1047 	softirq_init();
1048 	timekeeping_init();
1049 	time_init();
1050 
1051 	/* This must be after timekeeping is initialized */
1052 	random_init();
1053 
1054 	/* These make use of the fully initialized rng */
1055 	kfence_init();
1056 	boot_init_stack_canary();
1057 
1058 	perf_event_init();
1059 	profile_init();
1060 	call_function_init();
1061 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1062 
1063 	early_boot_irqs_disabled = false;
1064 	local_irq_enable();
1065 
1066 	kmem_cache_init_late();
1067 
1068 	/*
1069 	 * HACK ALERT! This is early. We're enabling the console before
1070 	 * we've done PCI setups etc, and console_init() must be aware of
1071 	 * this. But we do want output early, in case something goes wrong.
1072 	 */
1073 	console_init();
1074 	if (panic_later)
1075 		panic("Too many boot %s vars at `%s'", panic_later,
1076 		      panic_param);
1077 
1078 	lockdep_init();
1079 
1080 	/*
1081 	 * Need to run this when irqs are enabled, because it wants
1082 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1083 	 * too:
1084 	 */
1085 	locking_selftest();
1086 
1087 	/*
1088 	 * This needs to be called before any devices perform DMA
1089 	 * operations that might use the SWIOTLB bounce buffers. It will
1090 	 * mark the bounce buffers as decrypted so that their usage will
1091 	 * not cause "plain-text" data to be decrypted when accessed.
1092 	 */
1093 	mem_encrypt_init();
1094 
1095 #ifdef CONFIG_BLK_DEV_INITRD
1096 	if (initrd_start && !initrd_below_start_ok &&
1097 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1098 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1099 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1100 		    min_low_pfn);
1101 		initrd_start = 0;
1102 	}
1103 #endif
1104 	setup_per_cpu_pageset();
1105 	numa_policy_init();
1106 	acpi_early_init();
1107 	if (late_time_init)
1108 		late_time_init();
1109 	sched_clock_init();
1110 	calibrate_delay();
1111 	pid_idr_init();
1112 	anon_vma_init();
1113 #ifdef CONFIG_X86
1114 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1115 		efi_enter_virtual_mode();
1116 #endif
1117 	thread_stack_cache_init();
1118 	cred_init();
1119 	fork_init();
1120 	proc_caches_init();
1121 	uts_ns_init();
1122 	key_init();
1123 	security_init();
1124 	dbg_late_init();
1125 	net_ns_init();
1126 	vfs_caches_init();
1127 	pagecache_init();
1128 	signals_init();
1129 	seq_file_init();
1130 	proc_root_init();
1131 	nsfs_init();
1132 	cpuset_init();
1133 	cgroup_init();
1134 	taskstats_init_early();
1135 	delayacct_init();
1136 
1137 	poking_init();
1138 	check_bugs();
1139 
1140 	acpi_subsystem_init();
1141 	arch_post_acpi_subsys_init();
1142 	kcsan_init();
1143 
1144 	/* Do the rest non-__init'ed, we're now alive */
1145 	arch_call_rest_init();
1146 
1147 	prevent_tail_call_optimization();
1148 }
1149 
1150 /* Call all constructor functions linked into the kernel. */
1151 static void __init do_ctors(void)
1152 {
1153 /*
1154  * For UML, the constructors have already been called by the
1155  * normal setup code as it's just a normal ELF binary, so we
1156  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1157  * even on UML for modules.
1158  */
1159 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1160 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1161 
1162 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1163 		(*fn)();
1164 #endif
1165 }
1166 
1167 #ifdef CONFIG_KALLSYMS
1168 struct blacklist_entry {
1169 	struct list_head next;
1170 	char *buf;
1171 };
1172 
1173 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1174 
1175 static int __init initcall_blacklist(char *str)
1176 {
1177 	char *str_entry;
1178 	struct blacklist_entry *entry;
1179 
1180 	/* str argument is a comma-separated list of functions */
1181 	do {
1182 		str_entry = strsep(&str, ",");
1183 		if (str_entry) {
1184 			pr_debug("blacklisting initcall %s\n", str_entry);
1185 			entry = memblock_alloc(sizeof(*entry),
1186 					       SMP_CACHE_BYTES);
1187 			if (!entry)
1188 				panic("%s: Failed to allocate %zu bytes\n",
1189 				      __func__, sizeof(*entry));
1190 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1191 						    SMP_CACHE_BYTES);
1192 			if (!entry->buf)
1193 				panic("%s: Failed to allocate %zu bytes\n",
1194 				      __func__, strlen(str_entry) + 1);
1195 			strcpy(entry->buf, str_entry);
1196 			list_add(&entry->next, &blacklisted_initcalls);
1197 		}
1198 	} while (str_entry);
1199 
1200 	return 1;
1201 }
1202 
1203 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1204 {
1205 	struct blacklist_entry *entry;
1206 	char fn_name[KSYM_SYMBOL_LEN];
1207 	unsigned long addr;
1208 
1209 	if (list_empty(&blacklisted_initcalls))
1210 		return false;
1211 
1212 	addr = (unsigned long) dereference_function_descriptor(fn);
1213 	sprint_symbol_no_offset(fn_name, addr);
1214 
1215 	/*
1216 	 * fn will be "function_name [module_name]" where [module_name] is not
1217 	 * displayed for built-in init functions.  Strip off the [module_name].
1218 	 */
1219 	strreplace(fn_name, ' ', '\0');
1220 
1221 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1222 		if (!strcmp(fn_name, entry->buf)) {
1223 			pr_debug("initcall %s blacklisted\n", fn_name);
1224 			return true;
1225 		}
1226 	}
1227 
1228 	return false;
1229 }
1230 #else
1231 static int __init initcall_blacklist(char *str)
1232 {
1233 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1234 	return 0;
1235 }
1236 
1237 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1238 {
1239 	return false;
1240 }
1241 #endif
1242 __setup("initcall_blacklist=", initcall_blacklist);
1243 
1244 static __init_or_module void
1245 trace_initcall_start_cb(void *data, initcall_t fn)
1246 {
1247 	ktime_t *calltime = data;
1248 
1249 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1250 	*calltime = ktime_get();
1251 }
1252 
1253 static __init_or_module void
1254 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1255 {
1256 	ktime_t rettime, *calltime = data;
1257 
1258 	rettime = ktime_get();
1259 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1260 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1261 }
1262 
1263 static ktime_t initcall_calltime;
1264 
1265 #ifdef TRACEPOINTS_ENABLED
1266 static void __init initcall_debug_enable(void)
1267 {
1268 	int ret;
1269 
1270 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1271 					    &initcall_calltime);
1272 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1273 					      &initcall_calltime);
1274 	WARN(ret, "Failed to register initcall tracepoints\n");
1275 }
1276 # define do_trace_initcall_start	trace_initcall_start
1277 # define do_trace_initcall_finish	trace_initcall_finish
1278 #else
1279 static inline void do_trace_initcall_start(initcall_t fn)
1280 {
1281 	if (!initcall_debug)
1282 		return;
1283 	trace_initcall_start_cb(&initcall_calltime, fn);
1284 }
1285 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1286 {
1287 	if (!initcall_debug)
1288 		return;
1289 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1290 }
1291 #endif /* !TRACEPOINTS_ENABLED */
1292 
1293 int __init_or_module do_one_initcall(initcall_t fn)
1294 {
1295 	int count = preempt_count();
1296 	char msgbuf[64];
1297 	int ret;
1298 
1299 	if (initcall_blacklisted(fn))
1300 		return -EPERM;
1301 
1302 	do_trace_initcall_start(fn);
1303 	ret = fn();
1304 	do_trace_initcall_finish(fn, ret);
1305 
1306 	msgbuf[0] = 0;
1307 
1308 	if (preempt_count() != count) {
1309 		sprintf(msgbuf, "preemption imbalance ");
1310 		preempt_count_set(count);
1311 	}
1312 	if (irqs_disabled()) {
1313 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1314 		local_irq_enable();
1315 	}
1316 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1317 
1318 	add_latent_entropy();
1319 	return ret;
1320 }
1321 
1322 
1323 extern initcall_entry_t __initcall_start[];
1324 extern initcall_entry_t __initcall0_start[];
1325 extern initcall_entry_t __initcall1_start[];
1326 extern initcall_entry_t __initcall2_start[];
1327 extern initcall_entry_t __initcall3_start[];
1328 extern initcall_entry_t __initcall4_start[];
1329 extern initcall_entry_t __initcall5_start[];
1330 extern initcall_entry_t __initcall6_start[];
1331 extern initcall_entry_t __initcall7_start[];
1332 extern initcall_entry_t __initcall_end[];
1333 
1334 static initcall_entry_t *initcall_levels[] __initdata = {
1335 	__initcall0_start,
1336 	__initcall1_start,
1337 	__initcall2_start,
1338 	__initcall3_start,
1339 	__initcall4_start,
1340 	__initcall5_start,
1341 	__initcall6_start,
1342 	__initcall7_start,
1343 	__initcall_end,
1344 };
1345 
1346 /* Keep these in sync with initcalls in include/linux/init.h */
1347 static const char *initcall_level_names[] __initdata = {
1348 	"pure",
1349 	"core",
1350 	"postcore",
1351 	"arch",
1352 	"subsys",
1353 	"fs",
1354 	"device",
1355 	"late",
1356 };
1357 
1358 static int __init ignore_unknown_bootoption(char *param, char *val,
1359 			       const char *unused, void *arg)
1360 {
1361 	return 0;
1362 }
1363 
1364 static void __init do_initcall_level(int level, char *command_line)
1365 {
1366 	initcall_entry_t *fn;
1367 
1368 	parse_args(initcall_level_names[level],
1369 		   command_line, __start___param,
1370 		   __stop___param - __start___param,
1371 		   level, level,
1372 		   NULL, ignore_unknown_bootoption);
1373 
1374 	trace_initcall_level(initcall_level_names[level]);
1375 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1376 		do_one_initcall(initcall_from_entry(fn));
1377 }
1378 
1379 static void __init do_initcalls(void)
1380 {
1381 	int level;
1382 	size_t len = strlen(saved_command_line) + 1;
1383 	char *command_line;
1384 
1385 	command_line = kzalloc(len, GFP_KERNEL);
1386 	if (!command_line)
1387 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1388 
1389 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1390 		/* Parser modifies command_line, restore it each time */
1391 		strcpy(command_line, saved_command_line);
1392 		do_initcall_level(level, command_line);
1393 	}
1394 
1395 	kfree(command_line);
1396 }
1397 
1398 /*
1399  * Ok, the machine is now initialized. None of the devices
1400  * have been touched yet, but the CPU subsystem is up and
1401  * running, and memory and process management works.
1402  *
1403  * Now we can finally start doing some real work..
1404  */
1405 static void __init do_basic_setup(void)
1406 {
1407 	cpuset_init_smp();
1408 	driver_init();
1409 	init_irq_proc();
1410 	do_ctors();
1411 	do_initcalls();
1412 }
1413 
1414 static void __init do_pre_smp_initcalls(void)
1415 {
1416 	initcall_entry_t *fn;
1417 
1418 	trace_initcall_level("early");
1419 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1420 		do_one_initcall(initcall_from_entry(fn));
1421 }
1422 
1423 static int run_init_process(const char *init_filename)
1424 {
1425 	const char *const *p;
1426 
1427 	argv_init[0] = init_filename;
1428 	pr_info("Run %s as init process\n", init_filename);
1429 	pr_debug("  with arguments:\n");
1430 	for (p = argv_init; *p; p++)
1431 		pr_debug("    %s\n", *p);
1432 	pr_debug("  with environment:\n");
1433 	for (p = envp_init; *p; p++)
1434 		pr_debug("    %s\n", *p);
1435 	return kernel_execve(init_filename, argv_init, envp_init);
1436 }
1437 
1438 static int try_to_run_init_process(const char *init_filename)
1439 {
1440 	int ret;
1441 
1442 	ret = run_init_process(init_filename);
1443 
1444 	if (ret && ret != -ENOENT) {
1445 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1446 		       init_filename, ret);
1447 	}
1448 
1449 	return ret;
1450 }
1451 
1452 static noinline void __init kernel_init_freeable(void);
1453 
1454 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1455 bool rodata_enabled __ro_after_init = true;
1456 
1457 #ifndef arch_parse_debug_rodata
1458 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1459 #endif
1460 
1461 static int __init set_debug_rodata(char *str)
1462 {
1463 	if (arch_parse_debug_rodata(str))
1464 		return 0;
1465 
1466 	if (str && !strcmp(str, "on"))
1467 		rodata_enabled = true;
1468 	else if (str && !strcmp(str, "off"))
1469 		rodata_enabled = false;
1470 	else
1471 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1472 	return 0;
1473 }
1474 early_param("rodata", set_debug_rodata);
1475 #endif
1476 
1477 #ifdef CONFIG_STRICT_KERNEL_RWX
1478 static void mark_readonly(void)
1479 {
1480 	if (rodata_enabled) {
1481 		/*
1482 		 * load_module() results in W+X mappings, which are cleaned
1483 		 * up with call_rcu().  Let's make sure that queued work is
1484 		 * flushed so that we don't hit false positives looking for
1485 		 * insecure pages which are W+X.
1486 		 */
1487 		rcu_barrier();
1488 		mark_rodata_ro();
1489 		rodata_test();
1490 	} else
1491 		pr_info("Kernel memory protection disabled.\n");
1492 }
1493 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1494 static inline void mark_readonly(void)
1495 {
1496 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1497 }
1498 #else
1499 static inline void mark_readonly(void)
1500 {
1501 	pr_warn("This architecture does not have kernel memory protection.\n");
1502 }
1503 #endif
1504 
1505 void __weak free_initmem(void)
1506 {
1507 	free_initmem_default(POISON_FREE_INITMEM);
1508 }
1509 
1510 static int __ref kernel_init(void *unused)
1511 {
1512 	int ret;
1513 
1514 	/*
1515 	 * Wait until kthreadd is all set-up.
1516 	 */
1517 	wait_for_completion(&kthreadd_done);
1518 
1519 	kernel_init_freeable();
1520 	/* need to finish all async __init code before freeing the memory */
1521 	async_synchronize_full();
1522 
1523 	system_state = SYSTEM_FREEING_INITMEM;
1524 	kprobe_free_init_mem();
1525 	ftrace_free_init_mem();
1526 	kgdb_free_init_mem();
1527 	exit_boot_config();
1528 	free_initmem();
1529 	mark_readonly();
1530 
1531 	/*
1532 	 * Kernel mappings are now finalized - update the userspace page-table
1533 	 * to finalize PTI.
1534 	 */
1535 	pti_finalize();
1536 
1537 	system_state = SYSTEM_RUNNING;
1538 	numa_default_policy();
1539 
1540 	rcu_end_inkernel_boot();
1541 
1542 	do_sysctl_args();
1543 
1544 	if (ramdisk_execute_command) {
1545 		ret = run_init_process(ramdisk_execute_command);
1546 		if (!ret)
1547 			return 0;
1548 		pr_err("Failed to execute %s (error %d)\n",
1549 		       ramdisk_execute_command, ret);
1550 	}
1551 
1552 	/*
1553 	 * We try each of these until one succeeds.
1554 	 *
1555 	 * The Bourne shell can be used instead of init if we are
1556 	 * trying to recover a really broken machine.
1557 	 */
1558 	if (execute_command) {
1559 		ret = run_init_process(execute_command);
1560 		if (!ret)
1561 			return 0;
1562 		panic("Requested init %s failed (error %d).",
1563 		      execute_command, ret);
1564 	}
1565 
1566 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1567 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1568 		if (ret)
1569 			pr_err("Default init %s failed (error %d)\n",
1570 			       CONFIG_DEFAULT_INIT, ret);
1571 		else
1572 			return 0;
1573 	}
1574 
1575 	if (!try_to_run_init_process("/sbin/init") ||
1576 	    !try_to_run_init_process("/etc/init") ||
1577 	    !try_to_run_init_process("/bin/init") ||
1578 	    !try_to_run_init_process("/bin/sh"))
1579 		return 0;
1580 
1581 	panic("No working init found.  Try passing init= option to kernel. "
1582 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1583 }
1584 
1585 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1586 void __init console_on_rootfs(void)
1587 {
1588 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1589 
1590 	if (IS_ERR(file)) {
1591 		pr_err("Warning: unable to open an initial console.\n");
1592 		return;
1593 	}
1594 	init_dup(file);
1595 	init_dup(file);
1596 	init_dup(file);
1597 	fput(file);
1598 }
1599 
1600 static noinline void __init kernel_init_freeable(void)
1601 {
1602 	/* Now the scheduler is fully set up and can do blocking allocations */
1603 	gfp_allowed_mask = __GFP_BITS_MASK;
1604 
1605 	/*
1606 	 * init can allocate pages on any node
1607 	 */
1608 	set_mems_allowed(node_states[N_MEMORY]);
1609 
1610 	cad_pid = get_pid(task_pid(current));
1611 
1612 	smp_prepare_cpus(setup_max_cpus);
1613 
1614 	workqueue_init();
1615 
1616 	init_mm_internals();
1617 
1618 	rcu_init_tasks_generic();
1619 	do_pre_smp_initcalls();
1620 	lockup_detector_init();
1621 
1622 	smp_init();
1623 	sched_init_smp();
1624 
1625 	padata_init();
1626 	page_alloc_init_late();
1627 	/* Initialize page ext after all struct pages are initialized. */
1628 	if (!early_page_ext_enabled())
1629 		page_ext_init();
1630 
1631 	do_basic_setup();
1632 
1633 	kunit_run_all_tests();
1634 
1635 	wait_for_initramfs();
1636 	console_on_rootfs();
1637 
1638 	/*
1639 	 * check if there is an early userspace init.  If yes, let it do all
1640 	 * the work
1641 	 */
1642 	if (init_eaccess(ramdisk_execute_command) != 0) {
1643 		ramdisk_execute_command = NULL;
1644 		prepare_namespace();
1645 	}
1646 
1647 	/*
1648 	 * Ok, we have completed the initial bootup, and
1649 	 * we're essentially up and running. Get rid of the
1650 	 * initmem segments and start the user-mode stuff..
1651 	 *
1652 	 * rootfs is available now, try loading the public keys
1653 	 * and default modules
1654 	 */
1655 
1656 	integrity_load_keys();
1657 }
1658