xref: /linux/init/main.c (revision 34f71a4a2de84dde52ccfcb96ce25240ea7981a8)
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/pid_namespace.h>
67 #include <linux/device.h>
68 #include <linux/kthread.h>
69 #include <linux/sched.h>
70 #include <linux/sched/init.h>
71 #include <linux/signal.h>
72 #include <linux/idr.h>
73 #include <linux/kgdb.h>
74 #include <linux/ftrace.h>
75 #include <linux/async.h>
76 #include <linux/sfi.h>
77 #include <linux/shmem_fs.h>
78 #include <linux/slab.h>
79 #include <linux/perf_event.h>
80 #include <linux/ptrace.h>
81 #include <linux/pti.h>
82 #include <linux/blkdev.h>
83 #include <linux/elevator.h>
84 #include <linux/sched/clock.h>
85 #include <linux/sched/task.h>
86 #include <linux/sched/task_stack.h>
87 #include <linux/context_tracking.h>
88 #include <linux/random.h>
89 #include <linux/list.h>
90 #include <linux/integrity.h>
91 #include <linux/proc_ns.h>
92 #include <linux/io.h>
93 #include <linux/cache.h>
94 #include <linux/rodata_test.h>
95 #include <linux/jump_label.h>
96 #include <linux/mem_encrypt.h>
97 
98 #include <asm/io.h>
99 #include <asm/bugs.h>
100 #include <asm/setup.h>
101 #include <asm/sections.h>
102 #include <asm/cacheflush.h>
103 
104 #define CREATE_TRACE_POINTS
105 #include <trace/events/initcall.h>
106 
107 static int kernel_init(void *);
108 
109 extern void init_IRQ(void);
110 extern void radix_tree_init(void);
111 
112 /*
113  * Debug helper: via this flag we know that we are in 'early bootup code'
114  * where only the boot processor is running with IRQ disabled.  This means
115  * two things - IRQ must not be enabled before the flag is cleared and some
116  * operations which are not allowed with IRQ disabled are allowed while the
117  * flag is set.
118  */
119 bool early_boot_irqs_disabled __read_mostly;
120 
121 enum system_states system_state __read_mostly;
122 EXPORT_SYMBOL(system_state);
123 
124 /*
125  * Boot command-line arguments
126  */
127 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
128 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
129 
130 extern void time_init(void);
131 /* Default late time init is NULL. archs can override this later. */
132 void (*__initdata late_time_init)(void);
133 
134 /* Untouched command line saved by arch-specific code. */
135 char __initdata boot_command_line[COMMAND_LINE_SIZE];
136 /* Untouched saved command line (eg. for /proc) */
137 char *saved_command_line;
138 /* Command line for parameter parsing */
139 static char *static_command_line;
140 /* Untouched extra command line */
141 static char *extra_command_line;
142 /* Extra init arguments */
143 static char *extra_init_args;
144 
145 static char *execute_command;
146 static char *ramdisk_execute_command;
147 
148 /*
149  * Used to generate warnings if static_key manipulation functions are used
150  * before jump_label_init is called.
151  */
152 bool static_key_initialized __read_mostly;
153 EXPORT_SYMBOL_GPL(static_key_initialized);
154 
155 /*
156  * If set, this is an indication to the drivers that reset the underlying
157  * device before going ahead with the initialization otherwise driver might
158  * rely on the BIOS and skip the reset operation.
159  *
160  * This is useful if kernel is booting in an unreliable environment.
161  * For ex. kdump situation where previous kernel has crashed, BIOS has been
162  * skipped and devices will be in unknown state.
163  */
164 unsigned int reset_devices;
165 EXPORT_SYMBOL(reset_devices);
166 
167 static int __init set_reset_devices(char *str)
168 {
169 	reset_devices = 1;
170 	return 1;
171 }
172 
173 __setup("reset_devices", set_reset_devices);
174 
175 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
176 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
177 static const char *panic_later, *panic_param;
178 
179 extern const struct obs_kernel_param __setup_start[], __setup_end[];
180 
181 static bool __init obsolete_checksetup(char *line)
182 {
183 	const struct obs_kernel_param *p;
184 	bool had_early_param = false;
185 
186 	p = __setup_start;
187 	do {
188 		int n = strlen(p->str);
189 		if (parameqn(line, p->str, n)) {
190 			if (p->early) {
191 				/* Already done in parse_early_param?
192 				 * (Needs exact match on param part).
193 				 * Keep iterating, as we can have early
194 				 * params and __setups of same names 8( */
195 				if (line[n] == '\0' || line[n] == '=')
196 					had_early_param = true;
197 			} else if (!p->setup_func) {
198 				pr_warn("Parameter %s is obsolete, ignored\n",
199 					p->str);
200 				return true;
201 			} else if (p->setup_func(line + n))
202 				return true;
203 		}
204 		p++;
205 	} while (p < __setup_end);
206 
207 	return had_early_param;
208 }
209 
210 /*
211  * This should be approx 2 Bo*oMips to start (note initial shift), and will
212  * still work even if initially too large, it will just take slightly longer
213  */
214 unsigned long loops_per_jiffy = (1<<12);
215 EXPORT_SYMBOL(loops_per_jiffy);
216 
217 static int __init debug_kernel(char *str)
218 {
219 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
220 	return 0;
221 }
222 
223 static int __init quiet_kernel(char *str)
224 {
225 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
226 	return 0;
227 }
228 
229 early_param("debug", debug_kernel);
230 early_param("quiet", quiet_kernel);
231 
232 static int __init loglevel(char *str)
233 {
234 	int newlevel;
235 
236 	/*
237 	 * Only update loglevel value when a correct setting was passed,
238 	 * to prevent blind crashes (when loglevel being set to 0) that
239 	 * are quite hard to debug
240 	 */
241 	if (get_option(&str, &newlevel)) {
242 		console_loglevel = newlevel;
243 		return 0;
244 	}
245 
246 	return -EINVAL;
247 }
248 
249 early_param("loglevel", loglevel);
250 
251 #ifdef CONFIG_BOOT_CONFIG
252 
253 char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
254 
255 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
256 
257 static int __init xbc_snprint_cmdline(char *buf, size_t size,
258 				      struct xbc_node *root)
259 {
260 	struct xbc_node *knode, *vnode;
261 	char *end = buf + size;
262 	char c = '\"';
263 	const char *val;
264 	int ret;
265 
266 	xbc_node_for_each_key_value(root, knode, val) {
267 		ret = xbc_node_compose_key_after(root, knode,
268 					xbc_namebuf, XBC_KEYLEN_MAX);
269 		if (ret < 0)
270 			return ret;
271 
272 		vnode = xbc_node_get_child(knode);
273 		ret = snprintf(buf, rest(buf, end), "%s%c", xbc_namebuf,
274 				vnode ? '=' : ' ');
275 		if (ret < 0)
276 			return ret;
277 		buf += ret;
278 		if (!vnode)
279 			continue;
280 
281 		c = '\"';
282 		xbc_array_for_each_value(vnode, val) {
283 			ret = snprintf(buf, rest(buf, end), "%c%s", c, val);
284 			if (ret < 0)
285 				return ret;
286 			buf += ret;
287 			c = ',';
288 		}
289 		if (rest(buf, end) > 2)
290 			strcpy(buf, "\" ");
291 		buf += 2;
292 	}
293 
294 	return buf - (end - size);
295 }
296 #undef rest
297 
298 /* Make an extra command line under given key word */
299 static char * __init xbc_make_cmdline(const char *key)
300 {
301 	struct xbc_node *root;
302 	char *new_cmdline;
303 	int ret, len = 0;
304 
305 	root = xbc_find_node(key);
306 	if (!root)
307 		return NULL;
308 
309 	/* Count required buffer size */
310 	len = xbc_snprint_cmdline(NULL, 0, root);
311 	if (len <= 0)
312 		return NULL;
313 
314 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
315 	if (!new_cmdline) {
316 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
317 		return NULL;
318 	}
319 
320 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
321 	if (ret < 0 || ret > len) {
322 		pr_err("Failed to print extra kernel cmdline.\n");
323 		return NULL;
324 	}
325 
326 	return new_cmdline;
327 }
328 
329 u32 boot_config_checksum(unsigned char *p, u32 size)
330 {
331 	u32 ret = 0;
332 
333 	while (size--)
334 		ret += *p++;
335 
336 	return ret;
337 }
338 
339 static void __init setup_boot_config(void)
340 {
341 	u32 size, csum;
342 	char *data, *copy;
343 	u32 *hdr;
344 
345 	if (!initrd_end)
346 		return;
347 
348 	hdr = (u32 *)(initrd_end - 8);
349 	size = hdr[0];
350 	csum = hdr[1];
351 
352 	if (size >= XBC_DATA_MAX)
353 		return;
354 
355 	data = ((void *)hdr) - size;
356 	if ((unsigned long)data < initrd_start)
357 		return;
358 
359 	if (boot_config_checksum((unsigned char *)data, size) != csum)
360 		return;
361 
362 	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
363 	if (!copy) {
364 		pr_err("Failed to allocate memory for boot config\n");
365 		return;
366 	}
367 
368 	memcpy(copy, data, size);
369 	copy[size] = '\0';
370 
371 	if (xbc_init(copy) < 0)
372 		pr_err("Failed to parse boot config\n");
373 	else {
374 		pr_info("Load boot config: %d bytes\n", size);
375 		/* keys starting with "kernel." are passed via cmdline */
376 		extra_command_line = xbc_make_cmdline("kernel");
377 		/* Also, "init." keys are init arguments */
378 		extra_init_args = xbc_make_cmdline("init");
379 	}
380 }
381 #else
382 #define setup_boot_config()	do { } while (0)
383 #endif
384 
385 /* Change NUL term back to "=", to make "param" the whole string. */
386 static int __init repair_env_string(char *param, char *val,
387 				    const char *unused, void *arg)
388 {
389 	if (val) {
390 		/* param=val or param="val"? */
391 		if (val == param+strlen(param)+1)
392 			val[-1] = '=';
393 		else if (val == param+strlen(param)+2) {
394 			val[-2] = '=';
395 			memmove(val-1, val, strlen(val)+1);
396 			val--;
397 		} else
398 			BUG();
399 	}
400 	return 0;
401 }
402 
403 /* Anything after -- gets handed straight to init. */
404 static int __init set_init_arg(char *param, char *val,
405 			       const char *unused, void *arg)
406 {
407 	unsigned int i;
408 
409 	if (panic_later)
410 		return 0;
411 
412 	repair_env_string(param, val, unused, NULL);
413 
414 	for (i = 0; argv_init[i]; i++) {
415 		if (i == MAX_INIT_ARGS) {
416 			panic_later = "init";
417 			panic_param = param;
418 			return 0;
419 		}
420 	}
421 	argv_init[i] = param;
422 	return 0;
423 }
424 
425 /*
426  * Unknown boot options get handed to init, unless they look like
427  * unused parameters (modprobe will find them in /proc/cmdline).
428  */
429 static int __init unknown_bootoption(char *param, char *val,
430 				     const char *unused, void *arg)
431 {
432 	repair_env_string(param, val, unused, NULL);
433 
434 	/* Handle obsolete-style parameters */
435 	if (obsolete_checksetup(param))
436 		return 0;
437 
438 	/* Unused module parameter. */
439 	if (strchr(param, '.') && (!val || strchr(param, '.') < val))
440 		return 0;
441 
442 	if (panic_later)
443 		return 0;
444 
445 	if (val) {
446 		/* Environment option */
447 		unsigned int i;
448 		for (i = 0; envp_init[i]; i++) {
449 			if (i == MAX_INIT_ENVS) {
450 				panic_later = "env";
451 				panic_param = param;
452 			}
453 			if (!strncmp(param, envp_init[i], val - param))
454 				break;
455 		}
456 		envp_init[i] = param;
457 	} else {
458 		/* Command line option */
459 		unsigned int i;
460 		for (i = 0; argv_init[i]; i++) {
461 			if (i == MAX_INIT_ARGS) {
462 				panic_later = "init";
463 				panic_param = param;
464 			}
465 		}
466 		argv_init[i] = param;
467 	}
468 	return 0;
469 }
470 
471 static int __init init_setup(char *str)
472 {
473 	unsigned int i;
474 
475 	execute_command = str;
476 	/*
477 	 * In case LILO is going to boot us with default command line,
478 	 * it prepends "auto" before the whole cmdline which makes
479 	 * the shell think it should execute a script with such name.
480 	 * So we ignore all arguments entered _before_ init=... [MJ]
481 	 */
482 	for (i = 1; i < MAX_INIT_ARGS; i++)
483 		argv_init[i] = NULL;
484 	return 1;
485 }
486 __setup("init=", init_setup);
487 
488 static int __init rdinit_setup(char *str)
489 {
490 	unsigned int i;
491 
492 	ramdisk_execute_command = str;
493 	/* See "auto" comment in init_setup */
494 	for (i = 1; i < MAX_INIT_ARGS; i++)
495 		argv_init[i] = NULL;
496 	return 1;
497 }
498 __setup("rdinit=", rdinit_setup);
499 
500 #ifndef CONFIG_SMP
501 static const unsigned int setup_max_cpus = NR_CPUS;
502 static inline void setup_nr_cpu_ids(void) { }
503 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
504 #endif
505 
506 /*
507  * We need to store the untouched command line for future reference.
508  * We also need to store the touched command line since the parameter
509  * parsing is performed in place, and we should allow a component to
510  * store reference of name/value for future reference.
511  */
512 static void __init setup_command_line(char *command_line)
513 {
514 	size_t len, xlen = 0, ilen = 0;
515 
516 	if (extra_command_line)
517 		xlen = strlen(extra_command_line);
518 	if (extra_init_args)
519 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
520 
521 	len = xlen + strlen(boot_command_line) + 1;
522 
523 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
524 	if (!saved_command_line)
525 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
526 
527 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
528 	if (!static_command_line)
529 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
530 
531 	if (xlen) {
532 		/*
533 		 * We have to put extra_command_line before boot command
534 		 * lines because there could be dashes (separator of init
535 		 * command line) in the command lines.
536 		 */
537 		strcpy(saved_command_line, extra_command_line);
538 		strcpy(static_command_line, extra_command_line);
539 	}
540 	strcpy(saved_command_line + xlen, boot_command_line);
541 	strcpy(static_command_line + xlen, command_line);
542 
543 	if (ilen) {
544 		/*
545 		 * Append supplemental init boot args to saved_command_line
546 		 * so that user can check what command line options passed
547 		 * to init.
548 		 */
549 		len = strlen(saved_command_line);
550 		if (!strstr(boot_command_line, " -- ")) {
551 			strcpy(saved_command_line + len, " -- ");
552 			len += 4;
553 		} else
554 			saved_command_line[len++] = ' ';
555 
556 		strcpy(saved_command_line + len, extra_init_args);
557 	}
558 }
559 
560 /*
561  * We need to finalize in a non-__init function or else race conditions
562  * between the root thread and the init thread may cause start_kernel to
563  * be reaped by free_initmem before the root thread has proceeded to
564  * cpu_idle.
565  *
566  * gcc-3.4 accidentally inlines this function, so use noinline.
567  */
568 
569 static __initdata DECLARE_COMPLETION(kthreadd_done);
570 
571 noinline void __ref rest_init(void)
572 {
573 	struct task_struct *tsk;
574 	int pid;
575 
576 	rcu_scheduler_starting();
577 	/*
578 	 * We need to spawn init first so that it obtains pid 1, however
579 	 * the init task will end up wanting to create kthreads, which, if
580 	 * we schedule it before we create kthreadd, will OOPS.
581 	 */
582 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
583 	/*
584 	 * Pin init on the boot CPU. Task migration is not properly working
585 	 * until sched_init_smp() has been run. It will set the allowed
586 	 * CPUs for init to the non isolated CPUs.
587 	 */
588 	rcu_read_lock();
589 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
590 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
591 	rcu_read_unlock();
592 
593 	numa_default_policy();
594 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
595 	rcu_read_lock();
596 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
597 	rcu_read_unlock();
598 
599 	/*
600 	 * Enable might_sleep() and smp_processor_id() checks.
601 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
602 	 * kernel_thread() would trigger might_sleep() splats. With
603 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
604 	 * already, but it's stuck on the kthreadd_done completion.
605 	 */
606 	system_state = SYSTEM_SCHEDULING;
607 
608 	complete(&kthreadd_done);
609 
610 	/*
611 	 * The boot idle thread must execute schedule()
612 	 * at least once to get things moving:
613 	 */
614 	schedule_preempt_disabled();
615 	/* Call into cpu_idle with preempt disabled */
616 	cpu_startup_entry(CPUHP_ONLINE);
617 }
618 
619 /* Check for early params. */
620 static int __init do_early_param(char *param, char *val,
621 				 const char *unused, void *arg)
622 {
623 	const struct obs_kernel_param *p;
624 
625 	for (p = __setup_start; p < __setup_end; p++) {
626 		if ((p->early && parameq(param, p->str)) ||
627 		    (strcmp(param, "console") == 0 &&
628 		     strcmp(p->str, "earlycon") == 0)
629 		) {
630 			if (p->setup_func(val) != 0)
631 				pr_warn("Malformed early option '%s'\n", param);
632 		}
633 	}
634 	/* We accept everything at this stage. */
635 	return 0;
636 }
637 
638 void __init parse_early_options(char *cmdline)
639 {
640 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
641 		   do_early_param);
642 }
643 
644 /* Arch code calls this early on, or if not, just before other parsing. */
645 void __init parse_early_param(void)
646 {
647 	static int done __initdata;
648 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
649 
650 	if (done)
651 		return;
652 
653 	/* All fall through to do_early_param. */
654 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
655 	parse_early_options(tmp_cmdline);
656 	done = 1;
657 }
658 
659 void __init __weak arch_post_acpi_subsys_init(void) { }
660 
661 void __init __weak smp_setup_processor_id(void)
662 {
663 }
664 
665 # if THREAD_SIZE >= PAGE_SIZE
666 void __init __weak thread_stack_cache_init(void)
667 {
668 }
669 #endif
670 
671 void __init __weak mem_encrypt_init(void) { }
672 
673 void __init __weak poking_init(void) { }
674 
675 void __init __weak pgtable_cache_init(void) { }
676 
677 bool initcall_debug;
678 core_param(initcall_debug, initcall_debug, bool, 0644);
679 
680 #ifdef TRACEPOINTS_ENABLED
681 static void __init initcall_debug_enable(void);
682 #else
683 static inline void initcall_debug_enable(void)
684 {
685 }
686 #endif
687 
688 /* Report memory auto-initialization states for this boot. */
689 static void __init report_meminit(void)
690 {
691 	const char *stack;
692 
693 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL))
694 		stack = "all";
695 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
696 		stack = "byref_all";
697 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
698 		stack = "byref";
699 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
700 		stack = "__user";
701 	else
702 		stack = "off";
703 
704 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
705 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
706 		want_init_on_free() ? "on" : "off");
707 	if (want_init_on_free())
708 		pr_info("mem auto-init: clearing system memory may take some time...\n");
709 }
710 
711 /*
712  * Set up kernel memory allocators
713  */
714 static void __init mm_init(void)
715 {
716 	/*
717 	 * page_ext requires contiguous pages,
718 	 * bigger than MAX_ORDER unless SPARSEMEM.
719 	 */
720 	page_ext_init_flatmem();
721 	report_meminit();
722 	mem_init();
723 	kmem_cache_init();
724 	kmemleak_init();
725 	pgtable_init();
726 	debug_objects_mem_init();
727 	vmalloc_init();
728 	ioremap_huge_init();
729 	/* Should be run before the first non-init thread is created */
730 	init_espfix_bsp();
731 	/* Should be run after espfix64 is set up. */
732 	pti_init();
733 }
734 
735 void __init __weak arch_call_rest_init(void)
736 {
737 	rest_init();
738 }
739 
740 asmlinkage __visible void __init start_kernel(void)
741 {
742 	char *command_line;
743 	char *after_dashes;
744 
745 	set_task_stack_end_magic(&init_task);
746 	smp_setup_processor_id();
747 	debug_objects_early_init();
748 
749 	cgroup_init_early();
750 
751 	local_irq_disable();
752 	early_boot_irqs_disabled = true;
753 
754 	/*
755 	 * Interrupts are still disabled. Do necessary setups, then
756 	 * enable them.
757 	 */
758 	boot_cpu_init();
759 	page_address_init();
760 	pr_notice("%s", linux_banner);
761 	early_security_init();
762 	setup_arch(&command_line);
763 	setup_boot_config();
764 	setup_command_line(command_line);
765 	setup_nr_cpu_ids();
766 	setup_per_cpu_areas();
767 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
768 	boot_cpu_hotplug_init();
769 
770 	build_all_zonelists(NULL);
771 	page_alloc_init();
772 
773 	pr_notice("Kernel command line: %s\n", saved_command_line);
774 	/* parameters may set static keys */
775 	jump_label_init();
776 	parse_early_param();
777 	after_dashes = parse_args("Booting kernel",
778 				  static_command_line, __start___param,
779 				  __stop___param - __start___param,
780 				  -1, -1, NULL, &unknown_bootoption);
781 	if (!IS_ERR_OR_NULL(after_dashes))
782 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
783 			   NULL, set_init_arg);
784 	if (extra_init_args)
785 		parse_args("Setting extra init args", extra_init_args,
786 			   NULL, 0, -1, -1, NULL, set_init_arg);
787 
788 	/*
789 	 * These use large bootmem allocations and must precede
790 	 * kmem_cache_init()
791 	 */
792 	setup_log_buf(0);
793 	vfs_caches_init_early();
794 	sort_main_extable();
795 	trap_init();
796 	mm_init();
797 
798 	ftrace_init();
799 
800 	/* trace_printk can be enabled here */
801 	early_trace_init();
802 
803 	/*
804 	 * Set up the scheduler prior starting any interrupts (such as the
805 	 * timer interrupt). Full topology setup happens at smp_init()
806 	 * time - but meanwhile we still have a functioning scheduler.
807 	 */
808 	sched_init();
809 	/*
810 	 * Disable preemption - early bootup scheduling is extremely
811 	 * fragile until we cpu_idle() for the first time.
812 	 */
813 	preempt_disable();
814 	if (WARN(!irqs_disabled(),
815 		 "Interrupts were enabled *very* early, fixing it\n"))
816 		local_irq_disable();
817 	radix_tree_init();
818 
819 	/*
820 	 * Set up housekeeping before setting up workqueues to allow the unbound
821 	 * workqueue to take non-housekeeping into account.
822 	 */
823 	housekeeping_init();
824 
825 	/*
826 	 * Allow workqueue creation and work item queueing/cancelling
827 	 * early.  Work item execution depends on kthreads and starts after
828 	 * workqueue_init().
829 	 */
830 	workqueue_init_early();
831 
832 	rcu_init();
833 
834 	/* Trace events are available after this */
835 	trace_init();
836 
837 	if (initcall_debug)
838 		initcall_debug_enable();
839 
840 	context_tracking_init();
841 	/* init some links before init_ISA_irqs() */
842 	early_irq_init();
843 	init_IRQ();
844 	tick_init();
845 	rcu_init_nohz();
846 	init_timers();
847 	hrtimers_init();
848 	softirq_init();
849 	timekeeping_init();
850 
851 	/*
852 	 * For best initial stack canary entropy, prepare it after:
853 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
854 	 * - timekeeping_init() for ktime entropy used in rand_initialize()
855 	 * - rand_initialize() to get any arch-specific entropy like RDRAND
856 	 * - add_latent_entropy() to get any latent entropy
857 	 * - adding command line entropy
858 	 */
859 	rand_initialize();
860 	add_latent_entropy();
861 	add_device_randomness(command_line, strlen(command_line));
862 	boot_init_stack_canary();
863 
864 	time_init();
865 	printk_safe_init();
866 	perf_event_init();
867 	profile_init();
868 	call_function_init();
869 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
870 
871 	early_boot_irqs_disabled = false;
872 	local_irq_enable();
873 
874 	kmem_cache_init_late();
875 
876 	/*
877 	 * HACK ALERT! This is early. We're enabling the console before
878 	 * we've done PCI setups etc, and console_init() must be aware of
879 	 * this. But we do want output early, in case something goes wrong.
880 	 */
881 	console_init();
882 	if (panic_later)
883 		panic("Too many boot %s vars at `%s'", panic_later,
884 		      panic_param);
885 
886 	lockdep_init();
887 
888 	/*
889 	 * Need to run this when irqs are enabled, because it wants
890 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
891 	 * too:
892 	 */
893 	locking_selftest();
894 
895 	/*
896 	 * This needs to be called before any devices perform DMA
897 	 * operations that might use the SWIOTLB bounce buffers. It will
898 	 * mark the bounce buffers as decrypted so that their usage will
899 	 * not cause "plain-text" data to be decrypted when accessed.
900 	 */
901 	mem_encrypt_init();
902 
903 #ifdef CONFIG_BLK_DEV_INITRD
904 	if (initrd_start && !initrd_below_start_ok &&
905 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
906 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
907 		    page_to_pfn(virt_to_page((void *)initrd_start)),
908 		    min_low_pfn);
909 		initrd_start = 0;
910 	}
911 #endif
912 	setup_per_cpu_pageset();
913 	numa_policy_init();
914 	acpi_early_init();
915 	if (late_time_init)
916 		late_time_init();
917 	sched_clock_init();
918 	calibrate_delay();
919 	pid_idr_init();
920 	anon_vma_init();
921 #ifdef CONFIG_X86
922 	if (efi_enabled(EFI_RUNTIME_SERVICES))
923 		efi_enter_virtual_mode();
924 #endif
925 	thread_stack_cache_init();
926 	cred_init();
927 	fork_init();
928 	proc_caches_init();
929 	uts_ns_init();
930 	buffer_init();
931 	key_init();
932 	security_init();
933 	dbg_late_init();
934 	vfs_caches_init();
935 	pagecache_init();
936 	signals_init();
937 	seq_file_init();
938 	proc_root_init();
939 	nsfs_init();
940 	cpuset_init();
941 	cgroup_init();
942 	taskstats_init_early();
943 	delayacct_init();
944 
945 	poking_init();
946 	check_bugs();
947 
948 	acpi_subsystem_init();
949 	arch_post_acpi_subsys_init();
950 	sfi_init_late();
951 
952 	/* Do the rest non-__init'ed, we're now alive */
953 	arch_call_rest_init();
954 }
955 
956 /* Call all constructor functions linked into the kernel. */
957 static void __init do_ctors(void)
958 {
959 #ifdef CONFIG_CONSTRUCTORS
960 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
961 
962 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
963 		(*fn)();
964 #endif
965 }
966 
967 #ifdef CONFIG_KALLSYMS
968 struct blacklist_entry {
969 	struct list_head next;
970 	char *buf;
971 };
972 
973 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
974 
975 static int __init initcall_blacklist(char *str)
976 {
977 	char *str_entry;
978 	struct blacklist_entry *entry;
979 
980 	/* str argument is a comma-separated list of functions */
981 	do {
982 		str_entry = strsep(&str, ",");
983 		if (str_entry) {
984 			pr_debug("blacklisting initcall %s\n", str_entry);
985 			entry = memblock_alloc(sizeof(*entry),
986 					       SMP_CACHE_BYTES);
987 			if (!entry)
988 				panic("%s: Failed to allocate %zu bytes\n",
989 				      __func__, sizeof(*entry));
990 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
991 						    SMP_CACHE_BYTES);
992 			if (!entry->buf)
993 				panic("%s: Failed to allocate %zu bytes\n",
994 				      __func__, strlen(str_entry) + 1);
995 			strcpy(entry->buf, str_entry);
996 			list_add(&entry->next, &blacklisted_initcalls);
997 		}
998 	} while (str_entry);
999 
1000 	return 0;
1001 }
1002 
1003 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1004 {
1005 	struct blacklist_entry *entry;
1006 	char fn_name[KSYM_SYMBOL_LEN];
1007 	unsigned long addr;
1008 
1009 	if (list_empty(&blacklisted_initcalls))
1010 		return false;
1011 
1012 	addr = (unsigned long) dereference_function_descriptor(fn);
1013 	sprint_symbol_no_offset(fn_name, addr);
1014 
1015 	/*
1016 	 * fn will be "function_name [module_name]" where [module_name] is not
1017 	 * displayed for built-in init functions.  Strip off the [module_name].
1018 	 */
1019 	strreplace(fn_name, ' ', '\0');
1020 
1021 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1022 		if (!strcmp(fn_name, entry->buf)) {
1023 			pr_debug("initcall %s blacklisted\n", fn_name);
1024 			return true;
1025 		}
1026 	}
1027 
1028 	return false;
1029 }
1030 #else
1031 static int __init initcall_blacklist(char *str)
1032 {
1033 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1034 	return 0;
1035 }
1036 
1037 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1038 {
1039 	return false;
1040 }
1041 #endif
1042 __setup("initcall_blacklist=", initcall_blacklist);
1043 
1044 static __init_or_module void
1045 trace_initcall_start_cb(void *data, initcall_t fn)
1046 {
1047 	ktime_t *calltime = (ktime_t *)data;
1048 
1049 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1050 	*calltime = ktime_get();
1051 }
1052 
1053 static __init_or_module void
1054 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1055 {
1056 	ktime_t *calltime = (ktime_t *)data;
1057 	ktime_t delta, rettime;
1058 	unsigned long long duration;
1059 
1060 	rettime = ktime_get();
1061 	delta = ktime_sub(rettime, *calltime);
1062 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1063 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1064 		 fn, ret, duration);
1065 }
1066 
1067 static ktime_t initcall_calltime;
1068 
1069 #ifdef TRACEPOINTS_ENABLED
1070 static void __init initcall_debug_enable(void)
1071 {
1072 	int ret;
1073 
1074 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1075 					    &initcall_calltime);
1076 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1077 					      &initcall_calltime);
1078 	WARN(ret, "Failed to register initcall tracepoints\n");
1079 }
1080 # define do_trace_initcall_start	trace_initcall_start
1081 # define do_trace_initcall_finish	trace_initcall_finish
1082 #else
1083 static inline void do_trace_initcall_start(initcall_t fn)
1084 {
1085 	if (!initcall_debug)
1086 		return;
1087 	trace_initcall_start_cb(&initcall_calltime, fn);
1088 }
1089 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1090 {
1091 	if (!initcall_debug)
1092 		return;
1093 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1094 }
1095 #endif /* !TRACEPOINTS_ENABLED */
1096 
1097 int __init_or_module do_one_initcall(initcall_t fn)
1098 {
1099 	int count = preempt_count();
1100 	char msgbuf[64];
1101 	int ret;
1102 
1103 	if (initcall_blacklisted(fn))
1104 		return -EPERM;
1105 
1106 	do_trace_initcall_start(fn);
1107 	ret = fn();
1108 	do_trace_initcall_finish(fn, ret);
1109 
1110 	msgbuf[0] = 0;
1111 
1112 	if (preempt_count() != count) {
1113 		sprintf(msgbuf, "preemption imbalance ");
1114 		preempt_count_set(count);
1115 	}
1116 	if (irqs_disabled()) {
1117 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1118 		local_irq_enable();
1119 	}
1120 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1121 
1122 	add_latent_entropy();
1123 	return ret;
1124 }
1125 
1126 
1127 extern initcall_entry_t __initcall_start[];
1128 extern initcall_entry_t __initcall0_start[];
1129 extern initcall_entry_t __initcall1_start[];
1130 extern initcall_entry_t __initcall2_start[];
1131 extern initcall_entry_t __initcall3_start[];
1132 extern initcall_entry_t __initcall4_start[];
1133 extern initcall_entry_t __initcall5_start[];
1134 extern initcall_entry_t __initcall6_start[];
1135 extern initcall_entry_t __initcall7_start[];
1136 extern initcall_entry_t __initcall_end[];
1137 
1138 static initcall_entry_t *initcall_levels[] __initdata = {
1139 	__initcall0_start,
1140 	__initcall1_start,
1141 	__initcall2_start,
1142 	__initcall3_start,
1143 	__initcall4_start,
1144 	__initcall5_start,
1145 	__initcall6_start,
1146 	__initcall7_start,
1147 	__initcall_end,
1148 };
1149 
1150 /* Keep these in sync with initcalls in include/linux/init.h */
1151 static const char *initcall_level_names[] __initdata = {
1152 	"pure",
1153 	"core",
1154 	"postcore",
1155 	"arch",
1156 	"subsys",
1157 	"fs",
1158 	"device",
1159 	"late",
1160 };
1161 
1162 static void __init do_initcall_level(int level, char *command_line)
1163 {
1164 	initcall_entry_t *fn;
1165 
1166 	parse_args(initcall_level_names[level],
1167 		   command_line, __start___param,
1168 		   __stop___param - __start___param,
1169 		   level, level,
1170 		   NULL, &repair_env_string);
1171 
1172 	trace_initcall_level(initcall_level_names[level]);
1173 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1174 		do_one_initcall(initcall_from_entry(fn));
1175 }
1176 
1177 static void __init do_initcalls(void)
1178 {
1179 	int level;
1180 	size_t len = strlen(saved_command_line) + 1;
1181 	char *command_line;
1182 
1183 	command_line = kzalloc(len, GFP_KERNEL);
1184 	if (!command_line)
1185 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1186 
1187 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1188 		/* Parser modifies command_line, restore it each time */
1189 		strcpy(command_line, saved_command_line);
1190 		do_initcall_level(level, command_line);
1191 	}
1192 
1193 	kfree(command_line);
1194 }
1195 
1196 /*
1197  * Ok, the machine is now initialized. None of the devices
1198  * have been touched yet, but the CPU subsystem is up and
1199  * running, and memory and process management works.
1200  *
1201  * Now we can finally start doing some real work..
1202  */
1203 static void __init do_basic_setup(void)
1204 {
1205 	cpuset_init_smp();
1206 	driver_init();
1207 	init_irq_proc();
1208 	do_ctors();
1209 	usermodehelper_enable();
1210 	do_initcalls();
1211 }
1212 
1213 static void __init do_pre_smp_initcalls(void)
1214 {
1215 	initcall_entry_t *fn;
1216 
1217 	trace_initcall_level("early");
1218 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1219 		do_one_initcall(initcall_from_entry(fn));
1220 }
1221 
1222 static int run_init_process(const char *init_filename)
1223 {
1224 	argv_init[0] = init_filename;
1225 	pr_info("Run %s as init process\n", init_filename);
1226 	return do_execve(getname_kernel(init_filename),
1227 		(const char __user *const __user *)argv_init,
1228 		(const char __user *const __user *)envp_init);
1229 }
1230 
1231 static int try_to_run_init_process(const char *init_filename)
1232 {
1233 	int ret;
1234 
1235 	ret = run_init_process(init_filename);
1236 
1237 	if (ret && ret != -ENOENT) {
1238 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1239 		       init_filename, ret);
1240 	}
1241 
1242 	return ret;
1243 }
1244 
1245 static noinline void __init kernel_init_freeable(void);
1246 
1247 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1248 bool rodata_enabled __ro_after_init = true;
1249 static int __init set_debug_rodata(char *str)
1250 {
1251 	return strtobool(str, &rodata_enabled);
1252 }
1253 __setup("rodata=", set_debug_rodata);
1254 #endif
1255 
1256 #ifdef CONFIG_STRICT_KERNEL_RWX
1257 static void mark_readonly(void)
1258 {
1259 	if (rodata_enabled) {
1260 		/*
1261 		 * load_module() results in W+X mappings, which are cleaned
1262 		 * up with call_rcu().  Let's make sure that queued work is
1263 		 * flushed so that we don't hit false positives looking for
1264 		 * insecure pages which are W+X.
1265 		 */
1266 		rcu_barrier();
1267 		mark_rodata_ro();
1268 		rodata_test();
1269 	} else
1270 		pr_info("Kernel memory protection disabled.\n");
1271 }
1272 #else
1273 static inline void mark_readonly(void)
1274 {
1275 	pr_warn("This architecture does not have kernel memory protection.\n");
1276 }
1277 #endif
1278 
1279 void __weak free_initmem(void)
1280 {
1281 	free_initmem_default(POISON_FREE_INITMEM);
1282 }
1283 
1284 static int __ref kernel_init(void *unused)
1285 {
1286 	int ret;
1287 
1288 	kernel_init_freeable();
1289 	/* need to finish all async __init code before freeing the memory */
1290 	async_synchronize_full();
1291 	ftrace_free_init_mem();
1292 	free_initmem();
1293 	mark_readonly();
1294 
1295 	/*
1296 	 * Kernel mappings are now finalized - update the userspace page-table
1297 	 * to finalize PTI.
1298 	 */
1299 	pti_finalize();
1300 
1301 	system_state = SYSTEM_RUNNING;
1302 	numa_default_policy();
1303 
1304 	rcu_end_inkernel_boot();
1305 
1306 	if (ramdisk_execute_command) {
1307 		ret = run_init_process(ramdisk_execute_command);
1308 		if (!ret)
1309 			return 0;
1310 		pr_err("Failed to execute %s (error %d)\n",
1311 		       ramdisk_execute_command, ret);
1312 	}
1313 
1314 	/*
1315 	 * We try each of these until one succeeds.
1316 	 *
1317 	 * The Bourne shell can be used instead of init if we are
1318 	 * trying to recover a really broken machine.
1319 	 */
1320 	if (execute_command) {
1321 		ret = run_init_process(execute_command);
1322 		if (!ret)
1323 			return 0;
1324 		panic("Requested init %s failed (error %d).",
1325 		      execute_command, ret);
1326 	}
1327 	if (!try_to_run_init_process("/sbin/init") ||
1328 	    !try_to_run_init_process("/etc/init") ||
1329 	    !try_to_run_init_process("/bin/init") ||
1330 	    !try_to_run_init_process("/bin/sh"))
1331 		return 0;
1332 
1333 	panic("No working init found.  Try passing init= option to kernel. "
1334 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1335 }
1336 
1337 void console_on_rootfs(void)
1338 {
1339 	/* Open the /dev/console as stdin, this should never fail */
1340 	if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1341 		pr_err("Warning: unable to open an initial console.\n");
1342 
1343 	/* create stdout/stderr */
1344 	(void) ksys_dup(0);
1345 	(void) ksys_dup(0);
1346 }
1347 
1348 static noinline void __init kernel_init_freeable(void)
1349 {
1350 	/*
1351 	 * Wait until kthreadd is all set-up.
1352 	 */
1353 	wait_for_completion(&kthreadd_done);
1354 
1355 	/* Now the scheduler is fully set up and can do blocking allocations */
1356 	gfp_allowed_mask = __GFP_BITS_MASK;
1357 
1358 	/*
1359 	 * init can allocate pages on any node
1360 	 */
1361 	set_mems_allowed(node_states[N_MEMORY]);
1362 
1363 	cad_pid = task_pid(current);
1364 
1365 	smp_prepare_cpus(setup_max_cpus);
1366 
1367 	workqueue_init();
1368 
1369 	init_mm_internals();
1370 
1371 	do_pre_smp_initcalls();
1372 	lockup_detector_init();
1373 
1374 	smp_init();
1375 	sched_init_smp();
1376 
1377 	page_alloc_init_late();
1378 	/* Initialize page ext after all struct pages are initialized. */
1379 	page_ext_init();
1380 
1381 	do_basic_setup();
1382 
1383 	console_on_rootfs();
1384 
1385 	/*
1386 	 * check if there is an early userspace init.  If yes, let it do all
1387 	 * the work
1388 	 */
1389 
1390 	if (!ramdisk_execute_command)
1391 		ramdisk_execute_command = "/init";
1392 
1393 	if (ksys_access((const char __user *)
1394 			ramdisk_execute_command, 0) != 0) {
1395 		ramdisk_execute_command = NULL;
1396 		prepare_namespace();
1397 	}
1398 
1399 	/*
1400 	 * Ok, we have completed the initial bootup, and
1401 	 * we're essentially up and running. Get rid of the
1402 	 * initmem segments and start the user-mode stuff..
1403 	 *
1404 	 * rootfs is available now, try loading the public keys
1405 	 * and default modules
1406 	 */
1407 
1408 	integrity_load_keys();
1409 }
1410