xref: /linux/init/main.c (revision 4793dae01f47754e288cdbb3a22581cac2317f2b)
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/export.h>
17 #include <linux/extable.h>
18 #include <linux/module.h>
19 #include <linux/proc_fs.h>
20 #include <linux/binfmts.h>
21 #include <linux/kernel.h>
22 #include <linux/syscalls.h>
23 #include <linux/stackprotector.h>
24 #include <linux/string.h>
25 #include <linux/ctype.h>
26 #include <linux/delay.h>
27 #include <linux/ioport.h>
28 #include <linux/init.h>
29 #include <linux/initrd.h>
30 #include <linux/memblock.h>
31 #include <linux/acpi.h>
32 #include <linux/bootconfig.h>
33 #include <linux/console.h>
34 #include <linux/nmi.h>
35 #include <linux/percpu.h>
36 #include <linux/kmod.h>
37 #include <linux/kprobes.h>
38 #include <linux/kmsan.h>
39 #include <linux/ksysfs.h>
40 #include <linux/vmalloc.h>
41 #include <linux/kernel_stat.h>
42 #include <linux/start_kernel.h>
43 #include <linux/security.h>
44 #include <linux/smp.h>
45 #include <linux/profile.h>
46 #include <linux/kfence.h>
47 #include <linux/rcupdate.h>
48 #include <linux/srcu.h>
49 #include <linux/moduleparam.h>
50 #include <linux/kallsyms.h>
51 #include <linux/buildid.h>
52 #include <linux/writeback.h>
53 #include <linux/cpu.h>
54 #include <linux/cpuset.h>
55 #include <linux/memcontrol.h>
56 #include <linux/cgroup.h>
57 #include <linux/tick.h>
58 #include <linux/sched/isolation.h>
59 #include <linux/interrupt.h>
60 #include <linux/taskstats_kern.h>
61 #include <linux/delayacct.h>
62 #include <linux/unistd.h>
63 #include <linux/utsname.h>
64 #include <linux/rmap.h>
65 #include <linux/mempolicy.h>
66 #include <linux/key.h>
67 #include <linux/debug_locks.h>
68 #include <linux/debugobjects.h>
69 #include <linux/lockdep.h>
70 #include <linux/kmemleak.h>
71 #include <linux/padata.h>
72 #include <linux/pid_namespace.h>
73 #include <linux/device/driver.h>
74 #include <linux/kthread.h>
75 #include <linux/sched.h>
76 #include <linux/sched/init.h>
77 #include <linux/signal.h>
78 #include <linux/idr.h>
79 #include <linux/kgdb.h>
80 #include <linux/ftrace.h>
81 #include <linux/async.h>
82 #include <linux/shmem_fs.h>
83 #include <linux/slab.h>
84 #include <linux/perf_event.h>
85 #include <linux/ptrace.h>
86 #include <linux/pti.h>
87 #include <linux/blkdev.h>
88 #include <linux/sched/clock.h>
89 #include <linux/sched/task.h>
90 #include <linux/sched/task_stack.h>
91 #include <linux/context_tracking.h>
92 #include <linux/random.h>
93 #include <linux/moduleloader.h>
94 #include <linux/list.h>
95 #include <linux/integrity.h>
96 #include <linux/proc_ns.h>
97 #include <linux/io.h>
98 #include <linux/cache.h>
99 #include <linux/rodata_test.h>
100 #include <linux/jump_label.h>
101 #include <linux/kcsan.h>
102 #include <linux/init_syscalls.h>
103 #include <linux/stackdepot.h>
104 #include <linux/randomize_kstack.h>
105 #include <linux/pidfs.h>
106 #include <linux/ptdump.h>
107 #include <linux/time_namespace.h>
108 #include <linux/unaligned.h>
109 #include <net/net_namespace.h>
110 
111 #include <asm/io.h>
112 #include <asm/setup.h>
113 #include <asm/sections.h>
114 #include <asm/cacheflush.h>
115 
116 #define CREATE_TRACE_POINTS
117 #include <trace/events/initcall.h>
118 
119 #include <kunit/test.h>
120 
121 static int kernel_init(void *);
122 
123 /*
124  * Debug helper: via this flag we know that we are in 'early bootup code'
125  * where only the boot processor is running with IRQ disabled.  This means
126  * two things - IRQ must not be enabled before the flag is cleared and some
127  * operations which are not allowed with IRQ disabled are allowed while the
128  * flag is set.
129  */
130 bool early_boot_irqs_disabled __read_mostly;
131 
132 enum system_states system_state __read_mostly;
133 EXPORT_SYMBOL(system_state);
134 
135 /*
136  * Boot command-line arguments
137  */
138 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
139 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
140 
141 /* Default late time init is NULL. archs can override this later. */
142 void (*__initdata late_time_init)(void);
143 
144 /* Untouched command line saved by arch-specific code. */
145 char __initdata boot_command_line[COMMAND_LINE_SIZE];
146 /* Untouched saved command line (eg. for /proc) */
147 char *saved_command_line __ro_after_init;
148 unsigned int saved_command_line_len __ro_after_init;
149 /* Command line for parameter parsing */
150 static char *static_command_line;
151 /* Untouched extra command line */
152 static char *extra_command_line;
153 /* Extra init arguments */
154 static char *extra_init_args;
155 
156 #ifdef CONFIG_BOOT_CONFIG
157 /* Is bootconfig on command line? */
158 static bool bootconfig_found;
159 static size_t initargs_offs;
160 #else
161 # define bootconfig_found false
162 # define initargs_offs 0
163 #endif
164 
165 static char *execute_command;
166 static char *ramdisk_execute_command = "/init";
167 static bool __initdata ramdisk_execute_command_set;
168 
169 /*
170  * Used to generate warnings if static_key manipulation functions are used
171  * before jump_label_init is called.
172  */
173 bool static_key_initialized __read_mostly;
174 EXPORT_SYMBOL_GPL(static_key_initialized);
175 
176 /*
177  * If set, this is an indication to the drivers that reset the underlying
178  * device before going ahead with the initialization otherwise driver might
179  * rely on the BIOS and skip the reset operation.
180  *
181  * This is useful if kernel is booting in an unreliable environment.
182  * For ex. kdump situation where previous kernel has crashed, BIOS has been
183  * skipped and devices will be in unknown state.
184  */
185 unsigned int reset_devices;
186 EXPORT_SYMBOL(reset_devices);
187 
set_reset_devices(char * str)188 static int __init set_reset_devices(char *str)
189 {
190 	reset_devices = 1;
191 	return 1;
192 }
193 
194 __setup("reset_devices", set_reset_devices);
195 
196 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
197 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
198 static const char *panic_later, *panic_param;
199 
obsolete_checksetup(char * line)200 static bool __init obsolete_checksetup(char *line)
201 {
202 	const struct obs_kernel_param *p;
203 	bool had_early_param = false;
204 
205 	p = __setup_start;
206 	do {
207 		int n = strlen(p->str);
208 		if (parameqn(line, p->str, n)) {
209 			if (p->early) {
210 				/* Already done in parse_early_param?
211 				 * (Needs exact match on param part).
212 				 * Keep iterating, as we can have early
213 				 * params and __setups of same names 8( */
214 				if (line[n] == '\0' || line[n] == '=')
215 					had_early_param = true;
216 			} else if (!p->setup_func) {
217 				pr_warn("Parameter %s is obsolete, ignored\n",
218 					p->str);
219 				return true;
220 			} else if (p->setup_func(line + n))
221 				return true;
222 		}
223 		p++;
224 	} while (p < __setup_end);
225 
226 	return had_early_param;
227 }
228 
229 /*
230  * This should be approx 2 Bo*oMips to start (note initial shift), and will
231  * still work even if initially too large, it will just take slightly longer
232  */
233 unsigned long loops_per_jiffy = (1<<12);
234 EXPORT_SYMBOL(loops_per_jiffy);
235 
debug_kernel(char * str)236 static int __init debug_kernel(char *str)
237 {
238 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
239 	return 0;
240 }
241 
quiet_kernel(char * str)242 static int __init quiet_kernel(char *str)
243 {
244 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
245 	return 0;
246 }
247 
248 early_param("debug", debug_kernel);
249 early_param("quiet", quiet_kernel);
250 
loglevel(char * str)251 static int __init loglevel(char *str)
252 {
253 	int newlevel;
254 
255 	/*
256 	 * Only update loglevel value when a correct setting was passed,
257 	 * to prevent blind crashes (when loglevel being set to 0) that
258 	 * are quite hard to debug
259 	 */
260 	if (get_option(&str, &newlevel)) {
261 		console_loglevel = newlevel;
262 		return 0;
263 	}
264 
265 	return -EINVAL;
266 }
267 
268 early_param("loglevel", loglevel);
269 
270 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(size_t * _size)271 static void * __init get_boot_config_from_initrd(size_t *_size)
272 {
273 	u32 size, csum;
274 	char *data;
275 	u8 *hdr;
276 	int i;
277 
278 	if (!initrd_end)
279 		return NULL;
280 
281 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
282 	/*
283 	 * Since Grub may align the size of initrd to 4, we must
284 	 * check the preceding 3 bytes as well.
285 	 */
286 	for (i = 0; i < 4; i++) {
287 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
288 			goto found;
289 		data--;
290 	}
291 	return NULL;
292 
293 found:
294 	hdr = (u8 *)(data - 8);
295 	size = get_unaligned_le32(hdr);
296 	csum = get_unaligned_le32(hdr + 4);
297 
298 	data = ((void *)hdr) - size;
299 	if ((unsigned long)data < initrd_start) {
300 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
301 			size, initrd_end - initrd_start);
302 		return NULL;
303 	}
304 
305 	if (xbc_calc_checksum(data, size) != csum) {
306 		pr_err("bootconfig checksum failed\n");
307 		return NULL;
308 	}
309 
310 	/* Remove bootconfig from initramfs/initrd */
311 	initrd_end = (unsigned long)data;
312 	if (_size)
313 		*_size = size;
314 
315 	return data;
316 }
317 #else
get_boot_config_from_initrd(size_t * _size)318 static void * __init get_boot_config_from_initrd(size_t *_size)
319 {
320 	return NULL;
321 }
322 #endif
323 
324 #ifdef CONFIG_BOOT_CONFIG
325 
326 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
327 
328 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
329 
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)330 static int __init xbc_snprint_cmdline(char *buf, size_t size,
331 				      struct xbc_node *root)
332 {
333 	struct xbc_node *knode, *vnode;
334 	char *end = buf + size;
335 	const char *val, *q;
336 	int ret;
337 
338 	xbc_node_for_each_key_value(root, knode, val) {
339 		ret = xbc_node_compose_key_after(root, knode,
340 					xbc_namebuf, XBC_KEYLEN_MAX);
341 		if (ret < 0)
342 			return ret;
343 
344 		vnode = xbc_node_get_child(knode);
345 		if (!vnode) {
346 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
347 			if (ret < 0)
348 				return ret;
349 			buf += ret;
350 			continue;
351 		}
352 		xbc_array_for_each_value(vnode, val) {
353 			/*
354 			 * For prettier and more readable /proc/cmdline, only
355 			 * quote the value when necessary, i.e. when it contains
356 			 * whitespace.
357 			 */
358 			q = strpbrk(val, " \t\r\n") ? "\"" : "";
359 			ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
360 				       xbc_namebuf, q, val, q);
361 			if (ret < 0)
362 				return ret;
363 			buf += ret;
364 		}
365 	}
366 
367 	return buf - (end - size);
368 }
369 #undef rest
370 
371 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)372 static char * __init xbc_make_cmdline(const char *key)
373 {
374 	struct xbc_node *root;
375 	char *new_cmdline;
376 	int ret, len = 0;
377 
378 	root = xbc_find_node(key);
379 	if (!root)
380 		return NULL;
381 
382 	/* Count required buffer size */
383 	len = xbc_snprint_cmdline(NULL, 0, root);
384 	if (len <= 0)
385 		return NULL;
386 
387 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
388 	if (!new_cmdline) {
389 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
390 		return NULL;
391 	}
392 
393 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
394 	if (ret < 0 || ret > len) {
395 		pr_err("Failed to print extra kernel cmdline.\n");
396 		memblock_free(new_cmdline, len + 1);
397 		return NULL;
398 	}
399 
400 	return new_cmdline;
401 }
402 
bootconfig_params(char * param,char * val,const char * unused,void * arg)403 static int __init bootconfig_params(char *param, char *val,
404 				    const char *unused, void *arg)
405 {
406 	if (strcmp(param, "bootconfig") == 0) {
407 		bootconfig_found = true;
408 	}
409 	return 0;
410 }
411 
warn_bootconfig(char * str)412 static int __init warn_bootconfig(char *str)
413 {
414 	/* The 'bootconfig' has been handled by bootconfig_params(). */
415 	return 0;
416 }
417 
setup_boot_config(void)418 static void __init setup_boot_config(void)
419 {
420 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
421 	const char *msg, *data;
422 	int pos, ret;
423 	size_t size;
424 	char *err;
425 
426 	/* Cut out the bootconfig data even if we have no bootconfig option */
427 	data = get_boot_config_from_initrd(&size);
428 	/* If there is no bootconfig in initrd, try embedded one. */
429 	if (!data)
430 		data = xbc_get_embedded_bootconfig(&size);
431 
432 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
433 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
434 			 bootconfig_params);
435 
436 	if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
437 		return;
438 
439 	/* parse_args() stops at the next param of '--' and returns an address */
440 	if (err)
441 		initargs_offs = err - tmp_cmdline;
442 
443 	if (!data) {
444 		/* If user intended to use bootconfig, show an error level message */
445 		if (bootconfig_found)
446 			pr_err("'bootconfig' found on command line, but no bootconfig found\n");
447 		else
448 			pr_info("No bootconfig data provided, so skipping bootconfig");
449 		return;
450 	}
451 
452 	if (size >= XBC_DATA_MAX) {
453 		pr_err("bootconfig size %ld greater than max size %d\n",
454 			(long)size, XBC_DATA_MAX);
455 		return;
456 	}
457 
458 	ret = xbc_init(data, size, &msg, &pos);
459 	if (ret < 0) {
460 		if (pos < 0)
461 			pr_err("Failed to init bootconfig: %s.\n", msg);
462 		else
463 			pr_err("Failed to parse bootconfig: %s at %d.\n",
464 				msg, pos);
465 	} else {
466 		xbc_get_info(&ret, NULL);
467 		pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
468 		/* keys starting with "kernel." are passed via cmdline */
469 		extra_command_line = xbc_make_cmdline("kernel");
470 		/* Also, "init." keys are init arguments */
471 		extra_init_args = xbc_make_cmdline("init");
472 	}
473 	return;
474 }
475 
exit_boot_config(void)476 static void __init exit_boot_config(void)
477 {
478 	xbc_exit();
479 }
480 
481 #else	/* !CONFIG_BOOT_CONFIG */
482 
setup_boot_config(void)483 static void __init setup_boot_config(void)
484 {
485 	/* Remove bootconfig data from initrd */
486 	get_boot_config_from_initrd(NULL);
487 }
488 
warn_bootconfig(char * str)489 static int __init warn_bootconfig(char *str)
490 {
491 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
492 	return 0;
493 }
494 
495 #define exit_boot_config()	do {} while (0)
496 
497 #endif	/* CONFIG_BOOT_CONFIG */
498 
499 early_param("bootconfig", warn_bootconfig);
500 
cmdline_has_extra_options(void)501 bool __init cmdline_has_extra_options(void)
502 {
503 	return extra_command_line || extra_init_args;
504 }
505 
506 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)507 static void __init repair_env_string(char *param, char *val)
508 {
509 	if (val) {
510 		/* param=val or param="val"? */
511 		if (val == param+strlen(param)+1)
512 			val[-1] = '=';
513 		else if (val == param+strlen(param)+2) {
514 			val[-2] = '=';
515 			memmove(val-1, val, strlen(val)+1);
516 		} else
517 			BUG();
518 	}
519 }
520 
521 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)522 static int __init set_init_arg(char *param, char *val,
523 			       const char *unused, void *arg)
524 {
525 	unsigned int i;
526 
527 	if (panic_later)
528 		return 0;
529 
530 	repair_env_string(param, val);
531 
532 	for (i = 0; argv_init[i]; i++) {
533 		if (i == MAX_INIT_ARGS) {
534 			panic_later = "init";
535 			panic_param = param;
536 			return 0;
537 		}
538 	}
539 	argv_init[i] = param;
540 	return 0;
541 }
542 
543 /*
544  * Unknown boot options get handed to init, unless they look like
545  * unused parameters (modprobe will find them in /proc/cmdline).
546  */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)547 static int __init unknown_bootoption(char *param, char *val,
548 				     const char *unused, void *arg)
549 {
550 	size_t len = strlen(param);
551 	/*
552 	 * Well-known bootloader identifiers:
553 	 * 1. LILO/Grub pass "BOOT_IMAGE=...";
554 	 * 2. kexec/kdump (kexec-tools) pass "kexec".
555 	 */
556 	const char *bootloader[] = { "BOOT_IMAGE=", "kexec", NULL };
557 
558 	/* Handle params aliased to sysctls */
559 	if (sysctl_is_alias(param))
560 		return 0;
561 
562 	repair_env_string(param, val);
563 
564 	/* Handle bootloader identifier */
565 	for (int i = 0; bootloader[i]; i++) {
566 		if (strstarts(param, bootloader[i]))
567 			return 0;
568 	}
569 
570 	/* Handle obsolete-style parameters */
571 	if (obsolete_checksetup(param))
572 		return 0;
573 
574 	/* Unused module parameter. */
575 	if (strnchr(param, len, '.'))
576 		return 0;
577 
578 	if (panic_later)
579 		return 0;
580 
581 	if (val) {
582 		/* Environment option */
583 		unsigned int i;
584 		for (i = 0; envp_init[i]; i++) {
585 			if (i == MAX_INIT_ENVS) {
586 				panic_later = "env";
587 				panic_param = param;
588 			}
589 			if (!strncmp(param, envp_init[i], len+1))
590 				break;
591 		}
592 		envp_init[i] = param;
593 	} else {
594 		/* Command line option */
595 		unsigned int i;
596 		for (i = 0; argv_init[i]; i++) {
597 			if (i == MAX_INIT_ARGS) {
598 				panic_later = "init";
599 				panic_param = param;
600 			}
601 		}
602 		argv_init[i] = param;
603 	}
604 	return 0;
605 }
606 
init_setup(char * str)607 static int __init init_setup(char *str)
608 {
609 	unsigned int i;
610 
611 	execute_command = str;
612 	/*
613 	 * In case LILO is going to boot us with default command line,
614 	 * it prepends "auto" before the whole cmdline which makes
615 	 * the shell think it should execute a script with such name.
616 	 * So we ignore all arguments entered _before_ init=... [MJ]
617 	 */
618 	for (i = 1; i < MAX_INIT_ARGS; i++)
619 		argv_init[i] = NULL;
620 	return 1;
621 }
622 __setup("init=", init_setup);
623 
rdinit_setup(char * str)624 static int __init rdinit_setup(char *str)
625 {
626 	unsigned int i;
627 
628 	ramdisk_execute_command = str;
629 	ramdisk_execute_command_set = true;
630 	/* See "auto" comment in init_setup */
631 	for (i = 1; i < MAX_INIT_ARGS; i++)
632 		argv_init[i] = NULL;
633 	return 1;
634 }
635 __setup("rdinit=", rdinit_setup);
636 
637 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)638 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)639 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
640 #endif
641 
642 /*
643  * We need to store the untouched command line for future reference.
644  * We also need to store the touched command line since the parameter
645  * parsing is performed in place, and we should allow a component to
646  * store reference of name/value for future reference.
647  */
setup_command_line(char * command_line)648 static void __init setup_command_line(char *command_line)
649 {
650 	size_t len, xlen = 0, ilen = 0;
651 
652 	if (extra_command_line)
653 		xlen = strlen(extra_command_line);
654 	if (extra_init_args) {
655 		extra_init_args = strim(extra_init_args); /* remove trailing space */
656 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
657 	}
658 
659 	len = xlen + strlen(boot_command_line) + ilen + 1;
660 
661 	saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
662 
663 	len = xlen + strlen(command_line) + 1;
664 
665 	static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
666 
667 	if (xlen) {
668 		/*
669 		 * We have to put extra_command_line before boot command
670 		 * lines because there could be dashes (separator of init
671 		 * command line) in the command lines.
672 		 */
673 		strcpy(saved_command_line, extra_command_line);
674 		strcpy(static_command_line, extra_command_line);
675 	}
676 	strcpy(saved_command_line + xlen, boot_command_line);
677 	strcpy(static_command_line + xlen, command_line);
678 
679 	if (ilen) {
680 		/*
681 		 * Append supplemental init boot args to saved_command_line
682 		 * so that user can check what command line options passed
683 		 * to init.
684 		 * The order should always be
685 		 * " -- "[bootconfig init-param][cmdline init-param]
686 		 */
687 		if (initargs_offs) {
688 			len = xlen + initargs_offs;
689 			strcpy(saved_command_line + len, extra_init_args);
690 			len += ilen - 4;	/* strlen(extra_init_args) */
691 			strcpy(saved_command_line + len,
692 				boot_command_line + initargs_offs - 1);
693 		} else {
694 			len = strlen(saved_command_line);
695 			strcpy(saved_command_line + len, " -- ");
696 			len += 4;
697 			strcpy(saved_command_line + len, extra_init_args);
698 		}
699 	}
700 
701 	saved_command_line_len = strlen(saved_command_line);
702 }
703 
704 /*
705  * We need to finalize in a non-__init function or else race conditions
706  * between the root thread and the init thread may cause start_kernel to
707  * be reaped by free_initmem before the root thread has proceeded to
708  * cpu_idle.
709  *
710  * gcc-3.4 accidentally inlines this function, so use noinline.
711  */
712 
713 static __initdata DECLARE_COMPLETION(kthreadd_done);
714 
rest_init(void)715 static noinline void __ref __noreturn rest_init(void)
716 {
717 	struct task_struct *tsk;
718 	int pid;
719 
720 	rcu_scheduler_starting();
721 	/*
722 	 * We need to spawn init first so that it obtains pid 1, however
723 	 * the init task will end up wanting to create kthreads, which, if
724 	 * we schedule it before we create kthreadd, will OOPS.
725 	 */
726 	pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
727 	/*
728 	 * Pin init on the boot CPU. Task migration is not properly working
729 	 * until sched_init_smp() has been run. It will set the allowed
730 	 * CPUs for init to the non isolated CPUs.
731 	 */
732 	rcu_read_lock();
733 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
734 	tsk->flags |= PF_NO_SETAFFINITY;
735 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
736 	rcu_read_unlock();
737 
738 	numa_default_policy();
739 	pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
740 	rcu_read_lock();
741 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
742 	rcu_read_unlock();
743 
744 	/*
745 	 * Enable might_sleep() and smp_processor_id() checks.
746 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
747 	 * kernel_thread() would trigger might_sleep() splats. With
748 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
749 	 * already, but it's stuck on the kthreadd_done completion.
750 	 */
751 	system_state = SYSTEM_SCHEDULING;
752 
753 	complete(&kthreadd_done);
754 
755 	/*
756 	 * The boot idle thread must execute schedule()
757 	 * at least once to get things moving:
758 	 */
759 	schedule_preempt_disabled();
760 	/* Call into cpu_idle with preempt disabled */
761 	cpu_startup_entry(CPUHP_ONLINE);
762 }
763 
764 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)765 static int __init do_early_param(char *param, char *val,
766 				 const char *unused, void *arg)
767 {
768 	const struct obs_kernel_param *p;
769 
770 	for (p = __setup_start; p < __setup_end; p++) {
771 		if (p->early && parameq(param, p->str)) {
772 			if (p->setup_func(val) != 0)
773 				pr_warn("Malformed early option '%s'\n", param);
774 		}
775 	}
776 	/* We accept everything at this stage. */
777 	return 0;
778 }
779 
parse_early_options(char * cmdline)780 void __init parse_early_options(char *cmdline)
781 {
782 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
783 		   do_early_param);
784 }
785 
786 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)787 void __init parse_early_param(void)
788 {
789 	static int done __initdata;
790 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
791 
792 	if (done)
793 		return;
794 
795 	/* All fall through to do_early_param. */
796 	strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
797 	parse_early_options(tmp_cmdline);
798 	done = 1;
799 }
800 
arch_post_acpi_subsys_init(void)801 void __init __weak arch_post_acpi_subsys_init(void) { }
802 
smp_setup_processor_id(void)803 void __init __weak smp_setup_processor_id(void)
804 {
805 }
806 
smp_prepare_boot_cpu(void)807 void __init __weak smp_prepare_boot_cpu(void)
808 {
809 }
810 
811 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)812 void __init __weak thread_stack_cache_init(void)
813 {
814 }
815 #endif
816 
poking_init(void)817 void __init __weak poking_init(void) { }
818 
pgtable_cache_init(void)819 void __init __weak pgtable_cache_init(void) { }
820 
trap_init(void)821 void __init __weak trap_init(void) { }
822 
823 bool initcall_debug;
824 core_param(initcall_debug, initcall_debug, bool, 0644);
825 
826 #ifdef TRACEPOINTS_ENABLED
827 static void __init initcall_debug_enable(void);
828 #else
initcall_debug_enable(void)829 static inline void initcall_debug_enable(void)
830 {
831 }
832 #endif
833 
834 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
835 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
836 			   randomize_kstack_offset);
837 DEFINE_PER_CPU(struct rnd_state, kstack_rnd_state);
838 
random_kstack_init(void)839 static int __init random_kstack_init(void)
840 {
841 	prandom_seed_full_state(&kstack_rnd_state);
842 	return 0;
843 }
844 late_initcall(random_kstack_init);
845 
early_randomize_kstack_offset(char * buf)846 static int __init early_randomize_kstack_offset(char *buf)
847 {
848 	int ret;
849 	bool bool_result;
850 
851 	ret = kstrtobool(buf, &bool_result);
852 	if (ret)
853 		return ret;
854 
855 	if (bool_result)
856 		static_branch_enable(&randomize_kstack_offset);
857 	else
858 		static_branch_disable(&randomize_kstack_offset);
859 	return 0;
860 }
861 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
862 #endif
863 
print_unknown_bootoptions(void)864 static void __init print_unknown_bootoptions(void)
865 {
866 	char *unknown_options;
867 	char *end;
868 	const char *const *p;
869 	size_t len;
870 
871 	if (panic_later || (!argv_init[1] && !envp_init[2]))
872 		return;
873 
874 	/*
875 	 * Determine how many options we have to print out, plus a space
876 	 * before each
877 	 */
878 	len = 1; /* null terminator */
879 	for (p = &argv_init[1]; *p; p++) {
880 		len++;
881 		len += strlen(*p);
882 	}
883 	for (p = &envp_init[2]; *p; p++) {
884 		len++;
885 		len += strlen(*p);
886 	}
887 
888 	unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
889 	if (!unknown_options) {
890 		pr_err("%s: Failed to allocate %zu bytes\n",
891 			__func__, len);
892 		return;
893 	}
894 	end = unknown_options;
895 
896 	for (p = &argv_init[1]; *p; p++)
897 		end += sprintf(end, " %s", *p);
898 	for (p = &envp_init[2]; *p; p++)
899 		end += sprintf(end, " %s", *p);
900 
901 	/* Start at unknown_options[1] to skip the initial space */
902 	pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
903 		&unknown_options[1]);
904 	memblock_free(unknown_options, len);
905 }
906 
early_numa_node_init(void)907 static void __init early_numa_node_init(void)
908 {
909 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
910 #ifndef cpu_to_node
911 	int cpu;
912 
913 	/* The early_cpu_to_node() should be ready here. */
914 	for_each_possible_cpu(cpu)
915 		set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
916 #endif
917 #endif
918 }
919 
920 #define KERNEL_CMDLINE_PREFIX		"Kernel command line: "
921 #define KERNEL_CMDLINE_PREFIX_LEN	(sizeof(KERNEL_CMDLINE_PREFIX) - 1)
922 #define KERNEL_CMDLINE_CONTINUATION	" \\"
923 #define KERNEL_CMDLINE_CONTINUATION_LEN	(sizeof(KERNEL_CMDLINE_CONTINUATION) - 1)
924 
925 #define MIN_CMDLINE_LOG_WRAP_IDEAL_LEN	(KERNEL_CMDLINE_PREFIX_LEN + \
926 					 KERNEL_CMDLINE_CONTINUATION_LEN)
927 #define CMDLINE_LOG_WRAP_IDEAL_LEN	(CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN > \
928 					 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN ? \
929 					 CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN : \
930 					 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN)
931 
932 #define IDEAL_CMDLINE_LEN		(CMDLINE_LOG_WRAP_IDEAL_LEN - KERNEL_CMDLINE_PREFIX_LEN)
933 #define IDEAL_CMDLINE_SPLIT_LEN		(IDEAL_CMDLINE_LEN - KERNEL_CMDLINE_CONTINUATION_LEN)
934 
935 /**
936  * print_kernel_cmdline() - Print the kernel cmdline with wrapping.
937  * @cmdline: The cmdline to print.
938  *
939  * Print the kernel command line, trying to wrap based on the Kconfig knob
940  * CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN.
941  *
942  * Wrapping is based on spaces, ignoring quotes. All lines are prefixed
943  * with "Kernel command line: " and lines that are not the last line have
944  * a " \" suffix added to them. The prefix and suffix count towards the
945  * line length for wrapping purposes. The ideal length will be exceeded
946  * if no appropriate place to wrap is found.
947  *
948  * Example output if CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN is 40:
949  *   Kernel command line: loglevel=7 \
950  *   Kernel command line: init=/sbin/init \
951  *   Kernel command line: root=PARTUUID=8c3efc1a-768b-6642-8d0c-89eb782f19f0/PARTNROFF=1 \
952  *   Kernel command line: rootwait ro \
953  *   Kernel command line: my_quoted_arg="The \
954  *   Kernel command line: quick brown fox \
955  *   Kernel command line: jumps over the \
956  *   Kernel command line: lazy dog."
957  */
print_kernel_cmdline(const char * cmdline)958 static void __init print_kernel_cmdline(const char *cmdline)
959 {
960 	size_t len;
961 
962 	/* Config option of 0 or anything longer than the max disables wrapping */
963 	if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN == 0 ||
964 	    IDEAL_CMDLINE_LEN >= COMMAND_LINE_SIZE - 1) {
965 		pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline);
966 		return;
967 	}
968 
969 	len = strlen(cmdline);
970 	while (len > IDEAL_CMDLINE_LEN) {
971 		const char *first_space;
972 		const char *prev_cutoff;
973 		const char *cutoff;
974 		int to_print;
975 		size_t used;
976 
977 		/* Find the last ' ' that wouldn't make the line too long */
978 		prev_cutoff = NULL;
979 		cutoff = cmdline;
980 		while (true) {
981 			cutoff = strchr(cutoff + 1, ' ');
982 			if (!cutoff || cutoff - cmdline > IDEAL_CMDLINE_SPLIT_LEN)
983 				break;
984 			prev_cutoff = cutoff;
985 		}
986 		if (prev_cutoff)
987 			cutoff = prev_cutoff;
988 		else if (!cutoff)
989 			break;
990 
991 		/* Find the beginning and end of the string of spaces */
992 		first_space = cutoff;
993 		while (first_space > cmdline && first_space[-1] == ' ')
994 			first_space--;
995 		to_print = first_space - cmdline;
996 		while (*cutoff == ' ')
997 			cutoff++;
998 		used = cutoff - cmdline;
999 
1000 		/* If the whole string is used, break and do the final printout */
1001 		if (len == used)
1002 			break;
1003 
1004 		if (to_print)
1005 			pr_notice("%s%.*s%s\n", KERNEL_CMDLINE_PREFIX,
1006 				  to_print, cmdline, KERNEL_CMDLINE_CONTINUATION);
1007 
1008 		len -= used;
1009 		cmdline += used;
1010 	}
1011 	if (len)
1012 		pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline);
1013 }
1014 
1015 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)1016 void start_kernel(void)
1017 {
1018 	char *command_line;
1019 	char *after_dashes;
1020 
1021 	set_task_stack_end_magic(&init_task);
1022 	smp_setup_processor_id();
1023 	debug_objects_early_init();
1024 	init_vmlinux_build_id();
1025 
1026 	cgroup_init_early();
1027 
1028 	local_irq_disable();
1029 	early_boot_irqs_disabled = true;
1030 
1031 	/*
1032 	 * Interrupts are still disabled. Do necessary setups, then
1033 	 * enable them.
1034 	 */
1035 	boot_cpu_init();
1036 	page_address_init();
1037 	pr_notice("%s", linux_banner);
1038 	setup_arch(&command_line);
1039 	mm_core_init_early();
1040 	/* Static keys and static calls are needed by LSMs */
1041 	jump_label_init();
1042 	static_call_init();
1043 	early_security_init();
1044 	setup_boot_config();
1045 	setup_command_line(command_line);
1046 	setup_nr_cpu_ids();
1047 	setup_per_cpu_areas();
1048 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
1049 	early_numa_node_init();
1050 	boot_cpu_hotplug_init();
1051 
1052 	print_kernel_cmdline(saved_command_line);
1053 	/* parameters may set static keys */
1054 	parse_early_param();
1055 	after_dashes = parse_args("Booting kernel",
1056 				  static_command_line, __start___param,
1057 				  __stop___param - __start___param,
1058 				  -1, -1, NULL, &unknown_bootoption);
1059 	print_unknown_bootoptions();
1060 	if (!IS_ERR_OR_NULL(after_dashes))
1061 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
1062 			   NULL, set_init_arg);
1063 	if (extra_init_args)
1064 		parse_args("Setting extra init args", extra_init_args,
1065 			   NULL, 0, -1, -1, NULL, set_init_arg);
1066 
1067 	/* Architectural and non-timekeeping rng init, before allocator init */
1068 	random_init_early(command_line);
1069 
1070 	/*
1071 	 * These use large bootmem allocations and must precede
1072 	 * initalization of page allocator
1073 	 */
1074 	setup_log_buf(0);
1075 	vfs_caches_init_early();
1076 	sort_main_extable();
1077 	trap_init();
1078 	mm_core_init();
1079 	maple_tree_init();
1080 	poking_init();
1081 	ftrace_init();
1082 
1083 	/* trace_printk can be enabled here */
1084 	early_trace_init();
1085 
1086 	/*
1087 	 * Set up the scheduler prior starting any interrupts (such as the
1088 	 * timer interrupt). Full topology setup happens at smp_init()
1089 	 * time - but meanwhile we still have a functioning scheduler.
1090 	 */
1091 	sched_init();
1092 
1093 	if (WARN(!irqs_disabled(),
1094 		 "Interrupts were enabled *very* early, fixing it\n"))
1095 		local_irq_disable();
1096 	radix_tree_init();
1097 
1098 	/*
1099 	 * Set up housekeeping before setting up workqueues to allow the unbound
1100 	 * workqueue to take non-housekeeping into account.
1101 	 */
1102 	housekeeping_init();
1103 
1104 	/*
1105 	 * Allow workqueue creation and work item queueing/cancelling
1106 	 * early.  Work item execution depends on kthreads and starts after
1107 	 * workqueue_init().
1108 	 */
1109 	workqueue_init_early();
1110 
1111 	rcu_init();
1112 	kvfree_rcu_init();
1113 
1114 	/* Trace events are available after this */
1115 	trace_init();
1116 
1117 	if (initcall_debug)
1118 		initcall_debug_enable();
1119 
1120 	context_tracking_init();
1121 	/* init some links before init_ISA_irqs() */
1122 	early_irq_init();
1123 	init_IRQ();
1124 	tick_init();
1125 	rcu_init_nohz();
1126 	timers_init();
1127 	srcu_init();
1128 	hrtimers_init();
1129 	softirq_init();
1130 	timekeeping_init();
1131 	time_init();
1132 
1133 	/* This must be after timekeeping is initialized */
1134 	random_init();
1135 
1136 	/* These make use of the fully initialized rng */
1137 	kfence_init();
1138 	boot_init_stack_canary();
1139 
1140 	perf_event_init();
1141 	profile_init();
1142 	call_function_init();
1143 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1144 
1145 	early_boot_irqs_disabled = false;
1146 	local_irq_enable();
1147 
1148 	kmem_cache_init_late();
1149 
1150 	/*
1151 	 * HACK ALERT! This is early. We're enabling the console before
1152 	 * we've done PCI setups etc, and console_init() must be aware of
1153 	 * this. But we do want output early, in case something goes wrong.
1154 	 */
1155 	console_init();
1156 	if (panic_later)
1157 		panic("Too many boot %s vars at `%s'", panic_later,
1158 		      panic_param);
1159 
1160 	lockdep_init();
1161 
1162 	/*
1163 	 * Need to run this when irqs are enabled, because it wants
1164 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1165 	 * too:
1166 	 */
1167 	locking_selftest();
1168 
1169 #ifdef CONFIG_BLK_DEV_INITRD
1170 	if (initrd_start && !initrd_below_start_ok &&
1171 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1172 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1173 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1174 		    min_low_pfn);
1175 		initrd_start = 0;
1176 	}
1177 #endif
1178 	setup_per_cpu_pageset();
1179 	numa_policy_init();
1180 	acpi_early_init();
1181 	if (late_time_init)
1182 		late_time_init();
1183 	sched_clock_init();
1184 	calibrate_delay();
1185 
1186 	arch_cpu_finalize_init();
1187 
1188 	pid_idr_init();
1189 	anon_vma_init();
1190 	thread_stack_cache_init();
1191 	cred_init();
1192 	fork_init();
1193 	proc_caches_init();
1194 	uts_ns_init();
1195 	time_ns_init();
1196 	key_init();
1197 	security_init();
1198 	dbg_late_init();
1199 	net_ns_init();
1200 	vfs_caches_init();
1201 	pagecache_init();
1202 	signals_init();
1203 	seq_file_init();
1204 	proc_root_init();
1205 	nsfs_init();
1206 	pidfs_init();
1207 	cpuset_init();
1208 	mem_cgroup_init();
1209 	cgroup_init();
1210 	taskstats_init_early();
1211 	delayacct_init();
1212 
1213 	acpi_subsystem_init();
1214 	arch_post_acpi_subsys_init();
1215 	kcsan_init();
1216 
1217 	/* Do the rest non-__init'ed, we're now alive */
1218 	rest_init();
1219 
1220 	/*
1221 	 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1222 	 * and older.
1223 	 */
1224 #if !__has_attribute(__no_stack_protector__)
1225 	prevent_tail_call_optimization();
1226 #endif
1227 }
1228 
1229 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1230 static void __init do_ctors(void)
1231 {
1232 /*
1233  * For UML, the constructors have already been called by the
1234  * normal setup code as it's just a normal ELF binary, so we
1235  * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1236  * even on UML for modules.
1237  */
1238 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1239 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1240 
1241 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1242 		(*fn)();
1243 #endif
1244 }
1245 
1246 #ifdef CONFIG_KALLSYMS
1247 struct blacklist_entry {
1248 	struct list_head next;
1249 	char *buf;
1250 };
1251 
1252 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1253 
initcall_blacklist(char * str)1254 static int __init initcall_blacklist(char *str)
1255 {
1256 	char *str_entry;
1257 	struct blacklist_entry *entry;
1258 
1259 	/* str argument is a comma-separated list of functions */
1260 	do {
1261 		str_entry = strsep(&str, ",");
1262 		if (str_entry) {
1263 			pr_debug("blacklisting initcall %s\n", str_entry);
1264 			entry = memblock_alloc_or_panic(sizeof(*entry),
1265 					       SMP_CACHE_BYTES);
1266 			entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1,
1267 						    SMP_CACHE_BYTES);
1268 			strcpy(entry->buf, str_entry);
1269 			list_add(&entry->next, &blacklisted_initcalls);
1270 		}
1271 	} while (str_entry);
1272 
1273 	return 1;
1274 }
1275 
initcall_blacklisted(initcall_t fn)1276 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1277 {
1278 	struct blacklist_entry *entry;
1279 	char fn_name[KSYM_SYMBOL_LEN];
1280 	unsigned long addr;
1281 
1282 	if (list_empty(&blacklisted_initcalls))
1283 		return false;
1284 
1285 	addr = (unsigned long) dereference_function_descriptor(fn);
1286 	sprint_symbol_no_offset(fn_name, addr);
1287 
1288 	/*
1289 	 * fn will be "function_name [module_name]" where [module_name] is not
1290 	 * displayed for built-in init functions.  Strip off the [module_name].
1291 	 */
1292 	strreplace(fn_name, ' ', '\0');
1293 
1294 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1295 		if (!strcmp(fn_name, entry->buf)) {
1296 			pr_debug("initcall %s blacklisted\n", fn_name);
1297 			return true;
1298 		}
1299 	}
1300 
1301 	return false;
1302 }
1303 #else
initcall_blacklist(char * str)1304 static int __init initcall_blacklist(char *str)
1305 {
1306 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1307 	return 0;
1308 }
1309 
initcall_blacklisted(initcall_t fn)1310 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1311 {
1312 	return false;
1313 }
1314 #endif
1315 __setup("initcall_blacklist=", initcall_blacklist);
1316 
1317 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1318 trace_initcall_start_cb(void *data, initcall_t fn)
1319 {
1320 	ktime_t *calltime = data;
1321 
1322 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1323 	*calltime = ktime_get();
1324 }
1325 
1326 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1327 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1328 {
1329 	ktime_t rettime, *calltime = data;
1330 
1331 	rettime = ktime_get();
1332 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1333 		 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1334 }
1335 
1336 static __init_or_module void
trace_initcall_level_cb(void * data,const char * level)1337 trace_initcall_level_cb(void *data, const char *level)
1338 {
1339 	printk(KERN_DEBUG "entering initcall level: %s\n", level);
1340 }
1341 
1342 static ktime_t initcall_calltime;
1343 
1344 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1345 static void __init initcall_debug_enable(void)
1346 {
1347 	int ret;
1348 
1349 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1350 					    &initcall_calltime);
1351 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1352 					      &initcall_calltime);
1353 	ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL);
1354 	WARN(ret, "Failed to register initcall tracepoints\n");
1355 }
1356 # define do_trace_initcall_start	trace_initcall_start
1357 # define do_trace_initcall_finish	trace_initcall_finish
1358 # define do_trace_initcall_level	trace_initcall_level
1359 #else
do_trace_initcall_start(initcall_t fn)1360 static inline void do_trace_initcall_start(initcall_t fn)
1361 {
1362 	if (!initcall_debug)
1363 		return;
1364 	trace_initcall_start_cb(&initcall_calltime, fn);
1365 }
do_trace_initcall_finish(initcall_t fn,int ret)1366 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1367 {
1368 	if (!initcall_debug)
1369 		return;
1370 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1371 }
do_trace_initcall_level(const char * level)1372 static inline void do_trace_initcall_level(const char *level)
1373 {
1374 	if (!initcall_debug)
1375 		return;
1376 	trace_initcall_level_cb(NULL, level);
1377 }
1378 #endif /* !TRACEPOINTS_ENABLED */
1379 
do_one_initcall(initcall_t fn)1380 int __init_or_module do_one_initcall(initcall_t fn)
1381 {
1382 	int count = preempt_count();
1383 	char msgbuf[64];
1384 	int ret;
1385 
1386 	if (initcall_blacklisted(fn))
1387 		return -EPERM;
1388 
1389 	do_trace_initcall_start(fn);
1390 	ret = fn();
1391 	do_trace_initcall_finish(fn, ret);
1392 
1393 	msgbuf[0] = 0;
1394 
1395 	if (preempt_count() != count) {
1396 		sprintf(msgbuf, "preemption imbalance ");
1397 		preempt_count_set(count);
1398 	}
1399 	if (irqs_disabled()) {
1400 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1401 		local_irq_enable();
1402 	}
1403 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1404 
1405 	add_latent_entropy();
1406 	return ret;
1407 }
1408 
1409 
1410 static initcall_entry_t *initcall_levels[] __initdata = {
1411 	__initcall0_start,
1412 	__initcall1_start,
1413 	__initcall2_start,
1414 	__initcall3_start,
1415 	__initcall4_start,
1416 	__initcall5_start,
1417 	__initcall6_start,
1418 	__initcall7_start,
1419 	__initcall_end,
1420 };
1421 
1422 /* Keep these in sync with initcalls in include/linux/init.h */
1423 static const char *initcall_level_names[] __initdata = {
1424 	"pure",
1425 	"core",
1426 	"postcore",
1427 	"arch",
1428 	"subsys",
1429 	"fs",
1430 	"device",
1431 	"late",
1432 };
1433 
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1434 static int __init ignore_unknown_bootoption(char *param, char *val,
1435 			       const char *unused, void *arg)
1436 {
1437 	return 0;
1438 }
1439 
do_initcall_level(int level,char * command_line)1440 static void __init do_initcall_level(int level, char *command_line)
1441 {
1442 	initcall_entry_t *fn;
1443 
1444 	parse_args(initcall_level_names[level],
1445 		   command_line, __start___param,
1446 		   __stop___param - __start___param,
1447 		   level, level,
1448 		   NULL, ignore_unknown_bootoption);
1449 
1450 	do_trace_initcall_level(initcall_level_names[level]);
1451 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1452 		do_one_initcall(initcall_from_entry(fn));
1453 }
1454 
do_initcalls(void)1455 static void __init do_initcalls(void)
1456 {
1457 	int level;
1458 	size_t len = saved_command_line_len + 1;
1459 	char *command_line;
1460 
1461 	command_line = kzalloc(len, GFP_KERNEL);
1462 	if (!command_line)
1463 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1464 
1465 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1466 		/* Parser modifies command_line, restore it each time */
1467 		strcpy(command_line, saved_command_line);
1468 		do_initcall_level(level, command_line);
1469 	}
1470 
1471 	kfree(command_line);
1472 }
1473 
1474 /*
1475  * Ok, the machine is now initialized. None of the devices
1476  * have been touched yet, but the CPU subsystem is up and
1477  * running, and memory and process management works.
1478  *
1479  * Now we can finally start doing some real work..
1480  */
do_basic_setup(void)1481 static void __init do_basic_setup(void)
1482 {
1483 	cpuset_init_smp();
1484 	ksysfs_init();
1485 	driver_init();
1486 	init_irq_proc();
1487 	do_ctors();
1488 	do_initcalls();
1489 }
1490 
do_pre_smp_initcalls(void)1491 static void __init do_pre_smp_initcalls(void)
1492 {
1493 	initcall_entry_t *fn;
1494 
1495 	do_trace_initcall_level("early");
1496 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1497 		do_one_initcall(initcall_from_entry(fn));
1498 }
1499 
run_init_process(const char * init_filename)1500 static int run_init_process(const char *init_filename)
1501 {
1502 	const char *const *p;
1503 
1504 	argv_init[0] = init_filename;
1505 	pr_info("Run %s as init process\n", init_filename);
1506 	pr_debug("  with arguments:\n");
1507 	for (p = argv_init; *p; p++)
1508 		pr_debug("    %s\n", *p);
1509 	pr_debug("  with environment:\n");
1510 	for (p = envp_init; *p; p++)
1511 		pr_debug("    %s\n", *p);
1512 	return kernel_execve(init_filename, argv_init, envp_init);
1513 }
1514 
try_to_run_init_process(const char * init_filename)1515 static int try_to_run_init_process(const char *init_filename)
1516 {
1517 	int ret;
1518 
1519 	ret = run_init_process(init_filename);
1520 
1521 	if (ret && ret != -ENOENT) {
1522 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1523 		       init_filename, ret);
1524 	}
1525 
1526 	return ret;
1527 }
1528 
1529 static noinline void __init kernel_init_freeable(void);
1530 
1531 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1532 bool rodata_enabled __ro_after_init = true;
1533 
1534 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1535 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1536 #endif
1537 
set_debug_rodata(char * str)1538 static int __init set_debug_rodata(char *str)
1539 {
1540 	if (arch_parse_debug_rodata(str))
1541 		return 0;
1542 
1543 	if (str && !strcmp(str, "on"))
1544 		rodata_enabled = true;
1545 	else if (str && !strcmp(str, "off"))
1546 		rodata_enabled = false;
1547 	else
1548 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1549 	return 0;
1550 }
1551 early_param("rodata", set_debug_rodata);
1552 #endif
1553 
mark_readonly(void)1554 static void mark_readonly(void)
1555 {
1556 	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1557 		/*
1558 		 * load_module() results in W+X mappings, which are cleaned
1559 		 * up with init_free_wq. Let's make sure that queued work is
1560 		 * flushed so that we don't hit false positives looking for
1561 		 * insecure pages which are W+X.
1562 		 */
1563 		flush_module_init_free_work();
1564 		jump_label_init_ro();
1565 		mark_rodata_ro();
1566 		debug_checkwx();
1567 		rodata_test();
1568 	} else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1569 		pr_info("Kernel memory protection disabled.\n");
1570 	} else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1571 		pr_warn("Kernel memory protection not selected by kernel config.\n");
1572 	} else {
1573 		pr_warn("This architecture does not have kernel memory protection.\n");
1574 	}
1575 }
1576 
free_initmem(void)1577 void __weak free_initmem(void)
1578 {
1579 	free_initmem_default(POISON_FREE_INITMEM);
1580 }
1581 
kernel_init(void * unused)1582 static int __ref kernel_init(void *unused)
1583 {
1584 	int ret;
1585 
1586 	/*
1587 	 * Wait until kthreadd is all set-up.
1588 	 */
1589 	wait_for_completion(&kthreadd_done);
1590 
1591 	kernel_init_freeable();
1592 	/* need to finish all async __init code before freeing the memory */
1593 	async_synchronize_full();
1594 
1595 	system_state = SYSTEM_FREEING_INITMEM;
1596 	kprobe_free_init_mem();
1597 	ftrace_free_init_mem();
1598 	kgdb_free_init_mem();
1599 	exit_boot_config();
1600 	free_initmem();
1601 	mark_readonly();
1602 
1603 	/*
1604 	 * Kernel mappings are now finalized - update the userspace page-table
1605 	 * to finalize PTI.
1606 	 */
1607 	pti_finalize();
1608 
1609 	system_state = SYSTEM_RUNNING;
1610 	numa_default_policy();
1611 
1612 	rcu_end_inkernel_boot();
1613 
1614 	do_sysctl_args();
1615 
1616 	if (ramdisk_execute_command) {
1617 		ret = run_init_process(ramdisk_execute_command);
1618 		if (!ret)
1619 			return 0;
1620 		pr_err("Failed to execute %s (error %d)\n",
1621 		       ramdisk_execute_command, ret);
1622 	}
1623 
1624 	/*
1625 	 * We try each of these until one succeeds.
1626 	 *
1627 	 * The Bourne shell can be used instead of init if we are
1628 	 * trying to recover a really broken machine.
1629 	 */
1630 	if (execute_command) {
1631 		ret = run_init_process(execute_command);
1632 		if (!ret)
1633 			return 0;
1634 		panic("Requested init %s failed (error %d).",
1635 		      execute_command, ret);
1636 	}
1637 
1638 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1639 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1640 		if (ret)
1641 			pr_err("Default init %s failed (error %d)\n",
1642 			       CONFIG_DEFAULT_INIT, ret);
1643 		else
1644 			return 0;
1645 	}
1646 
1647 	if (!try_to_run_init_process("/sbin/init") ||
1648 	    !try_to_run_init_process("/etc/init") ||
1649 	    !try_to_run_init_process("/bin/init") ||
1650 	    !try_to_run_init_process("/bin/sh"))
1651 		return 0;
1652 
1653 	panic("No working init found.  Try passing init= option to kernel. "
1654 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1655 }
1656 
1657 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1658 void __init console_on_rootfs(void)
1659 {
1660 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1661 
1662 	if (IS_ERR(file)) {
1663 		pr_err("Warning: unable to open an initial console.\n");
1664 		return;
1665 	}
1666 	init_dup(file);
1667 	init_dup(file);
1668 	init_dup(file);
1669 	fput(file);
1670 }
1671 
kernel_init_freeable(void)1672 static noinline void __init kernel_init_freeable(void)
1673 {
1674 	/* Now the scheduler is fully set up and can do blocking allocations */
1675 	gfp_allowed_mask = __GFP_BITS_MASK;
1676 
1677 	/*
1678 	 * init can allocate pages on any node
1679 	 */
1680 	set_mems_allowed(node_states[N_MEMORY]);
1681 
1682 	cad_pid = get_pid(task_pid(current));
1683 
1684 	smp_prepare_cpus(setup_max_cpus);
1685 
1686 	workqueue_init();
1687 
1688 	init_mm_internals();
1689 
1690 	do_pre_smp_initcalls();
1691 	lockup_detector_init();
1692 
1693 	smp_init();
1694 	sched_init_smp();
1695 
1696 	workqueue_init_topology();
1697 	async_init();
1698 	padata_init();
1699 	page_alloc_init_late();
1700 
1701 	do_basic_setup();
1702 
1703 	kunit_run_all_tests();
1704 
1705 	wait_for_initramfs();
1706 	console_on_rootfs();
1707 
1708 	/*
1709 	 * check if there is an early userspace init.  If yes, let it do all
1710 	 * the work
1711 	 */
1712 	int ramdisk_command_access;
1713 	ramdisk_command_access = init_eaccess(ramdisk_execute_command);
1714 	if (ramdisk_command_access != 0) {
1715 		if (ramdisk_execute_command_set)
1716 			pr_warn("check access for rdinit=%s failed: %i, ignoring\n",
1717 				ramdisk_execute_command, ramdisk_command_access);
1718 		ramdisk_execute_command = NULL;
1719 		prepare_namespace();
1720 	}
1721 
1722 	/*
1723 	 * Ok, we have completed the initial bootup, and
1724 	 * we're essentially up and running. Get rid of the
1725 	 * initmem segments and start the user-mode stuff..
1726 	 *
1727 	 * rootfs is available now, try loading the public keys
1728 	 * and default modules
1729 	 */
1730 
1731 	integrity_load_keys();
1732 }
1733