xref: /kernel/linux/linux-5.10/init/main.c (revision 8c2ecf20)
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 *  linux/init/main.c
4 *
5 *  Copyright (C) 1991, 1992  Linus Torvalds
6 *
7 *  GK 2/5/95  -  Changed to support mounting root fs via NFS
8 *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
11 */
12
13#define DEBUG		/* Enable initcall_debug */
14
15#include <linux/types.h>
16#include <linux/extable.h>
17#include <linux/module.h>
18#include <linux/proc_fs.h>
19#include <linux/binfmts.h>
20#include <linux/kernel.h>
21#include <linux/syscalls.h>
22#include <linux/stackprotector.h>
23#include <linux/string.h>
24#include <linux/ctype.h>
25#include <linux/delay.h>
26#include <linux/ioport.h>
27#include <linux/init.h>
28#include <linux/initrd.h>
29#include <linux/memblock.h>
30#include <linux/acpi.h>
31#include <linux/bootconfig.h>
32#include <linux/console.h>
33#include <linux/nmi.h>
34#include <linux/percpu.h>
35#include <linux/kmod.h>
36#include <linux/kprobes.h>
37#include <linux/vmalloc.h>
38#include <linux/kernel_stat.h>
39#include <linux/start_kernel.h>
40#include <linux/security.h>
41#include <linux/smp.h>
42#include <linux/profile.h>
43#include <linux/rcupdate.h>
44#include <linux/moduleparam.h>
45#include <linux/kallsyms.h>
46#include <linux/writeback.h>
47#include <linux/cpu.h>
48#include <linux/cpuset.h>
49#include <linux/cgroup.h>
50#include <linux/efi.h>
51#include <linux/tick.h>
52#include <linux/sched/isolation.h>
53#include <linux/interrupt.h>
54#include <linux/taskstats_kern.h>
55#include <linux/delayacct.h>
56#include <linux/unistd.h>
57#include <linux/utsname.h>
58#include <linux/rmap.h>
59#include <linux/mempolicy.h>
60#include <linux/key.h>
61#include <linux/buffer_head.h>
62#include <linux/page_ext.h>
63#include <linux/debug_locks.h>
64#include <linux/debugobjects.h>
65#include <linux/lockdep.h>
66#include <linux/kmemleak.h>
67#include <linux/padata.h>
68#include <linux/pid_namespace.h>
69#include <linux/device/driver.h>
70#include <linux/kthread.h>
71#include <linux/sched.h>
72#include <linux/sched/init.h>
73#include <linux/signal.h>
74#include <linux/idr.h>
75#include <linux/kgdb.h>
76#include <linux/ftrace.h>
77#include <linux/async.h>
78#include <linux/sfi.h>
79#include <linux/shmem_fs.h>
80#include <linux/slab.h>
81#include <linux/perf_event.h>
82#include <linux/ptrace.h>
83#include <linux/pti.h>
84#include <linux/blkdev.h>
85#include <linux/elevator.h>
86#include <linux/sched/clock.h>
87#include <linux/sched/task.h>
88#include <linux/sched/task_stack.h>
89#include <linux/context_tracking.h>
90#include <linux/random.h>
91#include <linux/list.h>
92#include <linux/integrity.h>
93#include <linux/proc_ns.h>
94#include <linux/io.h>
95#include <linux/cache.h>
96#include <linux/rodata_test.h>
97#include <linux/jump_label.h>
98#include <linux/kcsan.h>
99#include <linux/init_syscalls.h>
100#ifdef CONFIG_RECLAIM_ACCT
101#include <linux/reclaim_acct.h>
102#endif
103
104#include <asm/io.h>
105#include <asm/setup.h>
106#include <asm/sections.h>
107#include <asm/cacheflush.h>
108
109#define CREATE_TRACE_POINTS
110#include <trace/events/initcall.h>
111
112#include <kunit/test.h>
113
114static int kernel_init(void *);
115
116extern void init_IRQ(void);
117extern void radix_tree_init(void);
118
119/*
120 * Debug helper: via this flag we know that we are in 'early bootup code'
121 * where only the boot processor is running with IRQ disabled.  This means
122 * two things - IRQ must not be enabled before the flag is cleared and some
123 * operations which are not allowed with IRQ disabled are allowed while the
124 * flag is set.
125 */
126bool early_boot_irqs_disabled __read_mostly;
127
128enum system_states system_state __read_mostly;
129EXPORT_SYMBOL(system_state);
130
131/*
132 * Boot command-line arguments
133 */
134#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
135#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
136
137extern void time_init(void);
138/* Default late time init is NULL. archs can override this later. */
139void (*__initdata late_time_init)(void);
140
141/* Untouched command line saved by arch-specific code. */
142char __initdata boot_command_line[COMMAND_LINE_SIZE];
143/* Untouched saved command line (eg. for /proc) */
144char *saved_command_line;
145/* Command line for parameter parsing */
146static char *static_command_line;
147/* Untouched extra command line */
148static char *extra_command_line;
149/* Extra init arguments */
150static char *extra_init_args;
151
152#ifdef CONFIG_BOOT_CONFIG
153/* Is bootconfig on command line? */
154static bool bootconfig_found;
155static bool initargs_found;
156#else
157# define bootconfig_found false
158# define initargs_found false
159#endif
160
161static char *execute_command;
162static char *ramdisk_execute_command = "/init";
163
164/*
165 * Used to generate warnings if static_key manipulation functions are used
166 * before jump_label_init is called.
167 */
168bool static_key_initialized __read_mostly;
169EXPORT_SYMBOL_GPL(static_key_initialized);
170
171/*
172 * If set, this is an indication to the drivers that reset the underlying
173 * device before going ahead with the initialization otherwise driver might
174 * rely on the BIOS and skip the reset operation.
175 *
176 * This is useful if kernel is booting in an unreliable environment.
177 * For ex. kdump situation where previous kernel has crashed, BIOS has been
178 * skipped and devices will be in unknown state.
179 */
180unsigned int reset_devices;
181EXPORT_SYMBOL(reset_devices);
182
183static int __init set_reset_devices(char *str)
184{
185	reset_devices = 1;
186	return 1;
187}
188
189__setup("reset_devices", set_reset_devices);
190
191static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
192const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
193static const char *panic_later, *panic_param;
194
195extern const struct obs_kernel_param __setup_start[], __setup_end[];
196
197static bool __init obsolete_checksetup(char *line)
198{
199	const struct obs_kernel_param *p;
200	bool had_early_param = false;
201
202	p = __setup_start;
203	do {
204		int n = strlen(p->str);
205		if (parameqn(line, p->str, n)) {
206			if (p->early) {
207				/* Already done in parse_early_param?
208				 * (Needs exact match on param part).
209				 * Keep iterating, as we can have early
210				 * params and __setups of same names 8( */
211				if (line[n] == '\0' || line[n] == '=')
212					had_early_param = true;
213			} else if (!p->setup_func) {
214				pr_warn("Parameter %s is obsolete, ignored\n",
215					p->str);
216				return true;
217			} else if (p->setup_func(line + n))
218				return true;
219		}
220		p++;
221	} while (p < __setup_end);
222
223	return had_early_param;
224}
225
226/*
227 * This should be approx 2 Bo*oMips to start (note initial shift), and will
228 * still work even if initially too large, it will just take slightly longer
229 */
230unsigned long loops_per_jiffy = (1<<12);
231EXPORT_SYMBOL(loops_per_jiffy);
232
233static int __init debug_kernel(char *str)
234{
235	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
236	return 0;
237}
238
239static int __init quiet_kernel(char *str)
240{
241	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
242	return 0;
243}
244
245early_param("debug", debug_kernel);
246early_param("quiet", quiet_kernel);
247
248static int __init loglevel(char *str)
249{
250	int newlevel;
251
252	/*
253	 * Only update loglevel value when a correct setting was passed,
254	 * to prevent blind crashes (when loglevel being set to 0) that
255	 * are quite hard to debug
256	 */
257	if (get_option(&str, &newlevel)) {
258		console_loglevel = newlevel;
259		return 0;
260	}
261
262	return -EINVAL;
263}
264
265early_param("loglevel", loglevel);
266
267#ifdef CONFIG_BLK_DEV_INITRD
268static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
269{
270	u32 size, csum;
271	char *data;
272	u32 *hdr;
273	int i;
274
275	if (!initrd_end)
276		return NULL;
277
278	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
279	/*
280	 * Since Grub may align the size of initrd to 4, we must
281	 * check the preceding 3 bytes as well.
282	 */
283	for (i = 0; i < 4; i++) {
284		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
285			goto found;
286		data--;
287	}
288	return NULL;
289
290found:
291	hdr = (u32 *)(data - 8);
292	size = le32_to_cpu(hdr[0]);
293	csum = le32_to_cpu(hdr[1]);
294
295	data = ((void *)hdr) - size;
296	if ((unsigned long)data < initrd_start) {
297		pr_err("bootconfig size %d is greater than initrd size %ld\n",
298			size, initrd_end - initrd_start);
299		return NULL;
300	}
301
302	/* Remove bootconfig from initramfs/initrd */
303	initrd_end = (unsigned long)data;
304	if (_size)
305		*_size = size;
306	if (_csum)
307		*_csum = csum;
308
309	return data;
310}
311#else
312static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
313{
314	return NULL;
315}
316#endif
317
318#ifdef CONFIG_BOOT_CONFIG
319
320static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
321
322#define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
323
324static int __init xbc_snprint_cmdline(char *buf, size_t size,
325				      struct xbc_node *root)
326{
327	struct xbc_node *knode, *vnode;
328	char *end = buf + size;
329	const char *val;
330	int ret;
331
332	xbc_node_for_each_key_value(root, knode, val) {
333		ret = xbc_node_compose_key_after(root, knode,
334					xbc_namebuf, XBC_KEYLEN_MAX);
335		if (ret < 0)
336			return ret;
337
338		vnode = xbc_node_get_child(knode);
339		if (!vnode) {
340			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
341			if (ret < 0)
342				return ret;
343			buf += ret;
344			continue;
345		}
346		xbc_array_for_each_value(vnode, val) {
347			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
348				       xbc_namebuf, val);
349			if (ret < 0)
350				return ret;
351			buf += ret;
352		}
353	}
354
355	return buf - (end - size);
356}
357#undef rest
358
359/* Make an extra command line under given key word */
360static char * __init xbc_make_cmdline(const char *key)
361{
362	struct xbc_node *root;
363	char *new_cmdline;
364	int ret, len = 0;
365
366	root = xbc_find_node(key);
367	if (!root)
368		return NULL;
369
370	/* Count required buffer size */
371	len = xbc_snprint_cmdline(NULL, 0, root);
372	if (len <= 0)
373		return NULL;
374
375	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
376	if (!new_cmdline) {
377		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
378		return NULL;
379	}
380
381	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
382	if (ret < 0 || ret > len) {
383		pr_err("Failed to print extra kernel cmdline.\n");
384		memblock_free(__pa(new_cmdline), len + 1);
385		return NULL;
386	}
387
388	return new_cmdline;
389}
390
391static u32 boot_config_checksum(unsigned char *p, u32 size)
392{
393	u32 ret = 0;
394
395	while (size--)
396		ret += *p++;
397
398	return ret;
399}
400
401static int __init bootconfig_params(char *param, char *val,
402				    const char *unused, void *arg)
403{
404	if (strcmp(param, "bootconfig") == 0) {
405		bootconfig_found = true;
406	}
407	return 0;
408}
409
410static void __init setup_boot_config(const char *cmdline)
411{
412	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
413	const char *msg;
414	int pos;
415	u32 size, csum;
416	char *data, *copy, *err;
417	int ret;
418
419	/* Cut out the bootconfig data even if we have no bootconfig option */
420	data = get_boot_config_from_initrd(&size, &csum);
421
422	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
423	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
424			 bootconfig_params);
425
426	if (IS_ERR(err) || !bootconfig_found)
427		return;
428
429	/* parse_args() stops at '--' and returns an address */
430	if (err)
431		initargs_found = true;
432
433	if (!data) {
434		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
435		return;
436	}
437
438	if (size >= XBC_DATA_MAX) {
439		pr_err("bootconfig size %d greater than max size %d\n",
440			size, XBC_DATA_MAX);
441		return;
442	}
443
444	if (boot_config_checksum((unsigned char *)data, size) != csum) {
445		pr_err("bootconfig checksum failed\n");
446		return;
447	}
448
449	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
450	if (!copy) {
451		pr_err("Failed to allocate memory for bootconfig\n");
452		return;
453	}
454
455	memcpy(copy, data, size);
456	copy[size] = '\0';
457
458	ret = xbc_init(copy, &msg, &pos);
459	if (ret < 0) {
460		if (pos < 0)
461			pr_err("Failed to init bootconfig: %s.\n", msg);
462		else
463			pr_err("Failed to parse bootconfig: %s at %d.\n",
464				msg, pos);
465	} else {
466		pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
467		/* keys starting with "kernel." are passed via cmdline */
468		extra_command_line = xbc_make_cmdline("kernel");
469		/* Also, "init." keys are init arguments */
470		extra_init_args = xbc_make_cmdline("init");
471	}
472	return;
473}
474
475#else
476
477static void __init setup_boot_config(const char *cmdline)
478{
479	/* Remove bootconfig data from initrd */
480	get_boot_config_from_initrd(NULL, NULL);
481}
482
483static int __init warn_bootconfig(char *str)
484{
485	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
486	return 0;
487}
488early_param("bootconfig", warn_bootconfig);
489
490#endif
491
492/* Change NUL term back to "=", to make "param" the whole string. */
493static 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. */
508static 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 */
533static 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
577static 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
594static 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
607static const unsigned int setup_max_cpus = NR_CPUS;
608static inline void setup_nr_cpu_ids(void) { }
609static 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 */
618static 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	len = xlen + strlen(command_line) + 1;
634
635	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
636	if (!static_command_line)
637		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
638
639	if (xlen) {
640		/*
641		 * We have to put extra_command_line before boot command
642		 * lines because there could be dashes (separator of init
643		 * command line) in the command lines.
644		 */
645		strcpy(saved_command_line, extra_command_line);
646		strcpy(static_command_line, extra_command_line);
647	}
648	strcpy(saved_command_line + xlen, boot_command_line);
649	strcpy(static_command_line + xlen, command_line);
650
651	if (ilen) {
652		/*
653		 * Append supplemental init boot args to saved_command_line
654		 * so that user can check what command line options passed
655		 * to init.
656		 */
657		len = strlen(saved_command_line);
658		if (initargs_found) {
659			saved_command_line[len++] = ' ';
660		} else {
661			strcpy(saved_command_line + len, " -- ");
662			len += 4;
663		}
664
665		strcpy(saved_command_line + len, extra_init_args);
666	}
667}
668
669/*
670 * We need to finalize in a non-__init function or else race conditions
671 * between the root thread and the init thread may cause start_kernel to
672 * be reaped by free_initmem before the root thread has proceeded to
673 * cpu_idle.
674 *
675 * gcc-3.4 accidentally inlines this function, so use noinline.
676 */
677
678static __initdata DECLARE_COMPLETION(kthreadd_done);
679
680noinline void __ref rest_init(void)
681{
682	struct task_struct *tsk;
683	int pid;
684
685	rcu_scheduler_starting();
686	/*
687	 * We need to spawn init first so that it obtains pid 1, however
688	 * the init task will end up wanting to create kthreads, which, if
689	 * we schedule it before we create kthreadd, will OOPS.
690	 */
691	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
692	/*
693	 * Pin init on the boot CPU. Task migration is not properly working
694	 * until sched_init_smp() has been run. It will set the allowed
695	 * CPUs for init to the non isolated CPUs.
696	 */
697	rcu_read_lock();
698	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
699	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
700	rcu_read_unlock();
701
702	numa_default_policy();
703	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
704	rcu_read_lock();
705	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
706	rcu_read_unlock();
707
708	/*
709	 * Enable might_sleep() and smp_processor_id() checks.
710	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
711	 * kernel_thread() would trigger might_sleep() splats. With
712	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
713	 * already, but it's stuck on the kthreadd_done completion.
714	 */
715	system_state = SYSTEM_SCHEDULING;
716
717	complete(&kthreadd_done);
718
719	/*
720	 * The boot idle thread must execute schedule()
721	 * at least once to get things moving:
722	 */
723	schedule_preempt_disabled();
724	/* Call into cpu_idle with preempt disabled */
725	cpu_startup_entry(CPUHP_ONLINE);
726}
727
728/* Check for early params. */
729static int __init do_early_param(char *param, char *val,
730				 const char *unused, void *arg)
731{
732	const struct obs_kernel_param *p;
733
734	for (p = __setup_start; p < __setup_end; p++) {
735		if ((p->early && parameq(param, p->str)) ||
736		    (strcmp(param, "console") == 0 &&
737		     strcmp(p->str, "earlycon") == 0)
738		) {
739			if (p->setup_func(val) != 0)
740				pr_warn("Malformed early option '%s'\n", param);
741		}
742	}
743	/* We accept everything at this stage. */
744	return 0;
745}
746
747void __init parse_early_options(char *cmdline)
748{
749	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
750		   do_early_param);
751}
752
753/* Arch code calls this early on, or if not, just before other parsing. */
754void __init parse_early_param(void)
755{
756	static int done __initdata;
757	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
758
759	if (done)
760		return;
761
762	/* All fall through to do_early_param. */
763	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
764	parse_early_options(tmp_cmdline);
765	done = 1;
766}
767
768void __init __weak arch_post_acpi_subsys_init(void) { }
769
770void __init __weak smp_setup_processor_id(void)
771{
772}
773
774# if THREAD_SIZE >= PAGE_SIZE
775void __init __weak thread_stack_cache_init(void)
776{
777}
778#endif
779
780void __init __weak poking_init(void) { }
781
782void __init __weak pgtable_cache_init(void) { }
783
784bool initcall_debug;
785core_param(initcall_debug, initcall_debug, bool, 0644);
786
787#ifdef TRACEPOINTS_ENABLED
788static void __init initcall_debug_enable(void);
789#else
790static inline void initcall_debug_enable(void)
791{
792}
793#endif
794
795/* Report memory auto-initialization states for this boot. */
796static void __init report_meminit(void)
797{
798	const char *stack;
799
800	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
801		stack = "all(pattern)";
802	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
803		stack = "all(zero)";
804	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
805		stack = "byref_all(zero)";
806	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
807		stack = "byref(zero)";
808	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
809		stack = "__user(zero)";
810	else
811		stack = "off";
812
813	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
814		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
815		want_init_on_free() ? "on" : "off");
816	if (want_init_on_free())
817		pr_info("mem auto-init: clearing system memory may take some time...\n");
818}
819
820/*
821 * Set up kernel memory allocators
822 */
823static void __init mm_init(void)
824{
825	/*
826	 * page_ext requires contiguous pages,
827	 * bigger than MAX_ORDER unless SPARSEMEM.
828	 */
829	page_ext_init_flatmem();
830	init_debug_pagealloc();
831	report_meminit();
832	mem_init();
833	/* page_owner must be initialized after buddy is ready */
834	page_ext_init_flatmem_late();
835	kmem_cache_init();
836	kmemleak_init();
837	pgtable_init();
838	debug_objects_mem_init();
839	vmalloc_init();
840	ioremap_huge_init();
841	/* Should be run before the first non-init thread is created */
842	init_espfix_bsp();
843	/* Should be run after espfix64 is set up. */
844	pti_init();
845	mm_cache_init();
846}
847
848void __init __weak arch_call_rest_init(void)
849{
850	rest_init();
851}
852
853asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
854{
855	char *command_line;
856	char *after_dashes;
857
858	set_task_stack_end_magic(&init_task);
859	smp_setup_processor_id();
860	debug_objects_early_init();
861
862	cgroup_init_early();
863
864	local_irq_disable();
865	early_boot_irqs_disabled = true;
866
867	/*
868	 * Interrupts are still disabled. Do necessary setups, then
869	 * enable them.
870	 */
871	boot_cpu_init();
872	page_address_init();
873	pr_notice("%s", linux_banner);
874	early_security_init();
875	setup_arch(&command_line);
876	setup_boot_config(command_line);
877	setup_command_line(command_line);
878	setup_nr_cpu_ids();
879	setup_per_cpu_areas();
880	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
881	boot_cpu_hotplug_init();
882
883	build_all_zonelists(NULL);
884	page_alloc_init();
885
886	pr_notice("Kernel command line: %s\n", saved_command_line);
887	/* parameters may set static keys */
888	jump_label_init();
889	parse_early_param();
890	after_dashes = parse_args("Booting kernel",
891				  static_command_line, __start___param,
892				  __stop___param - __start___param,
893				  -1, -1, NULL, &unknown_bootoption);
894	if (!IS_ERR_OR_NULL(after_dashes))
895		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
896			   NULL, set_init_arg);
897	if (extra_init_args)
898		parse_args("Setting extra init args", extra_init_args,
899			   NULL, 0, -1, -1, NULL, set_init_arg);
900
901	/*
902	 * These use large bootmem allocations and must precede
903	 * kmem_cache_init()
904	 */
905	setup_log_buf(0);
906	vfs_caches_init_early();
907	sort_main_extable();
908	trap_init();
909	mm_init();
910	poking_init();
911	ftrace_init();
912
913	/* trace_printk can be enabled here */
914	early_trace_init();
915
916	/*
917	 * Set up the scheduler prior starting any interrupts (such as the
918	 * timer interrupt). Full topology setup happens at smp_init()
919	 * time - but meanwhile we still have a functioning scheduler.
920	 */
921	sched_init();
922
923	if (WARN(!irqs_disabled(),
924		 "Interrupts were enabled *very* early, fixing it\n"))
925		local_irq_disable();
926	radix_tree_init();
927
928	/*
929	 * Set up housekeeping before setting up workqueues to allow the unbound
930	 * workqueue to take non-housekeeping into account.
931	 */
932	housekeeping_init();
933
934	/*
935	 * Allow workqueue creation and work item queueing/cancelling
936	 * early.  Work item execution depends on kthreads and starts after
937	 * workqueue_init().
938	 */
939	workqueue_init_early();
940
941	rcu_init();
942
943	/* Trace events are available after this */
944	trace_init();
945
946	if (initcall_debug)
947		initcall_debug_enable();
948
949	context_tracking_init();
950	/* init some links before init_ISA_irqs() */
951	early_irq_init();
952	init_IRQ();
953	tick_init();
954	rcu_init_nohz();
955	init_timers();
956	hrtimers_init();
957	softirq_init();
958	timekeeping_init();
959	time_init();
960
961	/*
962	 * For best initial stack canary entropy, prepare it after:
963	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
964	 * - timekeeping_init() for ktime entropy used in random_init()
965	 * - time_init() for making random_get_entropy() work on some platforms
966	 * - random_init() to initialize the RNG from from early entropy sources
967	 */
968	random_init(command_line);
969	boot_init_stack_canary();
970
971	perf_event_init();
972	profile_init();
973	call_function_init();
974	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
975
976	early_boot_irqs_disabled = false;
977	local_irq_enable();
978
979	kmem_cache_init_late();
980
981	/*
982	 * HACK ALERT! This is early. We're enabling the console before
983	 * we've done PCI setups etc, and console_init() must be aware of
984	 * this. But we do want output early, in case something goes wrong.
985	 */
986	console_init();
987	if (panic_later)
988		panic("Too many boot %s vars at `%s'", panic_later,
989		      panic_param);
990
991	lockdep_init();
992
993	/*
994	 * Need to run this when irqs are enabled, because it wants
995	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
996	 * too:
997	 */
998	locking_selftest();
999
1000#ifdef CONFIG_BLK_DEV_INITRD
1001	if (initrd_start && !initrd_below_start_ok &&
1002	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1003		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1004		    page_to_pfn(virt_to_page((void *)initrd_start)),
1005		    min_low_pfn);
1006		initrd_start = 0;
1007	}
1008#endif
1009	setup_per_cpu_pageset();
1010	numa_policy_init();
1011	acpi_early_init();
1012	if (late_time_init)
1013		late_time_init();
1014	sched_clock_init();
1015	calibrate_delay();
1016
1017	arch_cpu_finalize_init();
1018
1019	pid_idr_init();
1020	anon_vma_init();
1021#ifdef CONFIG_X86
1022	if (efi_enabled(EFI_RUNTIME_SERVICES))
1023		efi_enter_virtual_mode();
1024#endif
1025	thread_stack_cache_init();
1026	cred_init();
1027	fork_init();
1028	proc_caches_init();
1029	uts_ns_init();
1030	buffer_init();
1031	key_init();
1032	security_init();
1033	dbg_late_init();
1034	vfs_caches_init();
1035	pagecache_init();
1036	signals_init();
1037	seq_file_init();
1038	proc_root_init();
1039	nsfs_init();
1040	cpuset_init();
1041	cgroup_init();
1042	taskstats_init_early();
1043	delayacct_init();
1044#ifdef CONFIG_RECLAIM_ACCT
1045	reclaimacct_init();
1046#endif
1047
1048	acpi_subsystem_init();
1049	arch_post_acpi_subsys_init();
1050	sfi_init_late();
1051	kcsan_init();
1052
1053	/* Do the rest non-__init'ed, we're now alive */
1054	arch_call_rest_init();
1055
1056	prevent_tail_call_optimization();
1057}
1058
1059/* Call all constructor functions linked into the kernel. */
1060static void __init do_ctors(void)
1061{
1062#ifdef CONFIG_CONSTRUCTORS
1063	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1064
1065	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1066		(*fn)();
1067#endif
1068}
1069
1070#ifdef CONFIG_KALLSYMS
1071struct blacklist_entry {
1072	struct list_head next;
1073	char *buf;
1074};
1075
1076static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1077
1078static int __init initcall_blacklist(char *str)
1079{
1080	char *str_entry;
1081	struct blacklist_entry *entry;
1082
1083	/* str argument is a comma-separated list of functions */
1084	do {
1085		str_entry = strsep(&str, ",");
1086		if (str_entry) {
1087			pr_debug("blacklisting initcall %s\n", str_entry);
1088			entry = memblock_alloc(sizeof(*entry),
1089					       SMP_CACHE_BYTES);
1090			if (!entry)
1091				panic("%s: Failed to allocate %zu bytes\n",
1092				      __func__, sizeof(*entry));
1093			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1094						    SMP_CACHE_BYTES);
1095			if (!entry->buf)
1096				panic("%s: Failed to allocate %zu bytes\n",
1097				      __func__, strlen(str_entry) + 1);
1098			strcpy(entry->buf, str_entry);
1099			list_add(&entry->next, &blacklisted_initcalls);
1100		}
1101	} while (str_entry);
1102
1103	return 1;
1104}
1105
1106static bool __init_or_module initcall_blacklisted(initcall_t fn)
1107{
1108	struct blacklist_entry *entry;
1109	char fn_name[KSYM_SYMBOL_LEN];
1110	unsigned long addr;
1111
1112	if (list_empty(&blacklisted_initcalls))
1113		return false;
1114
1115	addr = (unsigned long) dereference_function_descriptor(fn);
1116	sprint_symbol_no_offset(fn_name, addr);
1117
1118	/*
1119	 * fn will be "function_name [module_name]" where [module_name] is not
1120	 * displayed for built-in init functions.  Strip off the [module_name].
1121	 */
1122	strreplace(fn_name, ' ', '\0');
1123
1124	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1125		if (!strcmp(fn_name, entry->buf)) {
1126			pr_debug("initcall %s blacklisted\n", fn_name);
1127			return true;
1128		}
1129	}
1130
1131	return false;
1132}
1133#else
1134static int __init initcall_blacklist(char *str)
1135{
1136	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1137	return 0;
1138}
1139
1140static bool __init_or_module initcall_blacklisted(initcall_t fn)
1141{
1142	return false;
1143}
1144#endif
1145__setup("initcall_blacklist=", initcall_blacklist);
1146
1147static __init_or_module void
1148trace_initcall_start_cb(void *data, initcall_t fn)
1149{
1150	ktime_t *calltime = (ktime_t *)data;
1151
1152	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1153	*calltime = ktime_get();
1154}
1155
1156static __init_or_module void
1157trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1158{
1159	ktime_t *calltime = (ktime_t *)data;
1160	ktime_t delta, rettime;
1161	unsigned long long duration;
1162
1163	rettime = ktime_get();
1164	delta = ktime_sub(rettime, *calltime);
1165	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1166	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1167		 fn, ret, duration);
1168}
1169
1170static ktime_t initcall_calltime;
1171
1172#ifdef TRACEPOINTS_ENABLED
1173static void __init initcall_debug_enable(void)
1174{
1175	int ret;
1176
1177	ret = register_trace_initcall_start(trace_initcall_start_cb,
1178					    &initcall_calltime);
1179	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1180					      &initcall_calltime);
1181	WARN(ret, "Failed to register initcall tracepoints\n");
1182}
1183# define do_trace_initcall_start	trace_initcall_start
1184# define do_trace_initcall_finish	trace_initcall_finish
1185#else
1186static inline void do_trace_initcall_start(initcall_t fn)
1187{
1188	if (!initcall_debug)
1189		return;
1190	trace_initcall_start_cb(&initcall_calltime, fn);
1191}
1192static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1193{
1194	if (!initcall_debug)
1195		return;
1196	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1197}
1198#endif /* !TRACEPOINTS_ENABLED */
1199
1200int __init_or_module do_one_initcall(initcall_t fn)
1201{
1202	int count = preempt_count();
1203	char msgbuf[64];
1204	int ret;
1205
1206	if (initcall_blacklisted(fn))
1207		return -EPERM;
1208
1209	do_trace_initcall_start(fn);
1210	ret = fn();
1211	do_trace_initcall_finish(fn, ret);
1212
1213	msgbuf[0] = 0;
1214
1215	if (preempt_count() != count) {
1216		sprintf(msgbuf, "preemption imbalance ");
1217		preempt_count_set(count);
1218	}
1219	if (irqs_disabled()) {
1220		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1221		local_irq_enable();
1222	}
1223	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1224
1225	add_latent_entropy();
1226	return ret;
1227}
1228
1229
1230extern initcall_entry_t __initcall_start[];
1231extern initcall_entry_t __initcall0_start[];
1232extern initcall_entry_t __initcall1_start[];
1233extern initcall_entry_t __initcall2_start[];
1234extern initcall_entry_t __initcall3_start[];
1235extern initcall_entry_t __initcall4_start[];
1236extern initcall_entry_t __initcall5_start[];
1237extern initcall_entry_t __initcall6_start[];
1238extern initcall_entry_t __initcall7_start[];
1239extern initcall_entry_t __initcall_end[];
1240
1241static initcall_entry_t *initcall_levels[] __initdata = {
1242	__initcall0_start,
1243	__initcall1_start,
1244	__initcall2_start,
1245	__initcall3_start,
1246	__initcall4_start,
1247	__initcall5_start,
1248	__initcall6_start,
1249	__initcall7_start,
1250	__initcall_end,
1251};
1252
1253/* Keep these in sync with initcalls in include/linux/init.h */
1254static const char *initcall_level_names[] __initdata = {
1255	"pure",
1256	"core",
1257	"postcore",
1258	"arch",
1259	"subsys",
1260	"fs",
1261	"device",
1262	"late",
1263};
1264
1265static int __init ignore_unknown_bootoption(char *param, char *val,
1266			       const char *unused, void *arg)
1267{
1268	return 0;
1269}
1270
1271static void __init do_initcall_level(int level, char *command_line)
1272{
1273	initcall_entry_t *fn;
1274
1275	parse_args(initcall_level_names[level],
1276		   command_line, __start___param,
1277		   __stop___param - __start___param,
1278		   level, level,
1279		   NULL, ignore_unknown_bootoption);
1280
1281	trace_initcall_level(initcall_level_names[level]);
1282	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1283		do_one_initcall(initcall_from_entry(fn));
1284}
1285
1286static void __init do_initcalls(void)
1287{
1288	int level;
1289	size_t len = strlen(saved_command_line) + 1;
1290	char *command_line;
1291
1292	command_line = kzalloc(len, GFP_KERNEL);
1293	if (!command_line)
1294		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1295
1296	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1297		/* Parser modifies command_line, restore it each time */
1298		strcpy(command_line, saved_command_line);
1299		do_initcall_level(level, command_line);
1300	}
1301
1302	kfree(command_line);
1303}
1304
1305/*
1306 * Ok, the machine is now initialized. None of the devices
1307 * have been touched yet, but the CPU subsystem is up and
1308 * running, and memory and process management works.
1309 *
1310 * Now we can finally start doing some real work..
1311 */
1312static void __init do_basic_setup(void)
1313{
1314	cpuset_init_smp();
1315	driver_init();
1316	init_irq_proc();
1317	do_ctors();
1318	usermodehelper_enable();
1319	do_initcalls();
1320}
1321
1322static void __init do_pre_smp_initcalls(void)
1323{
1324	initcall_entry_t *fn;
1325
1326	trace_initcall_level("early");
1327	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1328		do_one_initcall(initcall_from_entry(fn));
1329}
1330
1331static int run_init_process(const char *init_filename)
1332{
1333	const char *const *p;
1334
1335	argv_init[0] = init_filename;
1336	pr_info("Run %s as init process\n", init_filename);
1337	pr_debug("  with arguments:\n");
1338	for (p = argv_init; *p; p++)
1339		pr_debug("    %s\n", *p);
1340	pr_debug("  with environment:\n");
1341	for (p = envp_init; *p; p++)
1342		pr_debug("    %s\n", *p);
1343	return kernel_execve(init_filename, argv_init, envp_init);
1344}
1345
1346static int try_to_run_init_process(const char *init_filename)
1347{
1348	int ret;
1349
1350	ret = run_init_process(init_filename);
1351
1352	if (ret && ret != -ENOENT) {
1353		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1354		       init_filename, ret);
1355	}
1356
1357	return ret;
1358}
1359
1360static noinline void __init kernel_init_freeable(void);
1361
1362#if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1363bool rodata_enabled __ro_after_init = true;
1364static int __init set_debug_rodata(char *str)
1365{
1366	if (strtobool(str, &rodata_enabled))
1367		pr_warn("Invalid option string for rodata: '%s'\n", str);
1368	return 1;
1369}
1370__setup("rodata=", set_debug_rodata);
1371#endif
1372
1373#ifdef CONFIG_STRICT_KERNEL_RWX
1374static void mark_readonly(void)
1375{
1376	if (rodata_enabled) {
1377		/*
1378		 * load_module() results in W+X mappings, which are cleaned
1379		 * up with call_rcu().  Let's make sure that queued work is
1380		 * flushed so that we don't hit false positives looking for
1381		 * insecure pages which are W+X.
1382		 */
1383		rcu_barrier();
1384		mark_rodata_ro();
1385		rodata_test();
1386	} else
1387		pr_info("Kernel memory protection disabled.\n");
1388}
1389#elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1390static inline void mark_readonly(void)
1391{
1392	pr_warn("Kernel memory protection not selected by kernel config.\n");
1393}
1394#else
1395static inline void mark_readonly(void)
1396{
1397	pr_warn("This architecture does not have kernel memory protection.\n");
1398}
1399#endif
1400
1401void __weak free_initmem(void)
1402{
1403	free_initmem_default(POISON_FREE_INITMEM);
1404}
1405
1406static int __ref kernel_init(void *unused)
1407{
1408	int ret;
1409
1410	kernel_init_freeable();
1411	/* need to finish all async __init code before freeing the memory */
1412	async_synchronize_full();
1413	kprobe_free_init_mem();
1414	ftrace_free_init_mem();
1415	kgdb_free_init_mem();
1416	free_initmem();
1417	mark_readonly();
1418
1419	/*
1420	 * Kernel mappings are now finalized - update the userspace page-table
1421	 * to finalize PTI.
1422	 */
1423	pti_finalize();
1424
1425	system_state = SYSTEM_RUNNING;
1426	numa_default_policy();
1427
1428	rcu_end_inkernel_boot();
1429
1430	do_sysctl_args();
1431
1432	if (ramdisk_execute_command) {
1433		ret = run_init_process(ramdisk_execute_command);
1434		if (!ret)
1435			return 0;
1436		pr_err("Failed to execute %s (error %d)\n",
1437		       ramdisk_execute_command, ret);
1438	}
1439
1440	/*
1441	 * We try each of these until one succeeds.
1442	 *
1443	 * The Bourne shell can be used instead of init if we are
1444	 * trying to recover a really broken machine.
1445	 */
1446	if (execute_command) {
1447		ret = run_init_process(execute_command);
1448		if (!ret)
1449			return 0;
1450		panic("Requested init %s failed (error %d).",
1451		      execute_command, ret);
1452	}
1453
1454	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1455		ret = run_init_process(CONFIG_DEFAULT_INIT);
1456		if (ret)
1457			pr_err("Default init %s failed (error %d)\n",
1458			       CONFIG_DEFAULT_INIT, ret);
1459		else
1460			return 0;
1461	}
1462
1463	if (!try_to_run_init_process("/sbin/init") ||
1464	    !try_to_run_init_process("/etc/init") ||
1465	    !try_to_run_init_process("/bin/init") ||
1466	    !try_to_run_init_process("/bin/sh"))
1467		return 0;
1468
1469	panic("No working init found.  Try passing init= option to kernel. "
1470	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1471}
1472
1473/* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1474void __init console_on_rootfs(void)
1475{
1476	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1477
1478	if (IS_ERR(file)) {
1479		pr_err("Warning: unable to open an initial console.\n");
1480		return;
1481	}
1482	init_dup(file);
1483	init_dup(file);
1484	init_dup(file);
1485	fput(file);
1486}
1487
1488static noinline void __init kernel_init_freeable(void)
1489{
1490	/*
1491	 * Wait until kthreadd is all set-up.
1492	 */
1493	wait_for_completion(&kthreadd_done);
1494
1495	/* Now the scheduler is fully set up and can do blocking allocations */
1496	gfp_allowed_mask = __GFP_BITS_MASK;
1497
1498	/*
1499	 * init can allocate pages on any node
1500	 */
1501	set_mems_allowed(node_states[N_MEMORY]);
1502
1503	cad_pid = get_pid(task_pid(current));
1504
1505	smp_prepare_cpus(setup_max_cpus);
1506
1507	workqueue_init();
1508
1509	init_mm_internals();
1510
1511	rcu_init_tasks_generic();
1512	do_pre_smp_initcalls();
1513	lockup_detector_init();
1514
1515	smp_init();
1516	sched_init_smp();
1517
1518	padata_init();
1519	page_alloc_init_late();
1520	/* Initialize page ext after all struct pages are initialized. */
1521	page_ext_init();
1522
1523	do_basic_setup();
1524
1525	kunit_run_all_tests();
1526
1527	console_on_rootfs();
1528
1529	/*
1530	 * check if there is an early userspace init.  If yes, let it do all
1531	 * the work
1532	 */
1533	if (init_eaccess(ramdisk_execute_command) != 0) {
1534		ramdisk_execute_command = NULL;
1535		prepare_namespace();
1536	}
1537
1538	/*
1539	 * Ok, we have completed the initial bootup, and
1540	 * we're essentially up and running. Get rid of the
1541	 * initmem segments and start the user-mode stuff..
1542	 *
1543	 * rootfs is available now, try loading the public keys
1544	 * and default modules
1545	 */
1546
1547	integrity_load_keys();
1548}
1549