xref: /linux/init/main.c (revision 680fda4f671452d99401f294f20b60b3fdf24191)
1  // SPDX-License-Identifier: GPL-2.0-only
2  /*
3   *  linux/init/main.c
4   *
5   *  Copyright (C) 1991, 1992  Linus Torvalds
6   *
7   *  GK 2/5/95  -  Changed to support mounting root fs via NFS
8   *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9   *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10   *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
11   */
12  
13  #define DEBUG		/* Enable initcall_debug */
14  
15  #include <linux/types.h>
16  #include <linux/extable.h>
17  #include <linux/module.h>
18  #include <linux/proc_fs.h>
19  #include <linux/binfmts.h>
20  #include <linux/kernel.h>
21  #include <linux/syscalls.h>
22  #include <linux/stackprotector.h>
23  #include <linux/string.h>
24  #include <linux/ctype.h>
25  #include <linux/delay.h>
26  #include <linux/ioport.h>
27  #include <linux/init.h>
28  #include <linux/initrd.h>
29  #include <linux/memblock.h>
30  #include <linux/acpi.h>
31  #include <linux/bootconfig.h>
32  #include <linux/console.h>
33  #include <linux/nmi.h>
34  #include <linux/percpu.h>
35  #include <linux/kmod.h>
36  #include <linux/kprobes.h>
37  #include <linux/kmsan.h>
38  #include <linux/vmalloc.h>
39  #include <linux/kernel_stat.h>
40  #include <linux/start_kernel.h>
41  #include <linux/security.h>
42  #include <linux/smp.h>
43  #include <linux/profile.h>
44  #include <linux/kfence.h>
45  #include <linux/rcupdate.h>
46  #include <linux/srcu.h>
47  #include <linux/moduleparam.h>
48  #include <linux/kallsyms.h>
49  #include <linux/buildid.h>
50  #include <linux/writeback.h>
51  #include <linux/cpu.h>
52  #include <linux/cpuset.h>
53  #include <linux/cgroup.h>
54  #include <linux/efi.h>
55  #include <linux/tick.h>
56  #include <linux/sched/isolation.h>
57  #include <linux/interrupt.h>
58  #include <linux/taskstats_kern.h>
59  #include <linux/delayacct.h>
60  #include <linux/unistd.h>
61  #include <linux/utsname.h>
62  #include <linux/rmap.h>
63  #include <linux/mempolicy.h>
64  #include <linux/key.h>
65  #include <linux/debug_locks.h>
66  #include <linux/debugobjects.h>
67  #include <linux/lockdep.h>
68  #include <linux/kmemleak.h>
69  #include <linux/padata.h>
70  #include <linux/pid_namespace.h>
71  #include <linux/device/driver.h>
72  #include <linux/kthread.h>
73  #include <linux/sched.h>
74  #include <linux/sched/init.h>
75  #include <linux/signal.h>
76  #include <linux/idr.h>
77  #include <linux/kgdb.h>
78  #include <linux/ftrace.h>
79  #include <linux/async.h>
80  #include <linux/shmem_fs.h>
81  #include <linux/slab.h>
82  #include <linux/perf_event.h>
83  #include <linux/ptrace.h>
84  #include <linux/pti.h>
85  #include <linux/blkdev.h>
86  #include <linux/sched/clock.h>
87  #include <linux/sched/task.h>
88  #include <linux/sched/task_stack.h>
89  #include <linux/context_tracking.h>
90  #include <linux/random.h>
91  #include <linux/moduleloader.h>
92  #include <linux/list.h>
93  #include <linux/integrity.h>
94  #include <linux/proc_ns.h>
95  #include <linux/io.h>
96  #include <linux/cache.h>
97  #include <linux/rodata_test.h>
98  #include <linux/jump_label.h>
99  #include <linux/kcsan.h>
100  #include <linux/init_syscalls.h>
101  #include <linux/stackdepot.h>
102  #include <linux/randomize_kstack.h>
103  #include <linux/pidfs.h>
104  #include <linux/ptdump.h>
105  #include <net/net_namespace.h>
106  
107  #include <asm/io.h>
108  #include <asm/setup.h>
109  #include <asm/sections.h>
110  #include <asm/cacheflush.h>
111  
112  #define CREATE_TRACE_POINTS
113  #include <trace/events/initcall.h>
114  
115  #include <kunit/test.h>
116  
117  static int kernel_init(void *);
118  
119  /*
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   */
126  bool early_boot_irqs_disabled __read_mostly;
127  
128  enum system_states system_state __read_mostly;
129  EXPORT_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  
137  /* Default late time init is NULL. archs can override this later. */
138  void (*__initdata late_time_init)(void);
139  
140  /* Untouched command line saved by arch-specific code. */
141  char __initdata boot_command_line[COMMAND_LINE_SIZE];
142  /* Untouched saved command line (eg. for /proc) */
143  char *saved_command_line __ro_after_init;
144  unsigned int saved_command_line_len __ro_after_init;
145  /* Command line for parameter parsing */
146  static char *static_command_line;
147  /* Untouched extra command line */
148  static char *extra_command_line;
149  /* Extra init arguments */
150  static char *extra_init_args;
151  
152  #ifdef CONFIG_BOOT_CONFIG
153  /* Is bootconfig on command line? */
154  static bool bootconfig_found;
155  static size_t initargs_offs;
156  #else
157  # define bootconfig_found false
158  # define initargs_offs 0
159  #endif
160  
161  static char *execute_command;
162  static 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   */
168  bool static_key_initialized __read_mostly;
169  EXPORT_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   */
180  unsigned int reset_devices;
181  EXPORT_SYMBOL(reset_devices);
182  
183  static 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  
191  static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
192  const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
193  static const char *panic_later, *panic_param;
194  
195  static bool __init obsolete_checksetup(char *line)
196  {
197  	const struct obs_kernel_param *p;
198  	bool had_early_param = false;
199  
200  	p = __setup_start;
201  	do {
202  		int n = strlen(p->str);
203  		if (parameqn(line, p->str, n)) {
204  			if (p->early) {
205  				/* Already done in parse_early_param?
206  				 * (Needs exact match on param part).
207  				 * Keep iterating, as we can have early
208  				 * params and __setups of same names 8( */
209  				if (line[n] == '\0' || line[n] == '=')
210  					had_early_param = true;
211  			} else if (!p->setup_func) {
212  				pr_warn("Parameter %s is obsolete, ignored\n",
213  					p->str);
214  				return true;
215  			} else if (p->setup_func(line + n))
216  				return true;
217  		}
218  		p++;
219  	} while (p < __setup_end);
220  
221  	return had_early_param;
222  }
223  
224  /*
225   * This should be approx 2 Bo*oMips to start (note initial shift), and will
226   * still work even if initially too large, it will just take slightly longer
227   */
228  unsigned long loops_per_jiffy = (1<<12);
229  EXPORT_SYMBOL(loops_per_jiffy);
230  
231  static int __init debug_kernel(char *str)
232  {
233  	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
234  	return 0;
235  }
236  
237  static int __init quiet_kernel(char *str)
238  {
239  	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
240  	return 0;
241  }
242  
243  early_param("debug", debug_kernel);
244  early_param("quiet", quiet_kernel);
245  
246  static int __init loglevel(char *str)
247  {
248  	int newlevel;
249  
250  	/*
251  	 * Only update loglevel value when a correct setting was passed,
252  	 * to prevent blind crashes (when loglevel being set to 0) that
253  	 * are quite hard to debug
254  	 */
255  	if (get_option(&str, &newlevel)) {
256  		console_loglevel = newlevel;
257  		return 0;
258  	}
259  
260  	return -EINVAL;
261  }
262  
263  early_param("loglevel", loglevel);
264  
265  #ifdef CONFIG_BLK_DEV_INITRD
266  static void * __init get_boot_config_from_initrd(size_t *_size)
267  {
268  	u32 size, csum;
269  	char *data;
270  	u32 *hdr;
271  	int i;
272  
273  	if (!initrd_end)
274  		return NULL;
275  
276  	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
277  	/*
278  	 * Since Grub may align the size of initrd to 4, we must
279  	 * check the preceding 3 bytes as well.
280  	 */
281  	for (i = 0; i < 4; i++) {
282  		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
283  			goto found;
284  		data--;
285  	}
286  	return NULL;
287  
288  found:
289  	hdr = (u32 *)(data - 8);
290  	size = le32_to_cpu(hdr[0]);
291  	csum = le32_to_cpu(hdr[1]);
292  
293  	data = ((void *)hdr) - size;
294  	if ((unsigned long)data < initrd_start) {
295  		pr_err("bootconfig size %d is greater than initrd size %ld\n",
296  			size, initrd_end - initrd_start);
297  		return NULL;
298  	}
299  
300  	if (xbc_calc_checksum(data, size) != csum) {
301  		pr_err("bootconfig checksum failed\n");
302  		return NULL;
303  	}
304  
305  	/* Remove bootconfig from initramfs/initrd */
306  	initrd_end = (unsigned long)data;
307  	if (_size)
308  		*_size = size;
309  
310  	return data;
311  }
312  #else
313  static void * __init get_boot_config_from_initrd(size_t *_size)
314  {
315  	return NULL;
316  }
317  #endif
318  
319  #ifdef CONFIG_BOOT_CONFIG
320  
321  static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
322  
323  #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
324  
325  static int __init xbc_snprint_cmdline(char *buf, size_t size,
326  				      struct xbc_node *root)
327  {
328  	struct xbc_node *knode, *vnode;
329  	char *end = buf + size;
330  	const char *val;
331  	int ret;
332  
333  	xbc_node_for_each_key_value(root, knode, val) {
334  		ret = xbc_node_compose_key_after(root, knode,
335  					xbc_namebuf, XBC_KEYLEN_MAX);
336  		if (ret < 0)
337  			return ret;
338  
339  		vnode = xbc_node_get_child(knode);
340  		if (!vnode) {
341  			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
342  			if (ret < 0)
343  				return ret;
344  			buf += ret;
345  			continue;
346  		}
347  		xbc_array_for_each_value(vnode, val) {
348  			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
349  				       xbc_namebuf, val);
350  			if (ret < 0)
351  				return ret;
352  			buf += ret;
353  		}
354  	}
355  
356  	return buf - (end - size);
357  }
358  #undef rest
359  
360  /* Make an extra command line under given key word */
361  static char * __init xbc_make_cmdline(const char *key)
362  {
363  	struct xbc_node *root;
364  	char *new_cmdline;
365  	int ret, len = 0;
366  
367  	root = xbc_find_node(key);
368  	if (!root)
369  		return NULL;
370  
371  	/* Count required buffer size */
372  	len = xbc_snprint_cmdline(NULL, 0, root);
373  	if (len <= 0)
374  		return NULL;
375  
376  	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
377  	if (!new_cmdline) {
378  		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
379  		return NULL;
380  	}
381  
382  	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
383  	if (ret < 0 || ret > len) {
384  		pr_err("Failed to print extra kernel cmdline.\n");
385  		memblock_free(new_cmdline, len + 1);
386  		return NULL;
387  	}
388  
389  	return new_cmdline;
390  }
391  
392  static int __init bootconfig_params(char *param, char *val,
393  				    const char *unused, void *arg)
394  {
395  	if (strcmp(param, "bootconfig") == 0) {
396  		bootconfig_found = true;
397  	}
398  	return 0;
399  }
400  
401  static int __init warn_bootconfig(char *str)
402  {
403  	/* The 'bootconfig' has been handled by bootconfig_params(). */
404  	return 0;
405  }
406  
407  static void __init setup_boot_config(void)
408  {
409  	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
410  	const char *msg, *data;
411  	int pos, ret;
412  	size_t size;
413  	char *err;
414  
415  	/* Cut out the bootconfig data even if we have no bootconfig option */
416  	data = get_boot_config_from_initrd(&size);
417  	/* If there is no bootconfig in initrd, try embedded one. */
418  	if (!data)
419  		data = xbc_get_embedded_bootconfig(&size);
420  
421  	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
422  	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
423  			 bootconfig_params);
424  
425  	if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
426  		return;
427  
428  	/* parse_args() stops at the next param of '--' and returns an address */
429  	if (err)
430  		initargs_offs = err - tmp_cmdline;
431  
432  	if (!data) {
433  		/* If user intended to use bootconfig, show an error level message */
434  		if (bootconfig_found)
435  			pr_err("'bootconfig' found on command line, but no bootconfig found\n");
436  		else
437  			pr_info("No bootconfig data provided, so skipping bootconfig");
438  		return;
439  	}
440  
441  	if (size >= XBC_DATA_MAX) {
442  		pr_err("bootconfig size %ld greater than max size %d\n",
443  			(long)size, XBC_DATA_MAX);
444  		return;
445  	}
446  
447  	ret = xbc_init(data, size, &msg, &pos);
448  	if (ret < 0) {
449  		if (pos < 0)
450  			pr_err("Failed to init bootconfig: %s.\n", msg);
451  		else
452  			pr_err("Failed to parse bootconfig: %s at %d.\n",
453  				msg, pos);
454  	} else {
455  		xbc_get_info(&ret, NULL);
456  		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
457  		/* keys starting with "kernel." are passed via cmdline */
458  		extra_command_line = xbc_make_cmdline("kernel");
459  		/* Also, "init." keys are init arguments */
460  		extra_init_args = xbc_make_cmdline("init");
461  	}
462  	return;
463  }
464  
465  static void __init exit_boot_config(void)
466  {
467  	xbc_exit();
468  }
469  
470  #else	/* !CONFIG_BOOT_CONFIG */
471  
472  static void __init setup_boot_config(void)
473  {
474  	/* Remove bootconfig data from initrd */
475  	get_boot_config_from_initrd(NULL);
476  }
477  
478  static int __init warn_bootconfig(char *str)
479  {
480  	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
481  	return 0;
482  }
483  
484  #define exit_boot_config()	do {} while (0)
485  
486  #endif	/* CONFIG_BOOT_CONFIG */
487  
488  early_param("bootconfig", warn_bootconfig);
489  
490  bool __init cmdline_has_extra_options(void)
491  {
492  	return extra_command_line || extra_init_args;
493  }
494  
495  /* Change NUL term back to "=", to make "param" the whole string. */
496  static void __init repair_env_string(char *param, char *val)
497  {
498  	if (val) {
499  		/* param=val or param="val"? */
500  		if (val == param+strlen(param)+1)
501  			val[-1] = '=';
502  		else if (val == param+strlen(param)+2) {
503  			val[-2] = '=';
504  			memmove(val-1, val, strlen(val)+1);
505  		} else
506  			BUG();
507  	}
508  }
509  
510  /* Anything after -- gets handed straight to init. */
511  static int __init set_init_arg(char *param, char *val,
512  			       const char *unused, void *arg)
513  {
514  	unsigned int i;
515  
516  	if (panic_later)
517  		return 0;
518  
519  	repair_env_string(param, val);
520  
521  	for (i = 0; argv_init[i]; i++) {
522  		if (i == MAX_INIT_ARGS) {
523  			panic_later = "init";
524  			panic_param = param;
525  			return 0;
526  		}
527  	}
528  	argv_init[i] = param;
529  	return 0;
530  }
531  
532  /*
533   * Unknown boot options get handed to init, unless they look like
534   * unused parameters (modprobe will find them in /proc/cmdline).
535   */
536  static int __init unknown_bootoption(char *param, char *val,
537  				     const char *unused, void *arg)
538  {
539  	size_t len = strlen(param);
540  
541  	/* Handle params aliased to sysctls */
542  	if (sysctl_is_alias(param))
543  		return 0;
544  
545  	repair_env_string(param, val);
546  
547  	/* Handle obsolete-style parameters */
548  	if (obsolete_checksetup(param))
549  		return 0;
550  
551  	/* Unused module parameter. */
552  	if (strnchr(param, len, '.'))
553  		return 0;
554  
555  	if (panic_later)
556  		return 0;
557  
558  	if (val) {
559  		/* Environment option */
560  		unsigned int i;
561  		for (i = 0; envp_init[i]; i++) {
562  			if (i == MAX_INIT_ENVS) {
563  				panic_later = "env";
564  				panic_param = param;
565  			}
566  			if (!strncmp(param, envp_init[i], len+1))
567  				break;
568  		}
569  		envp_init[i] = param;
570  	} else {
571  		/* Command line option */
572  		unsigned int i;
573  		for (i = 0; argv_init[i]; i++) {
574  			if (i == MAX_INIT_ARGS) {
575  				panic_later = "init";
576  				panic_param = param;
577  			}
578  		}
579  		argv_init[i] = param;
580  	}
581  	return 0;
582  }
583  
584  static int __init init_setup(char *str)
585  {
586  	unsigned int i;
587  
588  	execute_command = str;
589  	/*
590  	 * In case LILO is going to boot us with default command line,
591  	 * it prepends "auto" before the whole cmdline which makes
592  	 * the shell think it should execute a script with such name.
593  	 * So we ignore all arguments entered _before_ init=... [MJ]
594  	 */
595  	for (i = 1; i < MAX_INIT_ARGS; i++)
596  		argv_init[i] = NULL;
597  	return 1;
598  }
599  __setup("init=", init_setup);
600  
601  static int __init rdinit_setup(char *str)
602  {
603  	unsigned int i;
604  
605  	ramdisk_execute_command = str;
606  	/* See "auto" comment in init_setup */
607  	for (i = 1; i < MAX_INIT_ARGS; i++)
608  		argv_init[i] = NULL;
609  	return 1;
610  }
611  __setup("rdinit=", rdinit_setup);
612  
613  #ifndef CONFIG_SMP
614  static inline void setup_nr_cpu_ids(void) { }
615  static inline void smp_prepare_cpus(unsigned int maxcpus) { }
616  #endif
617  
618  /*
619   * We need to store the untouched command line for future reference.
620   * We also need to store the touched command line since the parameter
621   * parsing is performed in place, and we should allow a component to
622   * store reference of name/value for future reference.
623   */
624  static void __init setup_command_line(char *command_line)
625  {
626  	size_t len, xlen = 0, ilen = 0;
627  
628  	if (extra_command_line)
629  		xlen = strlen(extra_command_line);
630  	if (extra_init_args)
631  		ilen = strlen(extra_init_args) + 4; /* for " -- " */
632  
633  	len = xlen + strlen(boot_command_line) + 1;
634  
635  	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
636  	if (!saved_command_line)
637  		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
638  
639  	len = xlen + strlen(command_line) + 1;
640  
641  	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
642  	if (!static_command_line)
643  		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
644  
645  	if (xlen) {
646  		/*
647  		 * We have to put extra_command_line before boot command
648  		 * lines because there could be dashes (separator of init
649  		 * command line) in the command lines.
650  		 */
651  		strcpy(saved_command_line, extra_command_line);
652  		strcpy(static_command_line, extra_command_line);
653  	}
654  	strcpy(saved_command_line + xlen, boot_command_line);
655  	strcpy(static_command_line + xlen, command_line);
656  
657  	if (ilen) {
658  		/*
659  		 * Append supplemental init boot args to saved_command_line
660  		 * so that user can check what command line options passed
661  		 * to init.
662  		 * The order should always be
663  		 * " -- "[bootconfig init-param][cmdline init-param]
664  		 */
665  		if (initargs_offs) {
666  			len = xlen + initargs_offs;
667  			strcpy(saved_command_line + len, extra_init_args);
668  			len += ilen - 4;	/* strlen(extra_init_args) */
669  			strcpy(saved_command_line + len,
670  				boot_command_line + initargs_offs - 1);
671  		} else {
672  			len = strlen(saved_command_line);
673  			strcpy(saved_command_line + len, " -- ");
674  			len += 4;
675  			strcpy(saved_command_line + len, extra_init_args);
676  		}
677  	}
678  
679  	saved_command_line_len = strlen(saved_command_line);
680  }
681  
682  /*
683   * We need to finalize in a non-__init function or else race conditions
684   * between the root thread and the init thread may cause start_kernel to
685   * be reaped by free_initmem before the root thread has proceeded to
686   * cpu_idle.
687   *
688   * gcc-3.4 accidentally inlines this function, so use noinline.
689   */
690  
691  static __initdata DECLARE_COMPLETION(kthreadd_done);
692  
693  static noinline void __ref __noreturn rest_init(void)
694  {
695  	struct task_struct *tsk;
696  	int pid;
697  
698  	rcu_scheduler_starting();
699  	/*
700  	 * We need to spawn init first so that it obtains pid 1, however
701  	 * the init task will end up wanting to create kthreads, which, if
702  	 * we schedule it before we create kthreadd, will OOPS.
703  	 */
704  	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
705  	/*
706  	 * Pin init on the boot CPU. Task migration is not properly working
707  	 * until sched_init_smp() has been run. It will set the allowed
708  	 * CPUs for init to the non isolated CPUs.
709  	 */
710  	rcu_read_lock();
711  	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
712  	tsk->flags |= PF_NO_SETAFFINITY;
713  	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
714  	rcu_read_unlock();
715  
716  	numa_default_policy();
717  	pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
718  	rcu_read_lock();
719  	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
720  	rcu_read_unlock();
721  
722  	/*
723  	 * Enable might_sleep() and smp_processor_id() checks.
724  	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
725  	 * kernel_thread() would trigger might_sleep() splats. With
726  	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
727  	 * already, but it's stuck on the kthreadd_done completion.
728  	 */
729  	system_state = SYSTEM_SCHEDULING;
730  
731  	complete(&kthreadd_done);
732  
733  	/*
734  	 * The boot idle thread must execute schedule()
735  	 * at least once to get things moving:
736  	 */
737  	schedule_preempt_disabled();
738  	/* Call into cpu_idle with preempt disabled */
739  	cpu_startup_entry(CPUHP_ONLINE);
740  }
741  
742  /* Check for early params. */
743  static int __init do_early_param(char *param, char *val,
744  				 const char *unused, void *arg)
745  {
746  	const struct obs_kernel_param *p;
747  
748  	for (p = __setup_start; p < __setup_end; p++) {
749  		if ((p->early && parameq(param, p->str)) ||
750  		    (strcmp(param, "console") == 0 &&
751  		     strcmp(p->str, "earlycon") == 0)
752  		) {
753  			if (p->setup_func(val) != 0)
754  				pr_warn("Malformed early option '%s'\n", param);
755  		}
756  	}
757  	/* We accept everything at this stage. */
758  	return 0;
759  }
760  
761  void __init parse_early_options(char *cmdline)
762  {
763  	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
764  		   do_early_param);
765  }
766  
767  /* Arch code calls this early on, or if not, just before other parsing. */
768  void __init parse_early_param(void)
769  {
770  	static int done __initdata;
771  	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
772  
773  	if (done)
774  		return;
775  
776  	/* All fall through to do_early_param. */
777  	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
778  	parse_early_options(tmp_cmdline);
779  	done = 1;
780  }
781  
782  void __init __weak arch_post_acpi_subsys_init(void) { }
783  
784  void __init __weak smp_setup_processor_id(void)
785  {
786  }
787  
788  void __init __weak smp_prepare_boot_cpu(void)
789  {
790  }
791  
792  # if THREAD_SIZE >= PAGE_SIZE
793  void __init __weak thread_stack_cache_init(void)
794  {
795  }
796  #endif
797  
798  void __init __weak poking_init(void) { }
799  
800  void __init __weak pgtable_cache_init(void) { }
801  
802  void __init __weak trap_init(void) { }
803  
804  bool initcall_debug;
805  core_param(initcall_debug, initcall_debug, bool, 0644);
806  
807  #ifdef TRACEPOINTS_ENABLED
808  static void __init initcall_debug_enable(void);
809  #else
810  static inline void initcall_debug_enable(void)
811  {
812  }
813  #endif
814  
815  #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
816  DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
817  			   randomize_kstack_offset);
818  DEFINE_PER_CPU(u32, kstack_offset);
819  
820  static int __init early_randomize_kstack_offset(char *buf)
821  {
822  	int ret;
823  	bool bool_result;
824  
825  	ret = kstrtobool(buf, &bool_result);
826  	if (ret)
827  		return ret;
828  
829  	if (bool_result)
830  		static_branch_enable(&randomize_kstack_offset);
831  	else
832  		static_branch_disable(&randomize_kstack_offset);
833  	return 0;
834  }
835  early_param("randomize_kstack_offset", early_randomize_kstack_offset);
836  #endif
837  
838  static void __init print_unknown_bootoptions(void)
839  {
840  	char *unknown_options;
841  	char *end;
842  	const char *const *p;
843  	size_t len;
844  
845  	if (panic_later || (!argv_init[1] && !envp_init[2]))
846  		return;
847  
848  	/*
849  	 * Determine how many options we have to print out, plus a space
850  	 * before each
851  	 */
852  	len = 1; /* null terminator */
853  	for (p = &argv_init[1]; *p; p++) {
854  		len++;
855  		len += strlen(*p);
856  	}
857  	for (p = &envp_init[2]; *p; p++) {
858  		len++;
859  		len += strlen(*p);
860  	}
861  
862  	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
863  	if (!unknown_options) {
864  		pr_err("%s: Failed to allocate %zu bytes\n",
865  			__func__, len);
866  		return;
867  	}
868  	end = unknown_options;
869  
870  	for (p = &argv_init[1]; *p; p++)
871  		end += sprintf(end, " %s", *p);
872  	for (p = &envp_init[2]; *p; p++)
873  		end += sprintf(end, " %s", *p);
874  
875  	/* Start at unknown_options[1] to skip the initial space */
876  	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
877  		&unknown_options[1]);
878  	memblock_free(unknown_options, len);
879  }
880  
881  asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
882  void start_kernel(void)
883  {
884  	char *command_line;
885  	char *after_dashes;
886  
887  	set_task_stack_end_magic(&init_task);
888  	smp_setup_processor_id();
889  	debug_objects_early_init();
890  	init_vmlinux_build_id();
891  
892  	cgroup_init_early();
893  
894  	local_irq_disable();
895  	early_boot_irqs_disabled = true;
896  
897  	/*
898  	 * Interrupts are still disabled. Do necessary setups, then
899  	 * enable them.
900  	 */
901  	boot_cpu_init();
902  	page_address_init();
903  	pr_notice("%s", linux_banner);
904  	early_security_init();
905  	setup_arch(&command_line);
906  	setup_boot_config();
907  	setup_command_line(command_line);
908  	setup_nr_cpu_ids();
909  	setup_per_cpu_areas();
910  	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
911  	boot_cpu_hotplug_init();
912  
913  	pr_notice("Kernel command line: %s\n", saved_command_line);
914  	/* parameters may set static keys */
915  	jump_label_init();
916  	parse_early_param();
917  	after_dashes = parse_args("Booting kernel",
918  				  static_command_line, __start___param,
919  				  __stop___param - __start___param,
920  				  -1, -1, NULL, &unknown_bootoption);
921  	print_unknown_bootoptions();
922  	if (!IS_ERR_OR_NULL(after_dashes))
923  		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
924  			   NULL, set_init_arg);
925  	if (extra_init_args)
926  		parse_args("Setting extra init args", extra_init_args,
927  			   NULL, 0, -1, -1, NULL, set_init_arg);
928  
929  	/* Architectural and non-timekeeping rng init, before allocator init */
930  	random_init_early(command_line);
931  
932  	/*
933  	 * These use large bootmem allocations and must precede
934  	 * initalization of page allocator
935  	 */
936  	setup_log_buf(0);
937  	vfs_caches_init_early();
938  	sort_main_extable();
939  	trap_init();
940  	mm_core_init();
941  	poking_init();
942  	ftrace_init();
943  
944  	/* trace_printk can be enabled here */
945  	early_trace_init();
946  
947  	/*
948  	 * Set up the scheduler prior starting any interrupts (such as the
949  	 * timer interrupt). Full topology setup happens at smp_init()
950  	 * time - but meanwhile we still have a functioning scheduler.
951  	 */
952  	sched_init();
953  
954  	if (WARN(!irqs_disabled(),
955  		 "Interrupts were enabled *very* early, fixing it\n"))
956  		local_irq_disable();
957  	radix_tree_init();
958  	maple_tree_init();
959  
960  	/*
961  	 * Set up housekeeping before setting up workqueues to allow the unbound
962  	 * workqueue to take non-housekeeping into account.
963  	 */
964  	housekeeping_init();
965  
966  	/*
967  	 * Allow workqueue creation and work item queueing/cancelling
968  	 * early.  Work item execution depends on kthreads and starts after
969  	 * workqueue_init().
970  	 */
971  	workqueue_init_early();
972  
973  	rcu_init();
974  
975  	/* Trace events are available after this */
976  	trace_init();
977  
978  	if (initcall_debug)
979  		initcall_debug_enable();
980  
981  	context_tracking_init();
982  	/* init some links before init_ISA_irqs() */
983  	early_irq_init();
984  	init_IRQ();
985  	tick_init();
986  	rcu_init_nohz();
987  	init_timers();
988  	srcu_init();
989  	hrtimers_init();
990  	softirq_init();
991  	timekeeping_init();
992  	time_init();
993  
994  	/* This must be after timekeeping is initialized */
995  	random_init();
996  
997  	/* These make use of the fully initialized rng */
998  	kfence_init();
999  	boot_init_stack_canary();
1000  
1001  	perf_event_init();
1002  	profile_init();
1003  	call_function_init();
1004  	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1005  
1006  	early_boot_irqs_disabled = false;
1007  	local_irq_enable();
1008  
1009  	kmem_cache_init_late();
1010  
1011  	/*
1012  	 * HACK ALERT! This is early. We're enabling the console before
1013  	 * we've done PCI setups etc, and console_init() must be aware of
1014  	 * this. But we do want output early, in case something goes wrong.
1015  	 */
1016  	console_init();
1017  	if (panic_later)
1018  		panic("Too many boot %s vars at `%s'", panic_later,
1019  		      panic_param);
1020  
1021  	lockdep_init();
1022  
1023  	/*
1024  	 * Need to run this when irqs are enabled, because it wants
1025  	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1026  	 * too:
1027  	 */
1028  	locking_selftest();
1029  
1030  #ifdef CONFIG_BLK_DEV_INITRD
1031  	if (initrd_start && !initrd_below_start_ok &&
1032  	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1033  		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1034  		    page_to_pfn(virt_to_page((void *)initrd_start)),
1035  		    min_low_pfn);
1036  		initrd_start = 0;
1037  	}
1038  #endif
1039  	setup_per_cpu_pageset();
1040  	numa_policy_init();
1041  	acpi_early_init();
1042  	if (late_time_init)
1043  		late_time_init();
1044  	sched_clock_init();
1045  	calibrate_delay();
1046  
1047  	arch_cpu_finalize_init();
1048  
1049  	pid_idr_init();
1050  	anon_vma_init();
1051  #ifdef CONFIG_X86
1052  	if (efi_enabled(EFI_RUNTIME_SERVICES))
1053  		efi_enter_virtual_mode();
1054  #endif
1055  	thread_stack_cache_init();
1056  	cred_init();
1057  	fork_init();
1058  	proc_caches_init();
1059  	uts_ns_init();
1060  	key_init();
1061  	security_init();
1062  	dbg_late_init();
1063  	net_ns_init();
1064  	vfs_caches_init();
1065  	pagecache_init();
1066  	signals_init();
1067  	seq_file_init();
1068  	proc_root_init();
1069  	nsfs_init();
1070  	pidfs_init();
1071  	cpuset_init();
1072  	cgroup_init();
1073  	taskstats_init_early();
1074  	delayacct_init();
1075  
1076  	acpi_subsystem_init();
1077  	arch_post_acpi_subsys_init();
1078  	kcsan_init();
1079  
1080  	/* Do the rest non-__init'ed, we're now alive */
1081  	rest_init();
1082  
1083  	/*
1084  	 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1085  	 * and older.
1086  	 */
1087  #if !__has_attribute(__no_stack_protector__)
1088  	prevent_tail_call_optimization();
1089  #endif
1090  }
1091  
1092  /* Call all constructor functions linked into the kernel. */
1093  static void __init do_ctors(void)
1094  {
1095  /*
1096   * For UML, the constructors have already been called by the
1097   * normal setup code as it's just a normal ELF binary, so we
1098   * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1099   * even on UML for modules.
1100   */
1101  #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1102  	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1103  
1104  	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1105  		(*fn)();
1106  #endif
1107  }
1108  
1109  #ifdef CONFIG_KALLSYMS
1110  struct blacklist_entry {
1111  	struct list_head next;
1112  	char *buf;
1113  };
1114  
1115  static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1116  
1117  static int __init initcall_blacklist(char *str)
1118  {
1119  	char *str_entry;
1120  	struct blacklist_entry *entry;
1121  
1122  	/* str argument is a comma-separated list of functions */
1123  	do {
1124  		str_entry = strsep(&str, ",");
1125  		if (str_entry) {
1126  			pr_debug("blacklisting initcall %s\n", str_entry);
1127  			entry = memblock_alloc(sizeof(*entry),
1128  					       SMP_CACHE_BYTES);
1129  			if (!entry)
1130  				panic("%s: Failed to allocate %zu bytes\n",
1131  				      __func__, sizeof(*entry));
1132  			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1133  						    SMP_CACHE_BYTES);
1134  			if (!entry->buf)
1135  				panic("%s: Failed to allocate %zu bytes\n",
1136  				      __func__, strlen(str_entry) + 1);
1137  			strcpy(entry->buf, str_entry);
1138  			list_add(&entry->next, &blacklisted_initcalls);
1139  		}
1140  	} while (str_entry);
1141  
1142  	return 1;
1143  }
1144  
1145  static bool __init_or_module initcall_blacklisted(initcall_t fn)
1146  {
1147  	struct blacklist_entry *entry;
1148  	char fn_name[KSYM_SYMBOL_LEN];
1149  	unsigned long addr;
1150  
1151  	if (list_empty(&blacklisted_initcalls))
1152  		return false;
1153  
1154  	addr = (unsigned long) dereference_function_descriptor(fn);
1155  	sprint_symbol_no_offset(fn_name, addr);
1156  
1157  	/*
1158  	 * fn will be "function_name [module_name]" where [module_name] is not
1159  	 * displayed for built-in init functions.  Strip off the [module_name].
1160  	 */
1161  	strreplace(fn_name, ' ', '\0');
1162  
1163  	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1164  		if (!strcmp(fn_name, entry->buf)) {
1165  			pr_debug("initcall %s blacklisted\n", fn_name);
1166  			return true;
1167  		}
1168  	}
1169  
1170  	return false;
1171  }
1172  #else
1173  static int __init initcall_blacklist(char *str)
1174  {
1175  	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1176  	return 0;
1177  }
1178  
1179  static bool __init_or_module initcall_blacklisted(initcall_t fn)
1180  {
1181  	return false;
1182  }
1183  #endif
1184  __setup("initcall_blacklist=", initcall_blacklist);
1185  
1186  static __init_or_module void
1187  trace_initcall_start_cb(void *data, initcall_t fn)
1188  {
1189  	ktime_t *calltime = data;
1190  
1191  	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1192  	*calltime = ktime_get();
1193  }
1194  
1195  static __init_or_module void
1196  trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1197  {
1198  	ktime_t rettime, *calltime = data;
1199  
1200  	rettime = ktime_get();
1201  	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1202  		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1203  }
1204  
1205  static ktime_t initcall_calltime;
1206  
1207  #ifdef TRACEPOINTS_ENABLED
1208  static void __init initcall_debug_enable(void)
1209  {
1210  	int ret;
1211  
1212  	ret = register_trace_initcall_start(trace_initcall_start_cb,
1213  					    &initcall_calltime);
1214  	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1215  					      &initcall_calltime);
1216  	WARN(ret, "Failed to register initcall tracepoints\n");
1217  }
1218  # define do_trace_initcall_start	trace_initcall_start
1219  # define do_trace_initcall_finish	trace_initcall_finish
1220  #else
1221  static inline void do_trace_initcall_start(initcall_t fn)
1222  {
1223  	if (!initcall_debug)
1224  		return;
1225  	trace_initcall_start_cb(&initcall_calltime, fn);
1226  }
1227  static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1228  {
1229  	if (!initcall_debug)
1230  		return;
1231  	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1232  }
1233  #endif /* !TRACEPOINTS_ENABLED */
1234  
1235  int __init_or_module do_one_initcall(initcall_t fn)
1236  {
1237  	int count = preempt_count();
1238  	char msgbuf[64];
1239  	int ret;
1240  
1241  	if (initcall_blacklisted(fn))
1242  		return -EPERM;
1243  
1244  	do_trace_initcall_start(fn);
1245  	ret = fn();
1246  	do_trace_initcall_finish(fn, ret);
1247  
1248  	msgbuf[0] = 0;
1249  
1250  	if (preempt_count() != count) {
1251  		sprintf(msgbuf, "preemption imbalance ");
1252  		preempt_count_set(count);
1253  	}
1254  	if (irqs_disabled()) {
1255  		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1256  		local_irq_enable();
1257  	}
1258  	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1259  
1260  	add_latent_entropy();
1261  	return ret;
1262  }
1263  
1264  
1265  static initcall_entry_t *initcall_levels[] __initdata = {
1266  	__initcall0_start,
1267  	__initcall1_start,
1268  	__initcall2_start,
1269  	__initcall3_start,
1270  	__initcall4_start,
1271  	__initcall5_start,
1272  	__initcall6_start,
1273  	__initcall7_start,
1274  	__initcall_end,
1275  };
1276  
1277  /* Keep these in sync with initcalls in include/linux/init.h */
1278  static const char *initcall_level_names[] __initdata = {
1279  	"pure",
1280  	"core",
1281  	"postcore",
1282  	"arch",
1283  	"subsys",
1284  	"fs",
1285  	"device",
1286  	"late",
1287  };
1288  
1289  static int __init ignore_unknown_bootoption(char *param, char *val,
1290  			       const char *unused, void *arg)
1291  {
1292  	return 0;
1293  }
1294  
1295  static void __init do_initcall_level(int level, char *command_line)
1296  {
1297  	initcall_entry_t *fn;
1298  
1299  	parse_args(initcall_level_names[level],
1300  		   command_line, __start___param,
1301  		   __stop___param - __start___param,
1302  		   level, level,
1303  		   NULL, ignore_unknown_bootoption);
1304  
1305  	trace_initcall_level(initcall_level_names[level]);
1306  	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1307  		do_one_initcall(initcall_from_entry(fn));
1308  }
1309  
1310  static void __init do_initcalls(void)
1311  {
1312  	int level;
1313  	size_t len = saved_command_line_len + 1;
1314  	char *command_line;
1315  
1316  	command_line = kzalloc(len, GFP_KERNEL);
1317  	if (!command_line)
1318  		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1319  
1320  	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1321  		/* Parser modifies command_line, restore it each time */
1322  		strcpy(command_line, saved_command_line);
1323  		do_initcall_level(level, command_line);
1324  	}
1325  
1326  	kfree(command_line);
1327  }
1328  
1329  /*
1330   * Ok, the machine is now initialized. None of the devices
1331   * have been touched yet, but the CPU subsystem is up and
1332   * running, and memory and process management works.
1333   *
1334   * Now we can finally start doing some real work..
1335   */
1336  static void __init do_basic_setup(void)
1337  {
1338  	cpuset_init_smp();
1339  	driver_init();
1340  	init_irq_proc();
1341  	do_ctors();
1342  	do_initcalls();
1343  }
1344  
1345  static void __init do_pre_smp_initcalls(void)
1346  {
1347  	initcall_entry_t *fn;
1348  
1349  	trace_initcall_level("early");
1350  	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1351  		do_one_initcall(initcall_from_entry(fn));
1352  }
1353  
1354  static int run_init_process(const char *init_filename)
1355  {
1356  	const char *const *p;
1357  
1358  	argv_init[0] = init_filename;
1359  	pr_info("Run %s as init process\n", init_filename);
1360  	pr_debug("  with arguments:\n");
1361  	for (p = argv_init; *p; p++)
1362  		pr_debug("    %s\n", *p);
1363  	pr_debug("  with environment:\n");
1364  	for (p = envp_init; *p; p++)
1365  		pr_debug("    %s\n", *p);
1366  	return kernel_execve(init_filename, argv_init, envp_init);
1367  }
1368  
1369  static int try_to_run_init_process(const char *init_filename)
1370  {
1371  	int ret;
1372  
1373  	ret = run_init_process(init_filename);
1374  
1375  	if (ret && ret != -ENOENT) {
1376  		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1377  		       init_filename, ret);
1378  	}
1379  
1380  	return ret;
1381  }
1382  
1383  static noinline void __init kernel_init_freeable(void);
1384  
1385  #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1386  bool rodata_enabled __ro_after_init = true;
1387  
1388  #ifndef arch_parse_debug_rodata
1389  static inline bool arch_parse_debug_rodata(char *str) { return false; }
1390  #endif
1391  
1392  static int __init set_debug_rodata(char *str)
1393  {
1394  	if (arch_parse_debug_rodata(str))
1395  		return 0;
1396  
1397  	if (str && !strcmp(str, "on"))
1398  		rodata_enabled = true;
1399  	else if (str && !strcmp(str, "off"))
1400  		rodata_enabled = false;
1401  	else
1402  		pr_warn("Invalid option string for rodata: '%s'\n", str);
1403  	return 0;
1404  }
1405  early_param("rodata", set_debug_rodata);
1406  #endif
1407  
1408  static void mark_readonly(void)
1409  {
1410  	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1411  		/*
1412  		 * load_module() results in W+X mappings, which are cleaned
1413  		 * up with init_free_wq. Let's make sure that queued work is
1414  		 * flushed so that we don't hit false positives looking for
1415  		 * insecure pages which are W+X.
1416  		 */
1417  		flush_module_init_free_work();
1418  		mark_rodata_ro();
1419  		debug_checkwx();
1420  		rodata_test();
1421  	} else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1422  		pr_info("Kernel memory protection disabled.\n");
1423  	} else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1424  		pr_warn("Kernel memory protection not selected by kernel config.\n");
1425  	} else {
1426  		pr_warn("This architecture does not have kernel memory protection.\n");
1427  	}
1428  }
1429  
1430  void __weak free_initmem(void)
1431  {
1432  	free_initmem_default(POISON_FREE_INITMEM);
1433  }
1434  
1435  static int __ref kernel_init(void *unused)
1436  {
1437  	int ret;
1438  
1439  	/*
1440  	 * Wait until kthreadd is all set-up.
1441  	 */
1442  	wait_for_completion(&kthreadd_done);
1443  
1444  	kernel_init_freeable();
1445  	/* need to finish all async __init code before freeing the memory */
1446  	async_synchronize_full();
1447  
1448  	system_state = SYSTEM_FREEING_INITMEM;
1449  	kprobe_free_init_mem();
1450  	ftrace_free_init_mem();
1451  	kgdb_free_init_mem();
1452  	exit_boot_config();
1453  	free_initmem();
1454  	mark_readonly();
1455  
1456  	/*
1457  	 * Kernel mappings are now finalized - update the userspace page-table
1458  	 * to finalize PTI.
1459  	 */
1460  	pti_finalize();
1461  
1462  	system_state = SYSTEM_RUNNING;
1463  	numa_default_policy();
1464  
1465  	rcu_end_inkernel_boot();
1466  
1467  	do_sysctl_args();
1468  
1469  	if (ramdisk_execute_command) {
1470  		ret = run_init_process(ramdisk_execute_command);
1471  		if (!ret)
1472  			return 0;
1473  		pr_err("Failed to execute %s (error %d)\n",
1474  		       ramdisk_execute_command, ret);
1475  	}
1476  
1477  	/*
1478  	 * We try each of these until one succeeds.
1479  	 *
1480  	 * The Bourne shell can be used instead of init if we are
1481  	 * trying to recover a really broken machine.
1482  	 */
1483  	if (execute_command) {
1484  		ret = run_init_process(execute_command);
1485  		if (!ret)
1486  			return 0;
1487  		panic("Requested init %s failed (error %d).",
1488  		      execute_command, ret);
1489  	}
1490  
1491  	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1492  		ret = run_init_process(CONFIG_DEFAULT_INIT);
1493  		if (ret)
1494  			pr_err("Default init %s failed (error %d)\n",
1495  			       CONFIG_DEFAULT_INIT, ret);
1496  		else
1497  			return 0;
1498  	}
1499  
1500  	if (!try_to_run_init_process("/sbin/init") ||
1501  	    !try_to_run_init_process("/etc/init") ||
1502  	    !try_to_run_init_process("/bin/init") ||
1503  	    !try_to_run_init_process("/bin/sh"))
1504  		return 0;
1505  
1506  	panic("No working init found.  Try passing init= option to kernel. "
1507  	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1508  }
1509  
1510  /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1511  void __init console_on_rootfs(void)
1512  {
1513  	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1514  
1515  	if (IS_ERR(file)) {
1516  		pr_err("Warning: unable to open an initial console.\n");
1517  		return;
1518  	}
1519  	init_dup(file);
1520  	init_dup(file);
1521  	init_dup(file);
1522  	fput(file);
1523  }
1524  
1525  static noinline void __init kernel_init_freeable(void)
1526  {
1527  	/* Now the scheduler is fully set up and can do blocking allocations */
1528  	gfp_allowed_mask = __GFP_BITS_MASK;
1529  
1530  	/*
1531  	 * init can allocate pages on any node
1532  	 */
1533  	set_mems_allowed(node_states[N_MEMORY]);
1534  
1535  	cad_pid = get_pid(task_pid(current));
1536  
1537  	smp_prepare_cpus(setup_max_cpus);
1538  
1539  	workqueue_init();
1540  
1541  	init_mm_internals();
1542  
1543  	rcu_init_tasks_generic();
1544  	do_pre_smp_initcalls();
1545  	lockup_detector_init();
1546  
1547  	smp_init();
1548  	sched_init_smp();
1549  
1550  	workqueue_init_topology();
1551  	async_init();
1552  	padata_init();
1553  	page_alloc_init_late();
1554  
1555  	do_basic_setup();
1556  
1557  	kunit_run_all_tests();
1558  
1559  	wait_for_initramfs();
1560  	console_on_rootfs();
1561  
1562  	/*
1563  	 * check if there is an early userspace init.  If yes, let it do all
1564  	 * the work
1565  	 */
1566  	if (init_eaccess(ramdisk_execute_command) != 0) {
1567  		ramdisk_execute_command = NULL;
1568  		prepare_namespace();
1569  	}
1570  
1571  	/*
1572  	 * Ok, we have completed the initial bootup, and
1573  	 * we're essentially up and running. Get rid of the
1574  	 * initmem segments and start the user-mode stuff..
1575  	 *
1576  	 * rootfs is available now, try loading the public keys
1577  	 * and default modules
1578  	 */
1579  
1580  	integrity_load_keys();
1581  }
1582