xref: /linux/init/main.c (revision 3fd6c59042dbba50391e30862beac979491145fe)
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 
set_reset_devices(char * str)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 
obsolete_checksetup(char * line)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 
debug_kernel(char * str)231 static int __init debug_kernel(char *str)
232 {
233 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
234 	return 0;
235 }
236 
quiet_kernel(char * str)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 
loglevel(char * str)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
get_boot_config_from_initrd(size_t * _size)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
get_boot_config_from_initrd(size_t * _size)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 
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)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, *q;
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 			/*
349 			 * For prettier and more readable /proc/cmdline, only
350 			 * quote the value when necessary, i.e. when it contains
351 			 * whitespace.
352 			 */
353 			q = strpbrk(val, " \t\r\n") ? "\"" : "";
354 			ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
355 				       xbc_namebuf, q, val, q);
356 			if (ret < 0)
357 				return ret;
358 			buf += ret;
359 		}
360 	}
361 
362 	return buf - (end - size);
363 }
364 #undef rest
365 
366 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)367 static char * __init xbc_make_cmdline(const char *key)
368 {
369 	struct xbc_node *root;
370 	char *new_cmdline;
371 	int ret, len = 0;
372 
373 	root = xbc_find_node(key);
374 	if (!root)
375 		return NULL;
376 
377 	/* Count required buffer size */
378 	len = xbc_snprint_cmdline(NULL, 0, root);
379 	if (len <= 0)
380 		return NULL;
381 
382 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
383 	if (!new_cmdline) {
384 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
385 		return NULL;
386 	}
387 
388 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
389 	if (ret < 0 || ret > len) {
390 		pr_err("Failed to print extra kernel cmdline.\n");
391 		memblock_free(new_cmdline, len + 1);
392 		return NULL;
393 	}
394 
395 	return new_cmdline;
396 }
397 
bootconfig_params(char * param,char * val,const char * unused,void * arg)398 static int __init bootconfig_params(char *param, char *val,
399 				    const char *unused, void *arg)
400 {
401 	if (strcmp(param, "bootconfig") == 0) {
402 		bootconfig_found = true;
403 	}
404 	return 0;
405 }
406 
warn_bootconfig(char * str)407 static int __init warn_bootconfig(char *str)
408 {
409 	/* The 'bootconfig' has been handled by bootconfig_params(). */
410 	return 0;
411 }
412 
setup_boot_config(void)413 static void __init setup_boot_config(void)
414 {
415 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
416 	const char *msg, *data;
417 	int pos, ret;
418 	size_t size;
419 	char *err;
420 
421 	/* Cut out the bootconfig data even if we have no bootconfig option */
422 	data = get_boot_config_from_initrd(&size);
423 	/* If there is no bootconfig in initrd, try embedded one. */
424 	if (!data)
425 		data = xbc_get_embedded_bootconfig(&size);
426 
427 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
428 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
429 			 bootconfig_params);
430 
431 	if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
432 		return;
433 
434 	/* parse_args() stops at the next param of '--' and returns an address */
435 	if (err)
436 		initargs_offs = err - tmp_cmdline;
437 
438 	if (!data) {
439 		/* If user intended to use bootconfig, show an error level message */
440 		if (bootconfig_found)
441 			pr_err("'bootconfig' found on command line, but no bootconfig found\n");
442 		else
443 			pr_info("No bootconfig data provided, so skipping bootconfig");
444 		return;
445 	}
446 
447 	if (size >= XBC_DATA_MAX) {
448 		pr_err("bootconfig size %ld greater than max size %d\n",
449 			(long)size, XBC_DATA_MAX);
450 		return;
451 	}
452 
453 	ret = xbc_init(data, size, &msg, &pos);
454 	if (ret < 0) {
455 		if (pos < 0)
456 			pr_err("Failed to init bootconfig: %s.\n", msg);
457 		else
458 			pr_err("Failed to parse bootconfig: %s at %d.\n",
459 				msg, pos);
460 	} else {
461 		xbc_get_info(&ret, NULL);
462 		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
463 		/* keys starting with "kernel." are passed via cmdline */
464 		extra_command_line = xbc_make_cmdline("kernel");
465 		/* Also, "init." keys are init arguments */
466 		extra_init_args = xbc_make_cmdline("init");
467 	}
468 	return;
469 }
470 
exit_boot_config(void)471 static void __init exit_boot_config(void)
472 {
473 	xbc_exit();
474 }
475 
476 #else	/* !CONFIG_BOOT_CONFIG */
477 
setup_boot_config(void)478 static void __init setup_boot_config(void)
479 {
480 	/* Remove bootconfig data from initrd */
481 	get_boot_config_from_initrd(NULL);
482 }
483 
warn_bootconfig(char * str)484 static int __init warn_bootconfig(char *str)
485 {
486 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
487 	return 0;
488 }
489 
490 #define exit_boot_config()	do {} while (0)
491 
492 #endif	/* CONFIG_BOOT_CONFIG */
493 
494 early_param("bootconfig", warn_bootconfig);
495 
cmdline_has_extra_options(void)496 bool __init cmdline_has_extra_options(void)
497 {
498 	return extra_command_line || extra_init_args;
499 }
500 
501 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)502 static void __init repair_env_string(char *param, char *val)
503 {
504 	if (val) {
505 		/* param=val or param="val"? */
506 		if (val == param+strlen(param)+1)
507 			val[-1] = '=';
508 		else if (val == param+strlen(param)+2) {
509 			val[-2] = '=';
510 			memmove(val-1, val, strlen(val)+1);
511 		} else
512 			BUG();
513 	}
514 }
515 
516 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)517 static int __init set_init_arg(char *param, char *val,
518 			       const char *unused, void *arg)
519 {
520 	unsigned int i;
521 
522 	if (panic_later)
523 		return 0;
524 
525 	repair_env_string(param, val);
526 
527 	for (i = 0; argv_init[i]; i++) {
528 		if (i == MAX_INIT_ARGS) {
529 			panic_later = "init";
530 			panic_param = param;
531 			return 0;
532 		}
533 	}
534 	argv_init[i] = param;
535 	return 0;
536 }
537 
538 /*
539  * Unknown boot options get handed to init, unless they look like
540  * unused parameters (modprobe will find them in /proc/cmdline).
541  */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)542 static int __init unknown_bootoption(char *param, char *val,
543 				     const char *unused, void *arg)
544 {
545 	size_t len = strlen(param);
546 
547 	/* Handle params aliased to sysctls */
548 	if (sysctl_is_alias(param))
549 		return 0;
550 
551 	repair_env_string(param, val);
552 
553 	/* Handle obsolete-style parameters */
554 	if (obsolete_checksetup(param))
555 		return 0;
556 
557 	/* Unused module parameter. */
558 	if (strnchr(param, len, '.'))
559 		return 0;
560 
561 	if (panic_later)
562 		return 0;
563 
564 	if (val) {
565 		/* Environment option */
566 		unsigned int i;
567 		for (i = 0; envp_init[i]; i++) {
568 			if (i == MAX_INIT_ENVS) {
569 				panic_later = "env";
570 				panic_param = param;
571 			}
572 			if (!strncmp(param, envp_init[i], len+1))
573 				break;
574 		}
575 		envp_init[i] = param;
576 	} else {
577 		/* Command line option */
578 		unsigned int i;
579 		for (i = 0; argv_init[i]; i++) {
580 			if (i == MAX_INIT_ARGS) {
581 				panic_later = "init";
582 				panic_param = param;
583 			}
584 		}
585 		argv_init[i] = param;
586 	}
587 	return 0;
588 }
589 
init_setup(char * str)590 static int __init init_setup(char *str)
591 {
592 	unsigned int i;
593 
594 	execute_command = str;
595 	/*
596 	 * In case LILO is going to boot us with default command line,
597 	 * it prepends "auto" before the whole cmdline which makes
598 	 * the shell think it should execute a script with such name.
599 	 * So we ignore all arguments entered _before_ init=... [MJ]
600 	 */
601 	for (i = 1; i < MAX_INIT_ARGS; i++)
602 		argv_init[i] = NULL;
603 	return 1;
604 }
605 __setup("init=", init_setup);
606 
rdinit_setup(char * str)607 static int __init rdinit_setup(char *str)
608 {
609 	unsigned int i;
610 
611 	ramdisk_execute_command = str;
612 	/* See "auto" comment in init_setup */
613 	for (i = 1; i < MAX_INIT_ARGS; i++)
614 		argv_init[i] = NULL;
615 	return 1;
616 }
617 __setup("rdinit=", rdinit_setup);
618 
619 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)620 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)621 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
622 #endif
623 
624 /*
625  * We need to store the untouched command line for future reference.
626  * We also need to store the touched command line since the parameter
627  * parsing is performed in place, and we should allow a component to
628  * store reference of name/value for future reference.
629  */
setup_command_line(char * command_line)630 static void __init setup_command_line(char *command_line)
631 {
632 	size_t len, xlen = 0, ilen = 0;
633 
634 	if (extra_command_line)
635 		xlen = strlen(extra_command_line);
636 	if (extra_init_args) {
637 		extra_init_args = strim(extra_init_args); /* remove trailing space */
638 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
639 	}
640 
641 	len = xlen + strlen(boot_command_line) + ilen + 1;
642 
643 	saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
644 	if (!saved_command_line)
645 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
646 
647 	len = xlen + strlen(command_line) + 1;
648 
649 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
650 	if (!static_command_line)
651 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
652 
653 	if (xlen) {
654 		/*
655 		 * We have to put extra_command_line before boot command
656 		 * lines because there could be dashes (separator of init
657 		 * command line) in the command lines.
658 		 */
659 		strcpy(saved_command_line, extra_command_line);
660 		strcpy(static_command_line, extra_command_line);
661 	}
662 	strcpy(saved_command_line + xlen, boot_command_line);
663 	strcpy(static_command_line + xlen, command_line);
664 
665 	if (ilen) {
666 		/*
667 		 * Append supplemental init boot args to saved_command_line
668 		 * so that user can check what command line options passed
669 		 * to init.
670 		 * The order should always be
671 		 * " -- "[bootconfig init-param][cmdline init-param]
672 		 */
673 		if (initargs_offs) {
674 			len = xlen + initargs_offs;
675 			strcpy(saved_command_line + len, extra_init_args);
676 			len += ilen - 4;	/* strlen(extra_init_args) */
677 			strcpy(saved_command_line + len,
678 				boot_command_line + initargs_offs - 1);
679 		} else {
680 			len = strlen(saved_command_line);
681 			strcpy(saved_command_line + len, " -- ");
682 			len += 4;
683 			strcpy(saved_command_line + len, extra_init_args);
684 		}
685 	}
686 
687 	saved_command_line_len = strlen(saved_command_line);
688 }
689 
690 /*
691  * We need to finalize in a non-__init function or else race conditions
692  * between the root thread and the init thread may cause start_kernel to
693  * be reaped by free_initmem before the root thread has proceeded to
694  * cpu_idle.
695  *
696  * gcc-3.4 accidentally inlines this function, so use noinline.
697  */
698 
699 static __initdata DECLARE_COMPLETION(kthreadd_done);
700 
rest_init(void)701 static noinline void __ref __noreturn rest_init(void)
702 {
703 	struct task_struct *tsk;
704 	int pid;
705 
706 	rcu_scheduler_starting();
707 	/*
708 	 * We need to spawn init first so that it obtains pid 1, however
709 	 * the init task will end up wanting to create kthreads, which, if
710 	 * we schedule it before we create kthreadd, will OOPS.
711 	 */
712 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
713 	/*
714 	 * Pin init on the boot CPU. Task migration is not properly working
715 	 * until sched_init_smp() has been run. It will set the allowed
716 	 * CPUs for init to the non isolated CPUs.
717 	 */
718 	rcu_read_lock();
719 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
720 	tsk->flags |= PF_NO_SETAFFINITY;
721 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
722 	rcu_read_unlock();
723 
724 	numa_default_policy();
725 	pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
726 	rcu_read_lock();
727 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
728 	rcu_read_unlock();
729 
730 	/*
731 	 * Enable might_sleep() and smp_processor_id() checks.
732 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
733 	 * kernel_thread() would trigger might_sleep() splats. With
734 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
735 	 * already, but it's stuck on the kthreadd_done completion.
736 	 */
737 	system_state = SYSTEM_SCHEDULING;
738 
739 	complete(&kthreadd_done);
740 
741 	/*
742 	 * The boot idle thread must execute schedule()
743 	 * at least once to get things moving:
744 	 */
745 	schedule_preempt_disabled();
746 	/* Call into cpu_idle with preempt disabled */
747 	cpu_startup_entry(CPUHP_ONLINE);
748 }
749 
750 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)751 static int __init do_early_param(char *param, char *val,
752 				 const char *unused, void *arg)
753 {
754 	const struct obs_kernel_param *p;
755 
756 	for (p = __setup_start; p < __setup_end; p++) {
757 		if (p->early && parameq(param, p->str)) {
758 			if (p->setup_func(val) != 0)
759 				pr_warn("Malformed early option '%s'\n", param);
760 		}
761 	}
762 	/* We accept everything at this stage. */
763 	return 0;
764 }
765 
parse_early_options(char * cmdline)766 void __init parse_early_options(char *cmdline)
767 {
768 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
769 		   do_early_param);
770 }
771 
772 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)773 void __init parse_early_param(void)
774 {
775 	static int done __initdata;
776 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
777 
778 	if (done)
779 		return;
780 
781 	/* All fall through to do_early_param. */
782 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
783 	parse_early_options(tmp_cmdline);
784 	done = 1;
785 }
786 
arch_post_acpi_subsys_init(void)787 void __init __weak arch_post_acpi_subsys_init(void) { }
788 
smp_setup_processor_id(void)789 void __init __weak smp_setup_processor_id(void)
790 {
791 }
792 
smp_prepare_boot_cpu(void)793 void __init __weak smp_prepare_boot_cpu(void)
794 {
795 }
796 
797 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)798 void __init __weak thread_stack_cache_init(void)
799 {
800 }
801 #endif
802 
poking_init(void)803 void __init __weak poking_init(void) { }
804 
pgtable_cache_init(void)805 void __init __weak pgtable_cache_init(void) { }
806 
trap_init(void)807 void __init __weak trap_init(void) { }
808 
809 bool initcall_debug;
810 core_param(initcall_debug, initcall_debug, bool, 0644);
811 
812 #ifdef TRACEPOINTS_ENABLED
813 static void __init initcall_debug_enable(void);
814 #else
initcall_debug_enable(void)815 static inline void initcall_debug_enable(void)
816 {
817 }
818 #endif
819 
820 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
821 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
822 			   randomize_kstack_offset);
823 DEFINE_PER_CPU(u32, kstack_offset);
824 
early_randomize_kstack_offset(char * buf)825 static int __init early_randomize_kstack_offset(char *buf)
826 {
827 	int ret;
828 	bool bool_result;
829 
830 	ret = kstrtobool(buf, &bool_result);
831 	if (ret)
832 		return ret;
833 
834 	if (bool_result)
835 		static_branch_enable(&randomize_kstack_offset);
836 	else
837 		static_branch_disable(&randomize_kstack_offset);
838 	return 0;
839 }
840 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
841 #endif
842 
print_unknown_bootoptions(void)843 static void __init print_unknown_bootoptions(void)
844 {
845 	char *unknown_options;
846 	char *end;
847 	const char *const *p;
848 	size_t len;
849 
850 	if (panic_later || (!argv_init[1] && !envp_init[2]))
851 		return;
852 
853 	/*
854 	 * Determine how many options we have to print out, plus a space
855 	 * before each
856 	 */
857 	len = 1; /* null terminator */
858 	for (p = &argv_init[1]; *p; p++) {
859 		len++;
860 		len += strlen(*p);
861 	}
862 	for (p = &envp_init[2]; *p; p++) {
863 		len++;
864 		len += strlen(*p);
865 	}
866 
867 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
868 	if (!unknown_options) {
869 		pr_err("%s: Failed to allocate %zu bytes\n",
870 			__func__, len);
871 		return;
872 	}
873 	end = unknown_options;
874 
875 	for (p = &argv_init[1]; *p; p++)
876 		end += sprintf(end, " %s", *p);
877 	for (p = &envp_init[2]; *p; p++)
878 		end += sprintf(end, " %s", *p);
879 
880 	/* Start at unknown_options[1] to skip the initial space */
881 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
882 		&unknown_options[1]);
883 	memblock_free(unknown_options, len);
884 }
885 
early_numa_node_init(void)886 static void __init early_numa_node_init(void)
887 {
888 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
889 #ifndef cpu_to_node
890 	int cpu;
891 
892 	/* The early_cpu_to_node() should be ready here. */
893 	for_each_possible_cpu(cpu)
894 		set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
895 #endif
896 #endif
897 }
898 
899 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)900 void start_kernel(void)
901 {
902 	char *command_line;
903 	char *after_dashes;
904 
905 	set_task_stack_end_magic(&init_task);
906 	smp_setup_processor_id();
907 	debug_objects_early_init();
908 	init_vmlinux_build_id();
909 
910 	cgroup_init_early();
911 
912 	local_irq_disable();
913 	early_boot_irqs_disabled = true;
914 
915 	/*
916 	 * Interrupts are still disabled. Do necessary setups, then
917 	 * enable them.
918 	 */
919 	boot_cpu_init();
920 	page_address_init();
921 	pr_notice("%s", linux_banner);
922 	setup_arch(&command_line);
923 	/* Static keys and static calls are needed by LSMs */
924 	jump_label_init();
925 	static_call_init();
926 	early_security_init();
927 	setup_boot_config();
928 	setup_command_line(command_line);
929 	setup_nr_cpu_ids();
930 	setup_per_cpu_areas();
931 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
932 	early_numa_node_init();
933 	boot_cpu_hotplug_init();
934 
935 	pr_notice("Kernel command line: %s\n", saved_command_line);
936 	/* parameters may set static keys */
937 	parse_early_param();
938 	after_dashes = parse_args("Booting kernel",
939 				  static_command_line, __start___param,
940 				  __stop___param - __start___param,
941 				  -1, -1, NULL, &unknown_bootoption);
942 	print_unknown_bootoptions();
943 	if (!IS_ERR_OR_NULL(after_dashes))
944 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
945 			   NULL, set_init_arg);
946 	if (extra_init_args)
947 		parse_args("Setting extra init args", extra_init_args,
948 			   NULL, 0, -1, -1, NULL, set_init_arg);
949 
950 	/* Architectural and non-timekeeping rng init, before allocator init */
951 	random_init_early(command_line);
952 
953 	/*
954 	 * These use large bootmem allocations and must precede
955 	 * initalization of page allocator
956 	 */
957 	setup_log_buf(0);
958 	vfs_caches_init_early();
959 	sort_main_extable();
960 	trap_init();
961 	mm_core_init();
962 	poking_init();
963 	ftrace_init();
964 
965 	/* trace_printk can be enabled here */
966 	early_trace_init();
967 
968 	/*
969 	 * Set up the scheduler prior starting any interrupts (such as the
970 	 * timer interrupt). Full topology setup happens at smp_init()
971 	 * time - but meanwhile we still have a functioning scheduler.
972 	 */
973 	sched_init();
974 
975 	if (WARN(!irqs_disabled(),
976 		 "Interrupts were enabled *very* early, fixing it\n"))
977 		local_irq_disable();
978 	radix_tree_init();
979 	maple_tree_init();
980 
981 	/*
982 	 * Set up housekeeping before setting up workqueues to allow the unbound
983 	 * workqueue to take non-housekeeping into account.
984 	 */
985 	housekeeping_init();
986 
987 	/*
988 	 * Allow workqueue creation and work item queueing/cancelling
989 	 * early.  Work item execution depends on kthreads and starts after
990 	 * workqueue_init().
991 	 */
992 	workqueue_init_early();
993 
994 	rcu_init();
995 
996 	/* Trace events are available after this */
997 	trace_init();
998 
999 	if (initcall_debug)
1000 		initcall_debug_enable();
1001 
1002 	context_tracking_init();
1003 	/* init some links before init_ISA_irqs() */
1004 	early_irq_init();
1005 	init_IRQ();
1006 	tick_init();
1007 	rcu_init_nohz();
1008 	init_timers();
1009 	srcu_init();
1010 	hrtimers_init();
1011 	softirq_init();
1012 	timekeeping_init();
1013 	time_init();
1014 
1015 	/* This must be after timekeeping is initialized */
1016 	random_init();
1017 
1018 	/* These make use of the fully initialized rng */
1019 	kfence_init();
1020 	boot_init_stack_canary();
1021 
1022 	perf_event_init();
1023 	profile_init();
1024 	call_function_init();
1025 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1026 
1027 	early_boot_irqs_disabled = false;
1028 	local_irq_enable();
1029 
1030 	kmem_cache_init_late();
1031 
1032 	/*
1033 	 * HACK ALERT! This is early. We're enabling the console before
1034 	 * we've done PCI setups etc, and console_init() must be aware of
1035 	 * this. But we do want output early, in case something goes wrong.
1036 	 */
1037 	console_init();
1038 	if (panic_later)
1039 		panic("Too many boot %s vars at `%s'", panic_later,
1040 		      panic_param);
1041 
1042 	lockdep_init();
1043 
1044 	/*
1045 	 * Need to run this when irqs are enabled, because it wants
1046 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1047 	 * too:
1048 	 */
1049 	locking_selftest();
1050 
1051 #ifdef CONFIG_BLK_DEV_INITRD
1052 	if (initrd_start && !initrd_below_start_ok &&
1053 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1054 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1055 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1056 		    min_low_pfn);
1057 		initrd_start = 0;
1058 	}
1059 #endif
1060 	setup_per_cpu_pageset();
1061 	numa_policy_init();
1062 	acpi_early_init();
1063 	if (late_time_init)
1064 		late_time_init();
1065 	sched_clock_init();
1066 	calibrate_delay();
1067 
1068 	arch_cpu_finalize_init();
1069 
1070 	pid_idr_init();
1071 	anon_vma_init();
1072 #ifdef CONFIG_X86
1073 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1074 		efi_enter_virtual_mode();
1075 #endif
1076 	thread_stack_cache_init();
1077 	cred_init();
1078 	fork_init();
1079 	proc_caches_init();
1080 	uts_ns_init();
1081 	key_init();
1082 	security_init();
1083 	dbg_late_init();
1084 	net_ns_init();
1085 	vfs_caches_init();
1086 	pagecache_init();
1087 	signals_init();
1088 	seq_file_init();
1089 	proc_root_init();
1090 	nsfs_init();
1091 	pidfs_init();
1092 	cpuset_init();
1093 	cgroup_init();
1094 	taskstats_init_early();
1095 	delayacct_init();
1096 
1097 	acpi_subsystem_init();
1098 	arch_post_acpi_subsys_init();
1099 	kcsan_init();
1100 
1101 	/* Do the rest non-__init'ed, we're now alive */
1102 	rest_init();
1103 
1104 	/*
1105 	 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1106 	 * and older.
1107 	 */
1108 #if !__has_attribute(__no_stack_protector__)
1109 	prevent_tail_call_optimization();
1110 #endif
1111 }
1112 
1113 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1114 static void __init do_ctors(void)
1115 {
1116 /*
1117  * For UML, the constructors have already been called by the
1118  * normal setup code as it's just a normal ELF binary, so we
1119  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1120  * even on UML for modules.
1121  */
1122 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1123 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1124 
1125 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1126 		(*fn)();
1127 #endif
1128 }
1129 
1130 #ifdef CONFIG_KALLSYMS
1131 struct blacklist_entry {
1132 	struct list_head next;
1133 	char *buf;
1134 };
1135 
1136 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1137 
initcall_blacklist(char * str)1138 static int __init initcall_blacklist(char *str)
1139 {
1140 	char *str_entry;
1141 	struct blacklist_entry *entry;
1142 
1143 	/* str argument is a comma-separated list of functions */
1144 	do {
1145 		str_entry = strsep(&str, ",");
1146 		if (str_entry) {
1147 			pr_debug("blacklisting initcall %s\n", str_entry);
1148 			entry = memblock_alloc(sizeof(*entry),
1149 					       SMP_CACHE_BYTES);
1150 			if (!entry)
1151 				panic("%s: Failed to allocate %zu bytes\n",
1152 				      __func__, sizeof(*entry));
1153 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1154 						    SMP_CACHE_BYTES);
1155 			if (!entry->buf)
1156 				panic("%s: Failed to allocate %zu bytes\n",
1157 				      __func__, strlen(str_entry) + 1);
1158 			strcpy(entry->buf, str_entry);
1159 			list_add(&entry->next, &blacklisted_initcalls);
1160 		}
1161 	} while (str_entry);
1162 
1163 	return 1;
1164 }
1165 
initcall_blacklisted(initcall_t fn)1166 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1167 {
1168 	struct blacklist_entry *entry;
1169 	char fn_name[KSYM_SYMBOL_LEN];
1170 	unsigned long addr;
1171 
1172 	if (list_empty(&blacklisted_initcalls))
1173 		return false;
1174 
1175 	addr = (unsigned long) dereference_function_descriptor(fn);
1176 	sprint_symbol_no_offset(fn_name, addr);
1177 
1178 	/*
1179 	 * fn will be "function_name [module_name]" where [module_name] is not
1180 	 * displayed for built-in init functions.  Strip off the [module_name].
1181 	 */
1182 	strreplace(fn_name, ' ', '\0');
1183 
1184 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1185 		if (!strcmp(fn_name, entry->buf)) {
1186 			pr_debug("initcall %s blacklisted\n", fn_name);
1187 			return true;
1188 		}
1189 	}
1190 
1191 	return false;
1192 }
1193 #else
initcall_blacklist(char * str)1194 static int __init initcall_blacklist(char *str)
1195 {
1196 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1197 	return 0;
1198 }
1199 
initcall_blacklisted(initcall_t fn)1200 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1201 {
1202 	return false;
1203 }
1204 #endif
1205 __setup("initcall_blacklist=", initcall_blacklist);
1206 
1207 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1208 trace_initcall_start_cb(void *data, initcall_t fn)
1209 {
1210 	ktime_t *calltime = data;
1211 
1212 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1213 	*calltime = ktime_get();
1214 }
1215 
1216 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1217 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1218 {
1219 	ktime_t rettime, *calltime = data;
1220 
1221 	rettime = ktime_get();
1222 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1223 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1224 }
1225 
1226 static ktime_t initcall_calltime;
1227 
1228 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1229 static void __init initcall_debug_enable(void)
1230 {
1231 	int ret;
1232 
1233 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1234 					    &initcall_calltime);
1235 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1236 					      &initcall_calltime);
1237 	WARN(ret, "Failed to register initcall tracepoints\n");
1238 }
1239 # define do_trace_initcall_start	trace_initcall_start
1240 # define do_trace_initcall_finish	trace_initcall_finish
1241 #else
do_trace_initcall_start(initcall_t fn)1242 static inline void do_trace_initcall_start(initcall_t fn)
1243 {
1244 	if (!initcall_debug)
1245 		return;
1246 	trace_initcall_start_cb(&initcall_calltime, fn);
1247 }
do_trace_initcall_finish(initcall_t fn,int ret)1248 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1249 {
1250 	if (!initcall_debug)
1251 		return;
1252 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1253 }
1254 #endif /* !TRACEPOINTS_ENABLED */
1255 
do_one_initcall(initcall_t fn)1256 int __init_or_module do_one_initcall(initcall_t fn)
1257 {
1258 	int count = preempt_count();
1259 	char msgbuf[64];
1260 	int ret;
1261 
1262 	if (initcall_blacklisted(fn))
1263 		return -EPERM;
1264 
1265 	do_trace_initcall_start(fn);
1266 	ret = fn();
1267 	do_trace_initcall_finish(fn, ret);
1268 
1269 	msgbuf[0] = 0;
1270 
1271 	if (preempt_count() != count) {
1272 		sprintf(msgbuf, "preemption imbalance ");
1273 		preempt_count_set(count);
1274 	}
1275 	if (irqs_disabled()) {
1276 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1277 		local_irq_enable();
1278 	}
1279 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1280 
1281 	add_latent_entropy();
1282 	return ret;
1283 }
1284 
1285 
1286 static initcall_entry_t *initcall_levels[] __initdata = {
1287 	__initcall0_start,
1288 	__initcall1_start,
1289 	__initcall2_start,
1290 	__initcall3_start,
1291 	__initcall4_start,
1292 	__initcall5_start,
1293 	__initcall6_start,
1294 	__initcall7_start,
1295 	__initcall_end,
1296 };
1297 
1298 /* Keep these in sync with initcalls in include/linux/init.h */
1299 static const char *initcall_level_names[] __initdata = {
1300 	"pure",
1301 	"core",
1302 	"postcore",
1303 	"arch",
1304 	"subsys",
1305 	"fs",
1306 	"device",
1307 	"late",
1308 };
1309 
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1310 static int __init ignore_unknown_bootoption(char *param, char *val,
1311 			       const char *unused, void *arg)
1312 {
1313 	return 0;
1314 }
1315 
do_initcall_level(int level,char * command_line)1316 static void __init do_initcall_level(int level, char *command_line)
1317 {
1318 	initcall_entry_t *fn;
1319 
1320 	parse_args(initcall_level_names[level],
1321 		   command_line, __start___param,
1322 		   __stop___param - __start___param,
1323 		   level, level,
1324 		   NULL, ignore_unknown_bootoption);
1325 
1326 	trace_initcall_level(initcall_level_names[level]);
1327 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1328 		do_one_initcall(initcall_from_entry(fn));
1329 }
1330 
do_initcalls(void)1331 static void __init do_initcalls(void)
1332 {
1333 	int level;
1334 	size_t len = saved_command_line_len + 1;
1335 	char *command_line;
1336 
1337 	command_line = kzalloc(len, GFP_KERNEL);
1338 	if (!command_line)
1339 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1340 
1341 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1342 		/* Parser modifies command_line, restore it each time */
1343 		strcpy(command_line, saved_command_line);
1344 		do_initcall_level(level, command_line);
1345 	}
1346 
1347 	kfree(command_line);
1348 }
1349 
1350 /*
1351  * Ok, the machine is now initialized. None of the devices
1352  * have been touched yet, but the CPU subsystem is up and
1353  * running, and memory and process management works.
1354  *
1355  * Now we can finally start doing some real work..
1356  */
do_basic_setup(void)1357 static void __init do_basic_setup(void)
1358 {
1359 	cpuset_init_smp();
1360 	driver_init();
1361 	init_irq_proc();
1362 	do_ctors();
1363 	do_initcalls();
1364 }
1365 
do_pre_smp_initcalls(void)1366 static void __init do_pre_smp_initcalls(void)
1367 {
1368 	initcall_entry_t *fn;
1369 
1370 	trace_initcall_level("early");
1371 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1372 		do_one_initcall(initcall_from_entry(fn));
1373 }
1374 
run_init_process(const char * init_filename)1375 static int run_init_process(const char *init_filename)
1376 {
1377 	const char *const *p;
1378 
1379 	argv_init[0] = init_filename;
1380 	pr_info("Run %s as init process\n", init_filename);
1381 	pr_debug("  with arguments:\n");
1382 	for (p = argv_init; *p; p++)
1383 		pr_debug("    %s\n", *p);
1384 	pr_debug("  with environment:\n");
1385 	for (p = envp_init; *p; p++)
1386 		pr_debug("    %s\n", *p);
1387 	return kernel_execve(init_filename, argv_init, envp_init);
1388 }
1389 
try_to_run_init_process(const char * init_filename)1390 static int try_to_run_init_process(const char *init_filename)
1391 {
1392 	int ret;
1393 
1394 	ret = run_init_process(init_filename);
1395 
1396 	if (ret && ret != -ENOENT) {
1397 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1398 		       init_filename, ret);
1399 	}
1400 
1401 	return ret;
1402 }
1403 
1404 static noinline void __init kernel_init_freeable(void);
1405 
1406 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1407 bool rodata_enabled __ro_after_init = true;
1408 
1409 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1410 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1411 #endif
1412 
set_debug_rodata(char * str)1413 static int __init set_debug_rodata(char *str)
1414 {
1415 	if (arch_parse_debug_rodata(str))
1416 		return 0;
1417 
1418 	if (str && !strcmp(str, "on"))
1419 		rodata_enabled = true;
1420 	else if (str && !strcmp(str, "off"))
1421 		rodata_enabled = false;
1422 	else
1423 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1424 	return 0;
1425 }
1426 early_param("rodata", set_debug_rodata);
1427 #endif
1428 
mark_readonly(void)1429 static void mark_readonly(void)
1430 {
1431 	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1432 		/*
1433 		 * load_module() results in W+X mappings, which are cleaned
1434 		 * up with init_free_wq. Let's make sure that queued work is
1435 		 * flushed so that we don't hit false positives looking for
1436 		 * insecure pages which are W+X.
1437 		 */
1438 		flush_module_init_free_work();
1439 		jump_label_init_ro();
1440 		mark_rodata_ro();
1441 		debug_checkwx();
1442 		rodata_test();
1443 	} else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1444 		pr_info("Kernel memory protection disabled.\n");
1445 	} else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1446 		pr_warn("Kernel memory protection not selected by kernel config.\n");
1447 	} else {
1448 		pr_warn("This architecture does not have kernel memory protection.\n");
1449 	}
1450 }
1451 
free_initmem(void)1452 void __weak free_initmem(void)
1453 {
1454 	free_initmem_default(POISON_FREE_INITMEM);
1455 }
1456 
kernel_init(void * unused)1457 static int __ref kernel_init(void *unused)
1458 {
1459 	int ret;
1460 
1461 	/*
1462 	 * Wait until kthreadd is all set-up.
1463 	 */
1464 	wait_for_completion(&kthreadd_done);
1465 
1466 	kernel_init_freeable();
1467 	/* need to finish all async __init code before freeing the memory */
1468 	async_synchronize_full();
1469 
1470 	system_state = SYSTEM_FREEING_INITMEM;
1471 	kprobe_free_init_mem();
1472 	ftrace_free_init_mem();
1473 	kgdb_free_init_mem();
1474 	exit_boot_config();
1475 	free_initmem();
1476 	mark_readonly();
1477 
1478 	/*
1479 	 * Kernel mappings are now finalized - update the userspace page-table
1480 	 * to finalize PTI.
1481 	 */
1482 	pti_finalize();
1483 
1484 	system_state = SYSTEM_RUNNING;
1485 	numa_default_policy();
1486 
1487 	rcu_end_inkernel_boot();
1488 
1489 	do_sysctl_args();
1490 
1491 	if (ramdisk_execute_command) {
1492 		ret = run_init_process(ramdisk_execute_command);
1493 		if (!ret)
1494 			return 0;
1495 		pr_err("Failed to execute %s (error %d)\n",
1496 		       ramdisk_execute_command, ret);
1497 	}
1498 
1499 	/*
1500 	 * We try each of these until one succeeds.
1501 	 *
1502 	 * The Bourne shell can be used instead of init if we are
1503 	 * trying to recover a really broken machine.
1504 	 */
1505 	if (execute_command) {
1506 		ret = run_init_process(execute_command);
1507 		if (!ret)
1508 			return 0;
1509 		panic("Requested init %s failed (error %d).",
1510 		      execute_command, ret);
1511 	}
1512 
1513 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1514 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1515 		if (ret)
1516 			pr_err("Default init %s failed (error %d)\n",
1517 			       CONFIG_DEFAULT_INIT, ret);
1518 		else
1519 			return 0;
1520 	}
1521 
1522 	if (!try_to_run_init_process("/sbin/init") ||
1523 	    !try_to_run_init_process("/etc/init") ||
1524 	    !try_to_run_init_process("/bin/init") ||
1525 	    !try_to_run_init_process("/bin/sh"))
1526 		return 0;
1527 
1528 	panic("No working init found.  Try passing init= option to kernel. "
1529 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1530 }
1531 
1532 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1533 void __init console_on_rootfs(void)
1534 {
1535 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1536 
1537 	if (IS_ERR(file)) {
1538 		pr_err("Warning: unable to open an initial console.\n");
1539 		return;
1540 	}
1541 	init_dup(file);
1542 	init_dup(file);
1543 	init_dup(file);
1544 	fput(file);
1545 }
1546 
kernel_init_freeable(void)1547 static noinline void __init kernel_init_freeable(void)
1548 {
1549 	/* Now the scheduler is fully set up and can do blocking allocations */
1550 	gfp_allowed_mask = __GFP_BITS_MASK;
1551 
1552 	/*
1553 	 * init can allocate pages on any node
1554 	 */
1555 	set_mems_allowed(node_states[N_MEMORY]);
1556 
1557 	cad_pid = get_pid(task_pid(current));
1558 
1559 	smp_prepare_cpus(setup_max_cpus);
1560 
1561 	workqueue_init();
1562 
1563 	init_mm_internals();
1564 
1565 	rcu_init_tasks_generic();
1566 	do_pre_smp_initcalls();
1567 	lockup_detector_init();
1568 
1569 	smp_init();
1570 	sched_init_smp();
1571 
1572 	workqueue_init_topology();
1573 	async_init();
1574 	padata_init();
1575 	page_alloc_init_late();
1576 
1577 	do_basic_setup();
1578 
1579 	kunit_run_all_tests();
1580 
1581 	wait_for_initramfs();
1582 	console_on_rootfs();
1583 
1584 	/*
1585 	 * check if there is an early userspace init.  If yes, let it do all
1586 	 * the work
1587 	 */
1588 	if (init_eaccess(ramdisk_execute_command) != 0) {
1589 		ramdisk_execute_command = NULL;
1590 		prepare_namespace();
1591 	}
1592 
1593 	/*
1594 	 * Ok, we have completed the initial bootup, and
1595 	 * we're essentially up and running. Get rid of the
1596 	 * initmem segments and start the user-mode stuff..
1597 	 *
1598 	 * rootfs is available now, try loading the public keys
1599 	 * and default modules
1600 	 */
1601 
1602 	integrity_load_keys();
1603 }
1604