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