xref: /openbmc/linux/init/main.c (revision 11788d9b)
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/vmalloc.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/start_kernel.h>
39 #include <linux/security.h>
40 #include <linux/smp.h>
41 #include <linux/profile.h>
42 #include <linux/rcupdate.h>
43 #include <linux/moduleparam.h>
44 #include <linux/kallsyms.h>
45 #include <linux/writeback.h>
46 #include <linux/cpu.h>
47 #include <linux/cpuset.h>
48 #include <linux/cgroup.h>
49 #include <linux/efi.h>
50 #include <linux/tick.h>
51 #include <linux/sched/isolation.h>
52 #include <linux/interrupt.h>
53 #include <linux/taskstats_kern.h>
54 #include <linux/delayacct.h>
55 #include <linux/unistd.h>
56 #include <linux/utsname.h>
57 #include <linux/rmap.h>
58 #include <linux/mempolicy.h>
59 #include <linux/key.h>
60 #include <linux/buffer_head.h>
61 #include <linux/page_ext.h>
62 #include <linux/debug_locks.h>
63 #include <linux/debugobjects.h>
64 #include <linux/lockdep.h>
65 #include <linux/kmemleak.h>
66 #include <linux/padata.h>
67 #include <linux/pid_namespace.h>
68 #include <linux/device/driver.h>
69 #include <linux/kthread.h>
70 #include <linux/sched.h>
71 #include <linux/sched/init.h>
72 #include <linux/signal.h>
73 #include <linux/idr.h>
74 #include <linux/kgdb.h>
75 #include <linux/ftrace.h>
76 #include <linux/async.h>
77 #include <linux/sfi.h>
78 #include <linux/shmem_fs.h>
79 #include <linux/slab.h>
80 #include <linux/perf_event.h>
81 #include <linux/ptrace.h>
82 #include <linux/pti.h>
83 #include <linux/blkdev.h>
84 #include <linux/elevator.h>
85 #include <linux/sched/clock.h>
86 #include <linux/sched/task.h>
87 #include <linux/sched/task_stack.h>
88 #include <linux/context_tracking.h>
89 #include <linux/random.h>
90 #include <linux/list.h>
91 #include <linux/integrity.h>
92 #include <linux/proc_ns.h>
93 #include <linux/io.h>
94 #include <linux/cache.h>
95 #include <linux/rodata_test.h>
96 #include <linux/jump_label.h>
97 #include <linux/mem_encrypt.h>
98 #include <linux/kcsan.h>
99 #include <linux/init_syscalls.h>
100 
101 #include <asm/io.h>
102 #include <asm/bugs.h>
103 #include <asm/setup.h>
104 #include <asm/sections.h>
105 #include <asm/cacheflush.h>
106 
107 #define CREATE_TRACE_POINTS
108 #include <trace/events/initcall.h>
109 
110 static int kernel_init(void *);
111 
112 extern void init_IRQ(void);
113 extern void radix_tree_init(void);
114 
115 /*
116  * Debug helper: via this flag we know that we are in 'early bootup code'
117  * where only the boot processor is running with IRQ disabled.  This means
118  * two things - IRQ must not be enabled before the flag is cleared and some
119  * operations which are not allowed with IRQ disabled are allowed while the
120  * flag is set.
121  */
122 bool early_boot_irqs_disabled __read_mostly;
123 
124 enum system_states system_state __read_mostly;
125 EXPORT_SYMBOL(system_state);
126 
127 /*
128  * Boot command-line arguments
129  */
130 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
131 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
132 
133 extern void time_init(void);
134 /* Default late time init is NULL. archs can override this later. */
135 void (*__initdata late_time_init)(void);
136 
137 /* Untouched command line saved by arch-specific code. */
138 char __initdata boot_command_line[COMMAND_LINE_SIZE];
139 /* Untouched saved command line (eg. for /proc) */
140 char *saved_command_line;
141 /* Command line for parameter parsing */
142 static char *static_command_line;
143 /* Untouched extra command line */
144 static char *extra_command_line;
145 /* Extra init arguments */
146 static char *extra_init_args;
147 
148 #ifdef CONFIG_BOOT_CONFIG
149 /* Is bootconfig on command line? */
150 static bool bootconfig_found;
151 static bool initargs_found;
152 #else
153 # define bootconfig_found false
154 # define initargs_found false
155 #endif
156 
157 static char *execute_command;
158 static char *ramdisk_execute_command = "/init";
159 
160 /*
161  * Used to generate warnings if static_key manipulation functions are used
162  * before jump_label_init is called.
163  */
164 bool static_key_initialized __read_mostly;
165 EXPORT_SYMBOL_GPL(static_key_initialized);
166 
167 /*
168  * If set, this is an indication to the drivers that reset the underlying
169  * device before going ahead with the initialization otherwise driver might
170  * rely on the BIOS and skip the reset operation.
171  *
172  * This is useful if kernel is booting in an unreliable environment.
173  * For ex. kdump situation where previous kernel has crashed, BIOS has been
174  * skipped and devices will be in unknown state.
175  */
176 unsigned int reset_devices;
177 EXPORT_SYMBOL(reset_devices);
178 
179 static int __init set_reset_devices(char *str)
180 {
181 	reset_devices = 1;
182 	return 1;
183 }
184 
185 __setup("reset_devices", set_reset_devices);
186 
187 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
188 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
189 static const char *panic_later, *panic_param;
190 
191 extern const struct obs_kernel_param __setup_start[], __setup_end[];
192 
193 static bool __init obsolete_checksetup(char *line)
194 {
195 	const struct obs_kernel_param *p;
196 	bool had_early_param = false;
197 
198 	p = __setup_start;
199 	do {
200 		int n = strlen(p->str);
201 		if (parameqn(line, p->str, n)) {
202 			if (p->early) {
203 				/* Already done in parse_early_param?
204 				 * (Needs exact match on param part).
205 				 * Keep iterating, as we can have early
206 				 * params and __setups of same names 8( */
207 				if (line[n] == '\0' || line[n] == '=')
208 					had_early_param = true;
209 			} else if (!p->setup_func) {
210 				pr_warn("Parameter %s is obsolete, ignored\n",
211 					p->str);
212 				return true;
213 			} else if (p->setup_func(line + n))
214 				return true;
215 		}
216 		p++;
217 	} while (p < __setup_end);
218 
219 	return had_early_param;
220 }
221 
222 /*
223  * This should be approx 2 Bo*oMips to start (note initial shift), and will
224  * still work even if initially too large, it will just take slightly longer
225  */
226 unsigned long loops_per_jiffy = (1<<12);
227 EXPORT_SYMBOL(loops_per_jiffy);
228 
229 static int __init debug_kernel(char *str)
230 {
231 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
232 	return 0;
233 }
234 
235 static int __init quiet_kernel(char *str)
236 {
237 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
238 	return 0;
239 }
240 
241 early_param("debug", debug_kernel);
242 early_param("quiet", quiet_kernel);
243 
244 static int __init loglevel(char *str)
245 {
246 	int newlevel;
247 
248 	/*
249 	 * Only update loglevel value when a correct setting was passed,
250 	 * to prevent blind crashes (when loglevel being set to 0) that
251 	 * are quite hard to debug
252 	 */
253 	if (get_option(&str, &newlevel)) {
254 		console_loglevel = newlevel;
255 		return 0;
256 	}
257 
258 	return -EINVAL;
259 }
260 
261 early_param("loglevel", loglevel);
262 
263 #ifdef CONFIG_BLK_DEV_INITRD
264 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
265 {
266 	u32 size, csum;
267 	char *data;
268 	u32 *hdr;
269 
270 	if (!initrd_end)
271 		return NULL;
272 
273 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
274 	if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
275 		return NULL;
276 
277 	hdr = (u32 *)(data - 8);
278 	size = hdr[0];
279 	csum = hdr[1];
280 
281 	data = ((void *)hdr) - size;
282 	if ((unsigned long)data < initrd_start) {
283 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
284 			size, initrd_end - initrd_start);
285 		return NULL;
286 	}
287 
288 	/* Remove bootconfig from initramfs/initrd */
289 	initrd_end = (unsigned long)data;
290 	if (_size)
291 		*_size = size;
292 	if (_csum)
293 		*_csum = csum;
294 
295 	return data;
296 }
297 #else
298 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
299 {
300 	return NULL;
301 }
302 #endif
303 
304 #ifdef CONFIG_BOOT_CONFIG
305 
306 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
307 
308 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
309 
310 static int __init xbc_snprint_cmdline(char *buf, size_t size,
311 				      struct xbc_node *root)
312 {
313 	struct xbc_node *knode, *vnode;
314 	char *end = buf + size;
315 	const char *val;
316 	int ret;
317 
318 	xbc_node_for_each_key_value(root, knode, val) {
319 		ret = xbc_node_compose_key_after(root, knode,
320 					xbc_namebuf, XBC_KEYLEN_MAX);
321 		if (ret < 0)
322 			return ret;
323 
324 		vnode = xbc_node_get_child(knode);
325 		if (!vnode) {
326 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
327 			if (ret < 0)
328 				return ret;
329 			buf += ret;
330 			continue;
331 		}
332 		xbc_array_for_each_value(vnode, val) {
333 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
334 				       xbc_namebuf, val);
335 			if (ret < 0)
336 				return ret;
337 			buf += ret;
338 		}
339 	}
340 
341 	return buf - (end - size);
342 }
343 #undef rest
344 
345 /* Make an extra command line under given key word */
346 static char * __init xbc_make_cmdline(const char *key)
347 {
348 	struct xbc_node *root;
349 	char *new_cmdline;
350 	int ret, len = 0;
351 
352 	root = xbc_find_node(key);
353 	if (!root)
354 		return NULL;
355 
356 	/* Count required buffer size */
357 	len = xbc_snprint_cmdline(NULL, 0, root);
358 	if (len <= 0)
359 		return NULL;
360 
361 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
362 	if (!new_cmdline) {
363 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
364 		return NULL;
365 	}
366 
367 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
368 	if (ret < 0 || ret > len) {
369 		pr_err("Failed to print extra kernel cmdline.\n");
370 		return NULL;
371 	}
372 
373 	return new_cmdline;
374 }
375 
376 static u32 boot_config_checksum(unsigned char *p, u32 size)
377 {
378 	u32 ret = 0;
379 
380 	while (size--)
381 		ret += *p++;
382 
383 	return ret;
384 }
385 
386 static int __init bootconfig_params(char *param, char *val,
387 				    const char *unused, void *arg)
388 {
389 	if (strcmp(param, "bootconfig") == 0) {
390 		bootconfig_found = true;
391 	}
392 	return 0;
393 }
394 
395 static void __init setup_boot_config(const char *cmdline)
396 {
397 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
398 	const char *msg;
399 	int pos;
400 	u32 size, csum;
401 	char *data, *copy, *err;
402 	int ret;
403 
404 	/* Cut out the bootconfig data even if we have no bootconfig option */
405 	data = get_boot_config_from_initrd(&size, &csum);
406 
407 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
408 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
409 			 bootconfig_params);
410 
411 	if (IS_ERR(err) || !bootconfig_found)
412 		return;
413 
414 	/* parse_args() stops at '--' and returns an address */
415 	if (err)
416 		initargs_found = true;
417 
418 	if (!data) {
419 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
420 		return;
421 	}
422 
423 	if (size >= XBC_DATA_MAX) {
424 		pr_err("bootconfig size %d greater than max size %d\n",
425 			size, XBC_DATA_MAX);
426 		return;
427 	}
428 
429 	if (boot_config_checksum((unsigned char *)data, size) != csum) {
430 		pr_err("bootconfig checksum failed\n");
431 		return;
432 	}
433 
434 	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
435 	if (!copy) {
436 		pr_err("Failed to allocate memory for bootconfig\n");
437 		return;
438 	}
439 
440 	memcpy(copy, data, size);
441 	copy[size] = '\0';
442 
443 	ret = xbc_init(copy, &msg, &pos);
444 	if (ret < 0) {
445 		if (pos < 0)
446 			pr_err("Failed to init bootconfig: %s.\n", msg);
447 		else
448 			pr_err("Failed to parse bootconfig: %s at %d.\n",
449 				msg, pos);
450 	} else {
451 		pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
452 		/* keys starting with "kernel." are passed via cmdline */
453 		extra_command_line = xbc_make_cmdline("kernel");
454 		/* Also, "init." keys are init arguments */
455 		extra_init_args = xbc_make_cmdline("init");
456 	}
457 	return;
458 }
459 
460 #else
461 
462 static void __init setup_boot_config(const char *cmdline)
463 {
464 	/* Remove bootconfig data from initrd */
465 	get_boot_config_from_initrd(NULL, NULL);
466 }
467 
468 static int __init warn_bootconfig(char *str)
469 {
470 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
471 	return 0;
472 }
473 early_param("bootconfig", warn_bootconfig);
474 
475 #endif
476 
477 /* Change NUL term back to "=", to make "param" the whole string. */
478 static void __init repair_env_string(char *param, char *val)
479 {
480 	if (val) {
481 		/* param=val or param="val"? */
482 		if (val == param+strlen(param)+1)
483 			val[-1] = '=';
484 		else if (val == param+strlen(param)+2) {
485 			val[-2] = '=';
486 			memmove(val-1, val, strlen(val)+1);
487 		} else
488 			BUG();
489 	}
490 }
491 
492 /* Anything after -- gets handed straight to init. */
493 static int __init set_init_arg(char *param, char *val,
494 			       const char *unused, void *arg)
495 {
496 	unsigned int i;
497 
498 	if (panic_later)
499 		return 0;
500 
501 	repair_env_string(param, val);
502 
503 	for (i = 0; argv_init[i]; i++) {
504 		if (i == MAX_INIT_ARGS) {
505 			panic_later = "init";
506 			panic_param = param;
507 			return 0;
508 		}
509 	}
510 	argv_init[i] = param;
511 	return 0;
512 }
513 
514 /*
515  * Unknown boot options get handed to init, unless they look like
516  * unused parameters (modprobe will find them in /proc/cmdline).
517  */
518 static int __init unknown_bootoption(char *param, char *val,
519 				     const char *unused, void *arg)
520 {
521 	size_t len = strlen(param);
522 
523 	repair_env_string(param, val);
524 
525 	/* Handle obsolete-style parameters */
526 	if (obsolete_checksetup(param))
527 		return 0;
528 
529 	/* Unused module parameter. */
530 	if (strnchr(param, len, '.'))
531 		return 0;
532 
533 	if (panic_later)
534 		return 0;
535 
536 	if (val) {
537 		/* Environment option */
538 		unsigned int i;
539 		for (i = 0; envp_init[i]; i++) {
540 			if (i == MAX_INIT_ENVS) {
541 				panic_later = "env";
542 				panic_param = param;
543 			}
544 			if (!strncmp(param, envp_init[i], len+1))
545 				break;
546 		}
547 		envp_init[i] = param;
548 	} else {
549 		/* Command line option */
550 		unsigned int i;
551 		for (i = 0; argv_init[i]; i++) {
552 			if (i == MAX_INIT_ARGS) {
553 				panic_later = "init";
554 				panic_param = param;
555 			}
556 		}
557 		argv_init[i] = param;
558 	}
559 	return 0;
560 }
561 
562 static int __init init_setup(char *str)
563 {
564 	unsigned int i;
565 
566 	execute_command = str;
567 	/*
568 	 * In case LILO is going to boot us with default command line,
569 	 * it prepends "auto" before the whole cmdline which makes
570 	 * the shell think it should execute a script with such name.
571 	 * So we ignore all arguments entered _before_ init=... [MJ]
572 	 */
573 	for (i = 1; i < MAX_INIT_ARGS; i++)
574 		argv_init[i] = NULL;
575 	return 1;
576 }
577 __setup("init=", init_setup);
578 
579 static int __init rdinit_setup(char *str)
580 {
581 	unsigned int i;
582 
583 	ramdisk_execute_command = str;
584 	/* See "auto" comment in init_setup */
585 	for (i = 1; i < MAX_INIT_ARGS; i++)
586 		argv_init[i] = NULL;
587 	return 1;
588 }
589 __setup("rdinit=", rdinit_setup);
590 
591 #ifndef CONFIG_SMP
592 static const unsigned int setup_max_cpus = NR_CPUS;
593 static inline void setup_nr_cpu_ids(void) { }
594 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
595 #endif
596 
597 /*
598  * We need to store the untouched command line for future reference.
599  * We also need to store the touched command line since the parameter
600  * parsing is performed in place, and we should allow a component to
601  * store reference of name/value for future reference.
602  */
603 static void __init setup_command_line(char *command_line)
604 {
605 	size_t len, xlen = 0, ilen = 0;
606 
607 	if (extra_command_line)
608 		xlen = strlen(extra_command_line);
609 	if (extra_init_args)
610 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
611 
612 	len = xlen + strlen(boot_command_line) + 1;
613 
614 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
615 	if (!saved_command_line)
616 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
617 
618 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
619 	if (!static_command_line)
620 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
621 
622 	if (xlen) {
623 		/*
624 		 * We have to put extra_command_line before boot command
625 		 * lines because there could be dashes (separator of init
626 		 * command line) in the command lines.
627 		 */
628 		strcpy(saved_command_line, extra_command_line);
629 		strcpy(static_command_line, extra_command_line);
630 	}
631 	strcpy(saved_command_line + xlen, boot_command_line);
632 	strcpy(static_command_line + xlen, command_line);
633 
634 	if (ilen) {
635 		/*
636 		 * Append supplemental init boot args to saved_command_line
637 		 * so that user can check what command line options passed
638 		 * to init.
639 		 */
640 		len = strlen(saved_command_line);
641 		if (initargs_found) {
642 			saved_command_line[len++] = ' ';
643 		} else {
644 			strcpy(saved_command_line + len, " -- ");
645 			len += 4;
646 		}
647 
648 		strcpy(saved_command_line + len, extra_init_args);
649 	}
650 }
651 
652 /*
653  * We need to finalize in a non-__init function or else race conditions
654  * between the root thread and the init thread may cause start_kernel to
655  * be reaped by free_initmem before the root thread has proceeded to
656  * cpu_idle.
657  *
658  * gcc-3.4 accidentally inlines this function, so use noinline.
659  */
660 
661 static __initdata DECLARE_COMPLETION(kthreadd_done);
662 
663 noinline void __ref rest_init(void)
664 {
665 	struct task_struct *tsk;
666 	int pid;
667 
668 	rcu_scheduler_starting();
669 	/*
670 	 * We need to spawn init first so that it obtains pid 1, however
671 	 * the init task will end up wanting to create kthreads, which, if
672 	 * we schedule it before we create kthreadd, will OOPS.
673 	 */
674 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
675 	/*
676 	 * Pin init on the boot CPU. Task migration is not properly working
677 	 * until sched_init_smp() has been run. It will set the allowed
678 	 * CPUs for init to the non isolated CPUs.
679 	 */
680 	rcu_read_lock();
681 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
682 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
683 	rcu_read_unlock();
684 
685 	numa_default_policy();
686 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
687 	rcu_read_lock();
688 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
689 	rcu_read_unlock();
690 
691 	/*
692 	 * Enable might_sleep() and smp_processor_id() checks.
693 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
694 	 * kernel_thread() would trigger might_sleep() splats. With
695 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
696 	 * already, but it's stuck on the kthreadd_done completion.
697 	 */
698 	system_state = SYSTEM_SCHEDULING;
699 
700 	complete(&kthreadd_done);
701 
702 	/*
703 	 * The boot idle thread must execute schedule()
704 	 * at least once to get things moving:
705 	 */
706 	schedule_preempt_disabled();
707 	/* Call into cpu_idle with preempt disabled */
708 	cpu_startup_entry(CPUHP_ONLINE);
709 }
710 
711 /* Check for early params. */
712 static int __init do_early_param(char *param, char *val,
713 				 const char *unused, void *arg)
714 {
715 	const struct obs_kernel_param *p;
716 
717 	for (p = __setup_start; p < __setup_end; p++) {
718 		if ((p->early && parameq(param, p->str)) ||
719 		    (strcmp(param, "console") == 0 &&
720 		     strcmp(p->str, "earlycon") == 0)
721 		) {
722 			if (p->setup_func(val) != 0)
723 				pr_warn("Malformed early option '%s'\n", param);
724 		}
725 	}
726 	/* We accept everything at this stage. */
727 	return 0;
728 }
729 
730 void __init parse_early_options(char *cmdline)
731 {
732 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
733 		   do_early_param);
734 }
735 
736 /* Arch code calls this early on, or if not, just before other parsing. */
737 void __init parse_early_param(void)
738 {
739 	static int done __initdata;
740 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
741 
742 	if (done)
743 		return;
744 
745 	/* All fall through to do_early_param. */
746 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
747 	parse_early_options(tmp_cmdline);
748 	done = 1;
749 }
750 
751 void __init __weak arch_post_acpi_subsys_init(void) { }
752 
753 void __init __weak smp_setup_processor_id(void)
754 {
755 }
756 
757 # if THREAD_SIZE >= PAGE_SIZE
758 void __init __weak thread_stack_cache_init(void)
759 {
760 }
761 #endif
762 
763 void __init __weak mem_encrypt_init(void) { }
764 
765 void __init __weak poking_init(void) { }
766 
767 void __init __weak pgtable_cache_init(void) { }
768 
769 bool initcall_debug;
770 core_param(initcall_debug, initcall_debug, bool, 0644);
771 
772 #ifdef TRACEPOINTS_ENABLED
773 static void __init initcall_debug_enable(void);
774 #else
775 static inline void initcall_debug_enable(void)
776 {
777 }
778 #endif
779 
780 /* Report memory auto-initialization states for this boot. */
781 static void __init report_meminit(void)
782 {
783 	const char *stack;
784 
785 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
786 		stack = "all(pattern)";
787 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
788 		stack = "all(zero)";
789 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
790 		stack = "byref_all(zero)";
791 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
792 		stack = "byref(zero)";
793 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
794 		stack = "__user(zero)";
795 	else
796 		stack = "off";
797 
798 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
799 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
800 		want_init_on_free() ? "on" : "off");
801 	if (want_init_on_free())
802 		pr_info("mem auto-init: clearing system memory may take some time...\n");
803 }
804 
805 /*
806  * Set up kernel memory allocators
807  */
808 static void __init mm_init(void)
809 {
810 	/*
811 	 * page_ext requires contiguous pages,
812 	 * bigger than MAX_ORDER unless SPARSEMEM.
813 	 */
814 	page_ext_init_flatmem();
815 	init_debug_pagealloc();
816 	report_meminit();
817 	mem_init();
818 	kmem_cache_init();
819 	kmemleak_init();
820 	pgtable_init();
821 	debug_objects_mem_init();
822 	vmalloc_init();
823 	ioremap_huge_init();
824 	/* Should be run before the first non-init thread is created */
825 	init_espfix_bsp();
826 	/* Should be run after espfix64 is set up. */
827 	pti_init();
828 }
829 
830 void __init __weak arch_call_rest_init(void)
831 {
832 	rest_init();
833 }
834 
835 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
836 {
837 	char *command_line;
838 	char *after_dashes;
839 
840 	set_task_stack_end_magic(&init_task);
841 	smp_setup_processor_id();
842 	debug_objects_early_init();
843 
844 	cgroup_init_early();
845 
846 	local_irq_disable();
847 	early_boot_irqs_disabled = true;
848 
849 	/*
850 	 * Interrupts are still disabled. Do necessary setups, then
851 	 * enable them.
852 	 */
853 	boot_cpu_init();
854 	page_address_init();
855 	pr_notice("%s", linux_banner);
856 	early_security_init();
857 	setup_arch(&command_line);
858 	setup_boot_config(command_line);
859 	setup_command_line(command_line);
860 	setup_nr_cpu_ids();
861 	setup_per_cpu_areas();
862 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
863 	boot_cpu_hotplug_init();
864 
865 	build_all_zonelists(NULL);
866 	page_alloc_init();
867 
868 	pr_notice("Kernel command line: %s\n", saved_command_line);
869 	/* parameters may set static keys */
870 	jump_label_init();
871 	parse_early_param();
872 	after_dashes = parse_args("Booting kernel",
873 				  static_command_line, __start___param,
874 				  __stop___param - __start___param,
875 				  -1, -1, NULL, &unknown_bootoption);
876 	if (!IS_ERR_OR_NULL(after_dashes))
877 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
878 			   NULL, set_init_arg);
879 	if (extra_init_args)
880 		parse_args("Setting extra init args", extra_init_args,
881 			   NULL, 0, -1, -1, NULL, set_init_arg);
882 
883 	/*
884 	 * These use large bootmem allocations and must precede
885 	 * kmem_cache_init()
886 	 */
887 	setup_log_buf(0);
888 	vfs_caches_init_early();
889 	sort_main_extable();
890 	trap_init();
891 	mm_init();
892 
893 	ftrace_init();
894 
895 	/* trace_printk can be enabled here */
896 	early_trace_init();
897 
898 	/*
899 	 * Set up the scheduler prior starting any interrupts (such as the
900 	 * timer interrupt). Full topology setup happens at smp_init()
901 	 * time - but meanwhile we still have a functioning scheduler.
902 	 */
903 	sched_init();
904 	/*
905 	 * Disable preemption - early bootup scheduling is extremely
906 	 * fragile until we cpu_idle() for the first time.
907 	 */
908 	preempt_disable();
909 	if (WARN(!irqs_disabled(),
910 		 "Interrupts were enabled *very* early, fixing it\n"))
911 		local_irq_disable();
912 	radix_tree_init();
913 
914 	/*
915 	 * Set up housekeeping before setting up workqueues to allow the unbound
916 	 * workqueue to take non-housekeeping into account.
917 	 */
918 	housekeeping_init();
919 
920 	/*
921 	 * Allow workqueue creation and work item queueing/cancelling
922 	 * early.  Work item execution depends on kthreads and starts after
923 	 * workqueue_init().
924 	 */
925 	workqueue_init_early();
926 
927 	rcu_init();
928 
929 	/* Trace events are available after this */
930 	trace_init();
931 
932 	if (initcall_debug)
933 		initcall_debug_enable();
934 
935 	context_tracking_init();
936 	/* init some links before init_ISA_irqs() */
937 	early_irq_init();
938 	init_IRQ();
939 	tick_init();
940 	rcu_init_nohz();
941 	init_timers();
942 	hrtimers_init();
943 	softirq_init();
944 	timekeeping_init();
945 
946 	/*
947 	 * For best initial stack canary entropy, prepare it after:
948 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
949 	 * - timekeeping_init() for ktime entropy used in rand_initialize()
950 	 * - rand_initialize() to get any arch-specific entropy like RDRAND
951 	 * - add_latent_entropy() to get any latent entropy
952 	 * - adding command line entropy
953 	 */
954 	rand_initialize();
955 	add_latent_entropy();
956 	add_device_randomness(command_line, strlen(command_line));
957 	boot_init_stack_canary();
958 
959 	time_init();
960 	perf_event_init();
961 	profile_init();
962 	call_function_init();
963 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
964 
965 	early_boot_irqs_disabled = false;
966 	local_irq_enable();
967 
968 	kmem_cache_init_late();
969 
970 	/*
971 	 * HACK ALERT! This is early. We're enabling the console before
972 	 * we've done PCI setups etc, and console_init() must be aware of
973 	 * this. But we do want output early, in case something goes wrong.
974 	 */
975 	console_init();
976 	if (panic_later)
977 		panic("Too many boot %s vars at `%s'", panic_later,
978 		      panic_param);
979 
980 	lockdep_init();
981 
982 	/*
983 	 * Need to run this when irqs are enabled, because it wants
984 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
985 	 * too:
986 	 */
987 	locking_selftest();
988 
989 	/*
990 	 * This needs to be called before any devices perform DMA
991 	 * operations that might use the SWIOTLB bounce buffers. It will
992 	 * mark the bounce buffers as decrypted so that their usage will
993 	 * not cause "plain-text" data to be decrypted when accessed.
994 	 */
995 	mem_encrypt_init();
996 
997 #ifdef CONFIG_BLK_DEV_INITRD
998 	if (initrd_start && !initrd_below_start_ok &&
999 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1000 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1001 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1002 		    min_low_pfn);
1003 		initrd_start = 0;
1004 	}
1005 #endif
1006 	setup_per_cpu_pageset();
1007 	numa_policy_init();
1008 	acpi_early_init();
1009 	if (late_time_init)
1010 		late_time_init();
1011 	sched_clock_init();
1012 	calibrate_delay();
1013 	pid_idr_init();
1014 	anon_vma_init();
1015 #ifdef CONFIG_X86
1016 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1017 		efi_enter_virtual_mode();
1018 #endif
1019 	thread_stack_cache_init();
1020 	cred_init();
1021 	fork_init();
1022 	proc_caches_init();
1023 	uts_ns_init();
1024 	buffer_init();
1025 	key_init();
1026 	security_init();
1027 	dbg_late_init();
1028 	vfs_caches_init();
1029 	pagecache_init();
1030 	signals_init();
1031 	seq_file_init();
1032 	proc_root_init();
1033 	nsfs_init();
1034 	cpuset_init();
1035 	cgroup_init();
1036 	taskstats_init_early();
1037 	delayacct_init();
1038 
1039 	poking_init();
1040 	check_bugs();
1041 
1042 	acpi_subsystem_init();
1043 	arch_post_acpi_subsys_init();
1044 	sfi_init_late();
1045 	kcsan_init();
1046 
1047 	/* Do the rest non-__init'ed, we're now alive */
1048 	arch_call_rest_init();
1049 
1050 	prevent_tail_call_optimization();
1051 }
1052 
1053 /* Call all constructor functions linked into the kernel. */
1054 static void __init do_ctors(void)
1055 {
1056 #ifdef CONFIG_CONSTRUCTORS
1057 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1058 
1059 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1060 		(*fn)();
1061 #endif
1062 }
1063 
1064 #ifdef CONFIG_KALLSYMS
1065 struct blacklist_entry {
1066 	struct list_head next;
1067 	char *buf;
1068 };
1069 
1070 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1071 
1072 static int __init initcall_blacklist(char *str)
1073 {
1074 	char *str_entry;
1075 	struct blacklist_entry *entry;
1076 
1077 	/* str argument is a comma-separated list of functions */
1078 	do {
1079 		str_entry = strsep(&str, ",");
1080 		if (str_entry) {
1081 			pr_debug("blacklisting initcall %s\n", str_entry);
1082 			entry = memblock_alloc(sizeof(*entry),
1083 					       SMP_CACHE_BYTES);
1084 			if (!entry)
1085 				panic("%s: Failed to allocate %zu bytes\n",
1086 				      __func__, sizeof(*entry));
1087 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1088 						    SMP_CACHE_BYTES);
1089 			if (!entry->buf)
1090 				panic("%s: Failed to allocate %zu bytes\n",
1091 				      __func__, strlen(str_entry) + 1);
1092 			strcpy(entry->buf, str_entry);
1093 			list_add(&entry->next, &blacklisted_initcalls);
1094 		}
1095 	} while (str_entry);
1096 
1097 	return 0;
1098 }
1099 
1100 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1101 {
1102 	struct blacklist_entry *entry;
1103 	char fn_name[KSYM_SYMBOL_LEN];
1104 	unsigned long addr;
1105 
1106 	if (list_empty(&blacklisted_initcalls))
1107 		return false;
1108 
1109 	addr = (unsigned long) dereference_function_descriptor(fn);
1110 	sprint_symbol_no_offset(fn_name, addr);
1111 
1112 	/*
1113 	 * fn will be "function_name [module_name]" where [module_name] is not
1114 	 * displayed for built-in init functions.  Strip off the [module_name].
1115 	 */
1116 	strreplace(fn_name, ' ', '\0');
1117 
1118 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1119 		if (!strcmp(fn_name, entry->buf)) {
1120 			pr_debug("initcall %s blacklisted\n", fn_name);
1121 			return true;
1122 		}
1123 	}
1124 
1125 	return false;
1126 }
1127 #else
1128 static int __init initcall_blacklist(char *str)
1129 {
1130 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1131 	return 0;
1132 }
1133 
1134 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1135 {
1136 	return false;
1137 }
1138 #endif
1139 __setup("initcall_blacklist=", initcall_blacklist);
1140 
1141 static __init_or_module void
1142 trace_initcall_start_cb(void *data, initcall_t fn)
1143 {
1144 	ktime_t *calltime = (ktime_t *)data;
1145 
1146 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1147 	*calltime = ktime_get();
1148 }
1149 
1150 static __init_or_module void
1151 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1152 {
1153 	ktime_t *calltime = (ktime_t *)data;
1154 	ktime_t delta, rettime;
1155 	unsigned long long duration;
1156 
1157 	rettime = ktime_get();
1158 	delta = ktime_sub(rettime, *calltime);
1159 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1160 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1161 		 fn, ret, duration);
1162 }
1163 
1164 static ktime_t initcall_calltime;
1165 
1166 #ifdef TRACEPOINTS_ENABLED
1167 static void __init initcall_debug_enable(void)
1168 {
1169 	int ret;
1170 
1171 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1172 					    &initcall_calltime);
1173 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1174 					      &initcall_calltime);
1175 	WARN(ret, "Failed to register initcall tracepoints\n");
1176 }
1177 # define do_trace_initcall_start	trace_initcall_start
1178 # define do_trace_initcall_finish	trace_initcall_finish
1179 #else
1180 static inline void do_trace_initcall_start(initcall_t fn)
1181 {
1182 	if (!initcall_debug)
1183 		return;
1184 	trace_initcall_start_cb(&initcall_calltime, fn);
1185 }
1186 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1187 {
1188 	if (!initcall_debug)
1189 		return;
1190 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1191 }
1192 #endif /* !TRACEPOINTS_ENABLED */
1193 
1194 int __init_or_module do_one_initcall(initcall_t fn)
1195 {
1196 	int count = preempt_count();
1197 	char msgbuf[64];
1198 	int ret;
1199 
1200 	if (initcall_blacklisted(fn))
1201 		return -EPERM;
1202 
1203 	do_trace_initcall_start(fn);
1204 	ret = fn();
1205 	do_trace_initcall_finish(fn, ret);
1206 
1207 	msgbuf[0] = 0;
1208 
1209 	if (preempt_count() != count) {
1210 		sprintf(msgbuf, "preemption imbalance ");
1211 		preempt_count_set(count);
1212 	}
1213 	if (irqs_disabled()) {
1214 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1215 		local_irq_enable();
1216 	}
1217 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1218 
1219 	add_latent_entropy();
1220 	return ret;
1221 }
1222 
1223 
1224 extern initcall_entry_t __initcall_start[];
1225 extern initcall_entry_t __initcall0_start[];
1226 extern initcall_entry_t __initcall1_start[];
1227 extern initcall_entry_t __initcall2_start[];
1228 extern initcall_entry_t __initcall3_start[];
1229 extern initcall_entry_t __initcall4_start[];
1230 extern initcall_entry_t __initcall5_start[];
1231 extern initcall_entry_t __initcall6_start[];
1232 extern initcall_entry_t __initcall7_start[];
1233 extern initcall_entry_t __initcall_end[];
1234 
1235 static initcall_entry_t *initcall_levels[] __initdata = {
1236 	__initcall0_start,
1237 	__initcall1_start,
1238 	__initcall2_start,
1239 	__initcall3_start,
1240 	__initcall4_start,
1241 	__initcall5_start,
1242 	__initcall6_start,
1243 	__initcall7_start,
1244 	__initcall_end,
1245 };
1246 
1247 /* Keep these in sync with initcalls in include/linux/init.h */
1248 static const char *initcall_level_names[] __initdata = {
1249 	"pure",
1250 	"core",
1251 	"postcore",
1252 	"arch",
1253 	"subsys",
1254 	"fs",
1255 	"device",
1256 	"late",
1257 };
1258 
1259 static int __init ignore_unknown_bootoption(char *param, char *val,
1260 			       const char *unused, void *arg)
1261 {
1262 	return 0;
1263 }
1264 
1265 static void __init do_initcall_level(int level, char *command_line)
1266 {
1267 	initcall_entry_t *fn;
1268 
1269 	parse_args(initcall_level_names[level],
1270 		   command_line, __start___param,
1271 		   __stop___param - __start___param,
1272 		   level, level,
1273 		   NULL, ignore_unknown_bootoption);
1274 
1275 	trace_initcall_level(initcall_level_names[level]);
1276 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1277 		do_one_initcall(initcall_from_entry(fn));
1278 }
1279 
1280 static void __init do_initcalls(void)
1281 {
1282 	int level;
1283 	size_t len = strlen(saved_command_line) + 1;
1284 	char *command_line;
1285 
1286 	command_line = kzalloc(len, GFP_KERNEL);
1287 	if (!command_line)
1288 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1289 
1290 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1291 		/* Parser modifies command_line, restore it each time */
1292 		strcpy(command_line, saved_command_line);
1293 		do_initcall_level(level, command_line);
1294 	}
1295 
1296 	kfree(command_line);
1297 }
1298 
1299 /*
1300  * Ok, the machine is now initialized. None of the devices
1301  * have been touched yet, but the CPU subsystem is up and
1302  * running, and memory and process management works.
1303  *
1304  * Now we can finally start doing some real work..
1305  */
1306 static void __init do_basic_setup(void)
1307 {
1308 	cpuset_init_smp();
1309 	driver_init();
1310 	init_irq_proc();
1311 	do_ctors();
1312 	usermodehelper_enable();
1313 	do_initcalls();
1314 }
1315 
1316 static void __init do_pre_smp_initcalls(void)
1317 {
1318 	initcall_entry_t *fn;
1319 
1320 	trace_initcall_level("early");
1321 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1322 		do_one_initcall(initcall_from_entry(fn));
1323 }
1324 
1325 static int run_init_process(const char *init_filename)
1326 {
1327 	const char *const *p;
1328 
1329 	argv_init[0] = init_filename;
1330 	pr_info("Run %s as init process\n", init_filename);
1331 	pr_debug("  with arguments:\n");
1332 	for (p = argv_init; *p; p++)
1333 		pr_debug("    %s\n", *p);
1334 	pr_debug("  with environment:\n");
1335 	for (p = envp_init; *p; p++)
1336 		pr_debug("    %s\n", *p);
1337 	return kernel_execve(init_filename, argv_init, envp_init);
1338 }
1339 
1340 static int try_to_run_init_process(const char *init_filename)
1341 {
1342 	int ret;
1343 
1344 	ret = run_init_process(init_filename);
1345 
1346 	if (ret && ret != -ENOENT) {
1347 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1348 		       init_filename, ret);
1349 	}
1350 
1351 	return ret;
1352 }
1353 
1354 static noinline void __init kernel_init_freeable(void);
1355 
1356 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1357 bool rodata_enabled __ro_after_init = true;
1358 static int __init set_debug_rodata(char *str)
1359 {
1360 	return strtobool(str, &rodata_enabled);
1361 }
1362 __setup("rodata=", set_debug_rodata);
1363 #endif
1364 
1365 #ifdef CONFIG_STRICT_KERNEL_RWX
1366 static void mark_readonly(void)
1367 {
1368 	if (rodata_enabled) {
1369 		/*
1370 		 * load_module() results in W+X mappings, which are cleaned
1371 		 * up with call_rcu().  Let's make sure that queued work is
1372 		 * flushed so that we don't hit false positives looking for
1373 		 * insecure pages which are W+X.
1374 		 */
1375 		rcu_barrier();
1376 		mark_rodata_ro();
1377 		rodata_test();
1378 	} else
1379 		pr_info("Kernel memory protection disabled.\n");
1380 }
1381 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1382 static inline void mark_readonly(void)
1383 {
1384 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1385 }
1386 #else
1387 static inline void mark_readonly(void)
1388 {
1389 	pr_warn("This architecture does not have kernel memory protection.\n");
1390 }
1391 #endif
1392 
1393 void __weak free_initmem(void)
1394 {
1395 	free_initmem_default(POISON_FREE_INITMEM);
1396 }
1397 
1398 static int __ref kernel_init(void *unused)
1399 {
1400 	int ret;
1401 
1402 	kernel_init_freeable();
1403 	/* need to finish all async __init code before freeing the memory */
1404 	async_synchronize_full();
1405 	ftrace_free_init_mem();
1406 	free_initmem();
1407 	mark_readonly();
1408 
1409 	/*
1410 	 * Kernel mappings are now finalized - update the userspace page-table
1411 	 * to finalize PTI.
1412 	 */
1413 	pti_finalize();
1414 
1415 	system_state = SYSTEM_RUNNING;
1416 	numa_default_policy();
1417 
1418 	rcu_end_inkernel_boot();
1419 
1420 	do_sysctl_args();
1421 
1422 	if (ramdisk_execute_command) {
1423 		ret = run_init_process(ramdisk_execute_command);
1424 		if (!ret)
1425 			return 0;
1426 		pr_err("Failed to execute %s (error %d)\n",
1427 		       ramdisk_execute_command, ret);
1428 	}
1429 
1430 	/*
1431 	 * We try each of these until one succeeds.
1432 	 *
1433 	 * The Bourne shell can be used instead of init if we are
1434 	 * trying to recover a really broken machine.
1435 	 */
1436 	if (execute_command) {
1437 		ret = run_init_process(execute_command);
1438 		if (!ret)
1439 			return 0;
1440 		panic("Requested init %s failed (error %d).",
1441 		      execute_command, ret);
1442 	}
1443 
1444 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1445 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1446 		if (ret)
1447 			pr_err("Default init %s failed (error %d)\n",
1448 			       CONFIG_DEFAULT_INIT, ret);
1449 		else
1450 			return 0;
1451 	}
1452 
1453 	if (!try_to_run_init_process("/sbin/init") ||
1454 	    !try_to_run_init_process("/etc/init") ||
1455 	    !try_to_run_init_process("/bin/init") ||
1456 	    !try_to_run_init_process("/bin/sh"))
1457 		return 0;
1458 
1459 	panic("No working init found.  Try passing init= option to kernel. "
1460 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1461 }
1462 
1463 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1464 void __init console_on_rootfs(void)
1465 {
1466 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1467 
1468 	if (IS_ERR(file)) {
1469 		pr_err("Warning: unable to open an initial console.\n");
1470 		return;
1471 	}
1472 	init_dup(file);
1473 	init_dup(file);
1474 	init_dup(file);
1475 	fput(file);
1476 }
1477 
1478 static noinline void __init kernel_init_freeable(void)
1479 {
1480 	/*
1481 	 * Wait until kthreadd is all set-up.
1482 	 */
1483 	wait_for_completion(&kthreadd_done);
1484 
1485 	/* Now the scheduler is fully set up and can do blocking allocations */
1486 	gfp_allowed_mask = __GFP_BITS_MASK;
1487 
1488 	/*
1489 	 * init can allocate pages on any node
1490 	 */
1491 	set_mems_allowed(node_states[N_MEMORY]);
1492 
1493 	cad_pid = task_pid(current);
1494 
1495 	smp_prepare_cpus(setup_max_cpus);
1496 
1497 	workqueue_init();
1498 
1499 	init_mm_internals();
1500 
1501 	do_pre_smp_initcalls();
1502 	lockup_detector_init();
1503 
1504 	smp_init();
1505 	sched_init_smp();
1506 
1507 	padata_init();
1508 	page_alloc_init_late();
1509 	/* Initialize page ext after all struct pages are initialized. */
1510 	page_ext_init();
1511 
1512 	do_basic_setup();
1513 
1514 	console_on_rootfs();
1515 
1516 	/*
1517 	 * check if there is an early userspace init.  If yes, let it do all
1518 	 * the work
1519 	 */
1520 	if (init_eaccess(ramdisk_execute_command) != 0) {
1521 		ramdisk_execute_command = NULL;
1522 		prepare_namespace();
1523 	}
1524 
1525 	/*
1526 	 * Ok, we have completed the initial bootup, and
1527 	 * we're essentially up and running. Get rid of the
1528 	 * initmem segments and start the user-mode stuff..
1529 	 *
1530 	 * rootfs is available now, try loading the public keys
1531 	 * and default modules
1532 	 */
1533 
1534 	integrity_load_keys();
1535 }
1536