• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/init/main.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
8  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
11  */
12 
13 #define DEBUG		/* Enable initcall_debug */
14 
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/vmalloc.h>
38 #include <linux/kernel_stat.h>
39 #include <linux/start_kernel.h>
40 #include <linux/security.h>
41 #include <linux/smp.h>
42 #include <linux/profile.h>
43 #include <linux/rcupdate.h>
44 #include <linux/moduleparam.h>
45 #include <linux/kallsyms.h>
46 #include <linux/writeback.h>
47 #include <linux/cpu.h>
48 #include <linux/cpuset.h>
49 #include <linux/cgroup.h>
50 #include <linux/efi.h>
51 #include <linux/tick.h>
52 #include <linux/sched/isolation.h>
53 #include <linux/interrupt.h>
54 #include <linux/taskstats_kern.h>
55 #include <linux/delayacct.h>
56 #include <linux/unistd.h>
57 #include <linux/utsname.h>
58 #include <linux/rmap.h>
59 #include <linux/mempolicy.h>
60 #include <linux/key.h>
61 #include <linux/buffer_head.h>
62 #include <linux/page_ext.h>
63 #include <linux/debug_locks.h>
64 #include <linux/debugobjects.h>
65 #include <linux/lockdep.h>
66 #include <linux/kmemleak.h>
67 #include <linux/padata.h>
68 #include <linux/pid_namespace.h>
69 #include <linux/device/driver.h>
70 #include <linux/kthread.h>
71 #include <linux/sched.h>
72 #include <linux/sched/init.h>
73 #include <linux/signal.h>
74 #include <linux/idr.h>
75 #include <linux/kgdb.h>
76 #include <linux/ftrace.h>
77 #include <linux/async.h>
78 #include <linux/sfi.h>
79 #include <linux/shmem_fs.h>
80 #include <linux/slab.h>
81 #include <linux/perf_event.h>
82 #include <linux/ptrace.h>
83 #include <linux/pti.h>
84 #include <linux/blkdev.h>
85 #include <linux/elevator.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/list.h>
92 #include <linux/integrity.h>
93 #include <linux/proc_ns.h>
94 #include <linux/io.h>
95 #include <linux/cache.h>
96 #include <linux/rodata_test.h>
97 #include <linux/jump_label.h>
98 #include <linux/mem_encrypt.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #ifdef CONFIG_RECLAIM_ACCT
102 #include <linux/reclaim_acct.h>
103 #endif
104 
105 #include <asm/io.h>
106 #include <asm/bugs.h>
107 #include <asm/setup.h>
108 #include <asm/sections.h>
109 #include <asm/cacheflush.h>
110 
111 #define CREATE_TRACE_POINTS
112 #include <trace/events/initcall.h>
113 
114 #include <kunit/test.h>
115 
116 static int kernel_init(void *);
117 
118 extern void init_IRQ(void);
119 extern void radix_tree_init(void);
120 
121 /*
122  * Debug helper: via this flag we know that we are in 'early bootup code'
123  * where only the boot processor is running with IRQ disabled.  This means
124  * two things - IRQ must not be enabled before the flag is cleared and some
125  * operations which are not allowed with IRQ disabled are allowed while the
126  * flag is set.
127  */
128 bool early_boot_irqs_disabled __read_mostly;
129 
130 enum system_states system_state __read_mostly;
131 EXPORT_SYMBOL(system_state);
132 
133 /*
134  * Boot command-line arguments
135  */
136 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
137 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
138 
139 extern void time_init(void);
140 /* Default late time init is NULL. archs can override this later. */
141 void (*__initdata late_time_init)(void);
142 
143 /* Untouched command line saved by arch-specific code. */
144 char __initdata boot_command_line[COMMAND_LINE_SIZE];
145 /* Untouched saved command line (eg. for /proc) */
146 char *saved_command_line;
147 /* Command line for parameter parsing */
148 static char *static_command_line;
149 /* Untouched extra command line */
150 static char *extra_command_line;
151 /* Extra init arguments */
152 static char *extra_init_args;
153 
154 #ifdef CONFIG_BOOT_CONFIG
155 /* Is bootconfig on command line? */
156 static bool bootconfig_found;
157 static bool initargs_found;
158 #else
159 # define bootconfig_found false
160 # define initargs_found false
161 #endif
162 
163 static char *execute_command;
164 static char *ramdisk_execute_command = "/init";
165 
166 /*
167  * Used to generate warnings if static_key manipulation functions are used
168  * before jump_label_init is called.
169  */
170 bool static_key_initialized __read_mostly;
171 EXPORT_SYMBOL_GPL(static_key_initialized);
172 
173 /*
174  * If set, this is an indication to the drivers that reset the underlying
175  * device before going ahead with the initialization otherwise driver might
176  * rely on the BIOS and skip the reset operation.
177  *
178  * This is useful if kernel is booting in an unreliable environment.
179  * For ex. kdump situation where previous kernel has crashed, BIOS has been
180  * skipped and devices will be in unknown state.
181  */
182 unsigned int reset_devices;
183 EXPORT_SYMBOL(reset_devices);
184 
set_reset_devices(char * str)185 static int __init set_reset_devices(char *str)
186 {
187 	reset_devices = 1;
188 	return 1;
189 }
190 
191 __setup("reset_devices", set_reset_devices);
192 
193 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
194 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
195 static const char *panic_later, *panic_param;
196 
197 extern const struct obs_kernel_param __setup_start[], __setup_end[];
198 
obsolete_checksetup(char * line)199 static bool __init obsolete_checksetup(char *line)
200 {
201 	const struct obs_kernel_param *p;
202 	bool had_early_param = false;
203 
204 	p = __setup_start;
205 	do {
206 		int n = strlen(p->str);
207 		if (parameqn(line, p->str, n)) {
208 			if (p->early) {
209 				/* Already done in parse_early_param?
210 				 * (Needs exact match on param part).
211 				 * Keep iterating, as we can have early
212 				 * params and __setups of same names 8( */
213 				if (line[n] == '\0' || line[n] == '=')
214 					had_early_param = true;
215 			} else if (!p->setup_func) {
216 				pr_warn("Parameter %s is obsolete, ignored\n",
217 					p->str);
218 				return true;
219 			} else if (p->setup_func(line + n))
220 				return true;
221 		}
222 		p++;
223 	} while (p < __setup_end);
224 
225 	return had_early_param;
226 }
227 
228 /*
229  * This should be approx 2 Bo*oMips to start (note initial shift), and will
230  * still work even if initially too large, it will just take slightly longer
231  */
232 unsigned long loops_per_jiffy = (1<<12);
233 EXPORT_SYMBOL(loops_per_jiffy);
234 
debug_kernel(char * str)235 static int __init debug_kernel(char *str)
236 {
237 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
238 	return 0;
239 }
240 
quiet_kernel(char * str)241 static int __init quiet_kernel(char *str)
242 {
243 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
244 	return 0;
245 }
246 
247 early_param("debug", debug_kernel);
248 early_param("quiet", quiet_kernel);
249 
loglevel(char * str)250 static int __init loglevel(char *str)
251 {
252 	int newlevel;
253 
254 	/*
255 	 * Only update loglevel value when a correct setting was passed,
256 	 * to prevent blind crashes (when loglevel being set to 0) that
257 	 * are quite hard to debug
258 	 */
259 	if (get_option(&str, &newlevel)) {
260 		console_loglevel = newlevel;
261 		return 0;
262 	}
263 
264 	return -EINVAL;
265 }
266 
267 early_param("loglevel", loglevel);
268 
269 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(u32 * _size,u32 * _csum)270 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
271 {
272 	u32 size, csum;
273 	char *data;
274 	u32 *hdr;
275 	int i;
276 
277 	if (!initrd_end)
278 		return NULL;
279 
280 	data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
281 	/*
282 	 * Since Grub may align the size of initrd to 4, we must
283 	 * check the preceding 3 bytes as well.
284 	 */
285 	for (i = 0; i < 4; i++) {
286 		if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
287 			goto found;
288 		data--;
289 	}
290 	return NULL;
291 
292 found:
293 	hdr = (u32 *)(data - 8);
294 	size = le32_to_cpu(hdr[0]);
295 	csum = le32_to_cpu(hdr[1]);
296 
297 	data = ((void *)hdr) - size;
298 	if ((unsigned long)data < initrd_start) {
299 		pr_err("bootconfig size %d is greater than initrd size %ld\n",
300 			size, initrd_end - initrd_start);
301 		return NULL;
302 	}
303 
304 	/* Remove bootconfig from initramfs/initrd */
305 	initrd_end = (unsigned long)data;
306 	if (_size)
307 		*_size = size;
308 	if (_csum)
309 		*_csum = csum;
310 
311 	return data;
312 }
313 #else
get_boot_config_from_initrd(u32 * _size,u32 * _csum)314 static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
315 {
316 	return NULL;
317 }
318 #endif
319 
320 #ifdef CONFIG_BOOT_CONFIG
321 
322 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
323 
324 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
325 
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)326 static int __init xbc_snprint_cmdline(char *buf, size_t size,
327 				      struct xbc_node *root)
328 {
329 	struct xbc_node *knode, *vnode;
330 	char *end = buf + size;
331 	const char *val;
332 	int ret;
333 
334 	xbc_node_for_each_key_value(root, knode, val) {
335 		ret = xbc_node_compose_key_after(root, knode,
336 					xbc_namebuf, XBC_KEYLEN_MAX);
337 		if (ret < 0)
338 			return ret;
339 
340 		vnode = xbc_node_get_child(knode);
341 		if (!vnode) {
342 			ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
343 			if (ret < 0)
344 				return ret;
345 			buf += ret;
346 			continue;
347 		}
348 		xbc_array_for_each_value(vnode, val) {
349 			ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
350 				       xbc_namebuf, val);
351 			if (ret < 0)
352 				return ret;
353 			buf += ret;
354 		}
355 	}
356 
357 	return buf - (end - size);
358 }
359 #undef rest
360 
361 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)362 static char * __init xbc_make_cmdline(const char *key)
363 {
364 	struct xbc_node *root;
365 	char *new_cmdline;
366 	int ret, len = 0;
367 
368 	root = xbc_find_node(key);
369 	if (!root)
370 		return NULL;
371 
372 	/* Count required buffer size */
373 	len = xbc_snprint_cmdline(NULL, 0, root);
374 	if (len <= 0)
375 		return NULL;
376 
377 	new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
378 	if (!new_cmdline) {
379 		pr_err("Failed to allocate memory for extra kernel cmdline.\n");
380 		return NULL;
381 	}
382 
383 	ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
384 	if (ret < 0 || ret > len) {
385 		pr_err("Failed to print extra kernel cmdline.\n");
386 		memblock_free(__pa(new_cmdline), len + 1);
387 		return NULL;
388 	}
389 
390 	return new_cmdline;
391 }
392 
boot_config_checksum(unsigned char * p,u32 size)393 static u32 boot_config_checksum(unsigned char *p, u32 size)
394 {
395 	u32 ret = 0;
396 
397 	while (size--)
398 		ret += *p++;
399 
400 	return ret;
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 
setup_boot_config(const char * cmdline)412 static void __init setup_boot_config(const char *cmdline)
413 {
414 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
415 	const char *msg;
416 	int pos;
417 	u32 size, csum;
418 	char *data, *copy, *err;
419 	int ret;
420 
421 	/* Cut out the bootconfig data even if we have no bootconfig option */
422 	data = get_boot_config_from_initrd(&size, &csum);
423 
424 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
425 	err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
426 			 bootconfig_params);
427 
428 	if (IS_ERR(err) || !bootconfig_found)
429 		return;
430 
431 	/* parse_args() stops at '--' and returns an address */
432 	if (err)
433 		initargs_found = true;
434 
435 	if (!data) {
436 		pr_err("'bootconfig' found on command line, but no bootconfig found\n");
437 		return;
438 	}
439 
440 	if (size >= XBC_DATA_MAX) {
441 		pr_err("bootconfig size %d greater than max size %d\n",
442 			size, XBC_DATA_MAX);
443 		return;
444 	}
445 
446 	if (boot_config_checksum((unsigned char *)data, size) != csum) {
447 		pr_err("bootconfig checksum failed\n");
448 		return;
449 	}
450 
451 	copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
452 	if (!copy) {
453 		pr_err("Failed to allocate memory for bootconfig\n");
454 		return;
455 	}
456 
457 	memcpy(copy, data, size);
458 	copy[size] = '\0';
459 
460 	ret = xbc_init(copy, &msg, &pos);
461 	if (ret < 0) {
462 		if (pos < 0)
463 			pr_err("Failed to init bootconfig: %s.\n", msg);
464 		else
465 			pr_err("Failed to parse bootconfig: %s at %d.\n",
466 				msg, pos);
467 	} else {
468 		pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
469 		/* keys starting with "kernel." are passed via cmdline */
470 		extra_command_line = xbc_make_cmdline("kernel");
471 		/* Also, "init." keys are init arguments */
472 		extra_init_args = xbc_make_cmdline("init");
473 	}
474 	return;
475 }
476 
477 #else
478 
setup_boot_config(const char * cmdline)479 static void __init setup_boot_config(const char *cmdline)
480 {
481 	/* Remove bootconfig data from initrd */
482 	get_boot_config_from_initrd(NULL, NULL);
483 }
484 
warn_bootconfig(char * str)485 static int __init warn_bootconfig(char *str)
486 {
487 	pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
488 	return 0;
489 }
490 early_param("bootconfig", warn_bootconfig);
491 
492 #endif
493 
494 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)495 static void __init repair_env_string(char *param, char *val)
496 {
497 	if (val) {
498 		/* param=val or param="val"? */
499 		if (val == param+strlen(param)+1)
500 			val[-1] = '=';
501 		else if (val == param+strlen(param)+2) {
502 			val[-2] = '=';
503 			memmove(val-1, val, strlen(val)+1);
504 		} else
505 			BUG();
506 	}
507 }
508 
509 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)510 static int __init set_init_arg(char *param, char *val,
511 			       const char *unused, void *arg)
512 {
513 	unsigned int i;
514 
515 	if (panic_later)
516 		return 0;
517 
518 	repair_env_string(param, val);
519 
520 	for (i = 0; argv_init[i]; i++) {
521 		if (i == MAX_INIT_ARGS) {
522 			panic_later = "init";
523 			panic_param = param;
524 			return 0;
525 		}
526 	}
527 	argv_init[i] = param;
528 	return 0;
529 }
530 
531 /*
532  * Unknown boot options get handed to init, unless they look like
533  * unused parameters (modprobe will find them in /proc/cmdline).
534  */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)535 static int __init unknown_bootoption(char *param, char *val,
536 				     const char *unused, void *arg)
537 {
538 	size_t len = strlen(param);
539 
540 	repair_env_string(param, val);
541 
542 	/* Handle obsolete-style parameters */
543 	if (obsolete_checksetup(param))
544 		return 0;
545 
546 	/* Unused module parameter. */
547 	if (strnchr(param, len, '.'))
548 		return 0;
549 
550 	if (panic_later)
551 		return 0;
552 
553 	if (val) {
554 		/* Environment option */
555 		unsigned int i;
556 		for (i = 0; envp_init[i]; i++) {
557 			if (i == MAX_INIT_ENVS) {
558 				panic_later = "env";
559 				panic_param = param;
560 			}
561 			if (!strncmp(param, envp_init[i], len+1))
562 				break;
563 		}
564 		envp_init[i] = param;
565 	} else {
566 		/* Command line option */
567 		unsigned int i;
568 		for (i = 0; argv_init[i]; i++) {
569 			if (i == MAX_INIT_ARGS) {
570 				panic_later = "init";
571 				panic_param = param;
572 			}
573 		}
574 		argv_init[i] = param;
575 	}
576 	return 0;
577 }
578 
init_setup(char * str)579 static int __init init_setup(char *str)
580 {
581 	unsigned int i;
582 
583 	execute_command = str;
584 	/*
585 	 * In case LILO is going to boot us with default command line,
586 	 * it prepends "auto" before the whole cmdline which makes
587 	 * the shell think it should execute a script with such name.
588 	 * So we ignore all arguments entered _before_ init=... [MJ]
589 	 */
590 	for (i = 1; i < MAX_INIT_ARGS; i++)
591 		argv_init[i] = NULL;
592 	return 1;
593 }
594 __setup("init=", init_setup);
595 
rdinit_setup(char * str)596 static int __init rdinit_setup(char *str)
597 {
598 	unsigned int i;
599 
600 	ramdisk_execute_command = str;
601 	/* See "auto" comment in init_setup */
602 	for (i = 1; i < MAX_INIT_ARGS; i++)
603 		argv_init[i] = NULL;
604 	return 1;
605 }
606 __setup("rdinit=", rdinit_setup);
607 
608 #ifndef CONFIG_SMP
609 static const unsigned int setup_max_cpus = NR_CPUS;
setup_nr_cpu_ids(void)610 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)611 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
612 #endif
613 
614 /*
615  * We need to store the untouched command line for future reference.
616  * We also need to store the touched command line since the parameter
617  * parsing is performed in place, and we should allow a component to
618  * store reference of name/value for future reference.
619  */
setup_command_line(char * command_line)620 static void __init setup_command_line(char *command_line)
621 {
622 	size_t len, xlen = 0, ilen = 0;
623 
624 	if (extra_command_line)
625 		xlen = strlen(extra_command_line);
626 	if (extra_init_args)
627 		ilen = strlen(extra_init_args) + 4; /* for " -- " */
628 
629 	len = xlen + strlen(boot_command_line) + 1;
630 
631 	saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
632 	if (!saved_command_line)
633 		panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
634 
635 	static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
636 	if (!static_command_line)
637 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
638 
639 	if (xlen) {
640 		/*
641 		 * We have to put extra_command_line before boot command
642 		 * lines because there could be dashes (separator of init
643 		 * command line) in the command lines.
644 		 */
645 		strcpy(saved_command_line, extra_command_line);
646 		strcpy(static_command_line, extra_command_line);
647 	}
648 	strcpy(saved_command_line + xlen, boot_command_line);
649 	strcpy(static_command_line + xlen, command_line);
650 
651 	if (ilen) {
652 		/*
653 		 * Append supplemental init boot args to saved_command_line
654 		 * so that user can check what command line options passed
655 		 * to init.
656 		 */
657 		len = strlen(saved_command_line);
658 		if (initargs_found) {
659 			saved_command_line[len++] = ' ';
660 		} else {
661 			strcpy(saved_command_line + len, " -- ");
662 			len += 4;
663 		}
664 
665 		strcpy(saved_command_line + len, extra_init_args);
666 	}
667 }
668 
669 /*
670  * We need to finalize in a non-__init function or else race conditions
671  * between the root thread and the init thread may cause start_kernel to
672  * be reaped by free_initmem before the root thread has proceeded to
673  * cpu_idle.
674  *
675  * gcc-3.4 accidentally inlines this function, so use noinline.
676  */
677 
678 static __initdata DECLARE_COMPLETION(kthreadd_done);
679 
rest_init(void)680 noinline void __ref rest_init(void)
681 {
682 	struct task_struct *tsk;
683 	int pid;
684 
685 	rcu_scheduler_starting();
686 	/*
687 	 * We need to spawn init first so that it obtains pid 1, however
688 	 * the init task will end up wanting to create kthreads, which, if
689 	 * we schedule it before we create kthreadd, will OOPS.
690 	 */
691 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
692 	/*
693 	 * Pin init on the boot CPU. Task migration is not properly working
694 	 * until sched_init_smp() has been run. It will set the allowed
695 	 * CPUs for init to the non isolated CPUs.
696 	 */
697 	rcu_read_lock();
698 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
699 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
700 	rcu_read_unlock();
701 
702 	numa_default_policy();
703 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
704 	rcu_read_lock();
705 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
706 	rcu_read_unlock();
707 
708 	/*
709 	 * Enable might_sleep() and smp_processor_id() checks.
710 	 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
711 	 * kernel_thread() would trigger might_sleep() splats. With
712 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
713 	 * already, but it's stuck on the kthreadd_done completion.
714 	 */
715 	system_state = SYSTEM_SCHEDULING;
716 
717 	complete(&kthreadd_done);
718 
719 	/*
720 	 * The boot idle thread must execute schedule()
721 	 * at least once to get things moving:
722 	 */
723 	schedule_preempt_disabled();
724 	/* Call into cpu_idle with preempt disabled */
725 	cpu_startup_entry(CPUHP_ONLINE);
726 }
727 
728 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)729 static int __init do_early_param(char *param, char *val,
730 				 const char *unused, void *arg)
731 {
732 	const struct obs_kernel_param *p;
733 
734 	for (p = __setup_start; p < __setup_end; p++) {
735 		if ((p->early && parameq(param, p->str)) ||
736 		    (strcmp(param, "console") == 0 &&
737 		     strcmp(p->str, "earlycon") == 0)
738 		) {
739 			if (p->setup_func(val) != 0)
740 				pr_warn("Malformed early option '%s'\n", param);
741 		}
742 	}
743 	/* We accept everything at this stage. */
744 	return 0;
745 }
746 
parse_early_options(char * cmdline)747 void __init parse_early_options(char *cmdline)
748 {
749 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
750 		   do_early_param);
751 }
752 
753 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)754 void __init parse_early_param(void)
755 {
756 	static int done __initdata;
757 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
758 
759 	if (done)
760 		return;
761 
762 	/* All fall through to do_early_param. */
763 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
764 	parse_early_options(tmp_cmdline);
765 	done = 1;
766 }
767 
arch_post_acpi_subsys_init(void)768 void __init __weak arch_post_acpi_subsys_init(void) { }
769 
smp_setup_processor_id(void)770 void __init __weak smp_setup_processor_id(void)
771 {
772 }
773 
774 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)775 void __init __weak thread_stack_cache_init(void)
776 {
777 }
778 #endif
779 
mem_encrypt_init(void)780 void __init __weak mem_encrypt_init(void) { }
781 
poking_init(void)782 void __init __weak poking_init(void) { }
783 
pgtable_cache_init(void)784 void __init __weak pgtable_cache_init(void) { }
785 
786 bool initcall_debug;
787 core_param(initcall_debug, initcall_debug, bool, 0644);
788 
789 #ifdef TRACEPOINTS_ENABLED
790 static void __init initcall_debug_enable(void);
791 #else
initcall_debug_enable(void)792 static inline void initcall_debug_enable(void)
793 {
794 }
795 #endif
796 
797 /* Report memory auto-initialization states for this boot. */
report_meminit(void)798 static void __init report_meminit(void)
799 {
800 	const char *stack;
801 
802 	if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
803 		stack = "all(pattern)";
804 	else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
805 		stack = "all(zero)";
806 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
807 		stack = "byref_all(zero)";
808 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
809 		stack = "byref(zero)";
810 	else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
811 		stack = "__user(zero)";
812 	else
813 		stack = "off";
814 
815 	pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
816 		stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
817 		want_init_on_free() ? "on" : "off");
818 	if (want_init_on_free())
819 		pr_info("mem auto-init: clearing system memory may take some time...\n");
820 }
821 
822 /*
823  * Set up kernel memory allocators
824  */
mm_init(void)825 static void __init mm_init(void)
826 {
827 	/*
828 	 * page_ext requires contiguous pages,
829 	 * bigger than MAX_ORDER unless SPARSEMEM.
830 	 */
831 	page_ext_init_flatmem();
832 	init_debug_pagealloc();
833 	report_meminit();
834 	mem_init();
835 	/* page_owner must be initialized after buddy is ready */
836 	page_ext_init_flatmem_late();
837 	kmem_cache_init();
838 	kmemleak_init();
839 	pgtable_init();
840 	debug_objects_mem_init();
841 	vmalloc_init();
842 	ioremap_huge_init();
843 	/* Should be run before the first non-init thread is created */
844 	init_espfix_bsp();
845 	/* Should be run after espfix64 is set up. */
846 	pti_init();
847 }
848 
arch_call_rest_init(void)849 void __init __weak arch_call_rest_init(void)
850 {
851 	rest_init();
852 }
853 
start_kernel(void)854 asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
855 {
856 	char *command_line;
857 	char *after_dashes;
858 
859 	set_task_stack_end_magic(&init_task);
860 	smp_setup_processor_id();
861 	debug_objects_early_init();
862 
863 	cgroup_init_early();
864 
865 	local_irq_disable();
866 	early_boot_irqs_disabled = true;
867 
868 	/*
869 	 * Interrupts are still disabled. Do necessary setups, then
870 	 * enable them.
871 	 */
872 	boot_cpu_init();
873 	page_address_init();
874 	pr_notice("%s", linux_banner);
875 	early_security_init();
876 	setup_arch(&command_line);
877 	setup_boot_config(command_line);
878 	setup_command_line(command_line);
879 	setup_nr_cpu_ids();
880 	setup_per_cpu_areas();
881 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
882 	boot_cpu_hotplug_init();
883 
884 	build_all_zonelists(NULL);
885 	page_alloc_init();
886 
887 	pr_notice("Kernel command line: %s\n", saved_command_line);
888 	/* parameters may set static keys */
889 	jump_label_init();
890 	parse_early_param();
891 	after_dashes = parse_args("Booting kernel",
892 				  static_command_line, __start___param,
893 				  __stop___param - __start___param,
894 				  -1, -1, NULL, &unknown_bootoption);
895 	if (!IS_ERR_OR_NULL(after_dashes))
896 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
897 			   NULL, set_init_arg);
898 	if (extra_init_args)
899 		parse_args("Setting extra init args", extra_init_args,
900 			   NULL, 0, -1, -1, NULL, set_init_arg);
901 
902 	/*
903 	 * These use large bootmem allocations and must precede
904 	 * kmem_cache_init()
905 	 */
906 	setup_log_buf(0);
907 	vfs_caches_init_early();
908 	sort_main_extable();
909 	trap_init();
910 	mm_init();
911 
912 	ftrace_init();
913 
914 	/* trace_printk can be enabled here */
915 	early_trace_init();
916 
917 	/*
918 	 * Set up the scheduler prior starting any interrupts (such as the
919 	 * timer interrupt). Full topology setup happens at smp_init()
920 	 * time - but meanwhile we still have a functioning scheduler.
921 	 */
922 	sched_init();
923 
924 	if (WARN(!irqs_disabled(),
925 		 "Interrupts were enabled *very* early, fixing it\n"))
926 		local_irq_disable();
927 	radix_tree_init();
928 
929 	/*
930 	 * Set up housekeeping before setting up workqueues to allow the unbound
931 	 * workqueue to take non-housekeeping into account.
932 	 */
933 	housekeeping_init();
934 
935 	/*
936 	 * Allow workqueue creation and work item queueing/cancelling
937 	 * early.  Work item execution depends on kthreads and starts after
938 	 * workqueue_init().
939 	 */
940 	workqueue_init_early();
941 
942 	rcu_init();
943 
944 	/* Trace events are available after this */
945 	trace_init();
946 
947 	if (initcall_debug)
948 		initcall_debug_enable();
949 
950 	context_tracking_init();
951 	/* init some links before init_ISA_irqs() */
952 	early_irq_init();
953 	init_IRQ();
954 	tick_init();
955 	rcu_init_nohz();
956 	init_timers();
957 	hrtimers_init();
958 	softirq_init();
959 	timekeeping_init();
960 	time_init();
961 
962 	/*
963 	 * For best initial stack canary entropy, prepare it after:
964 	 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
965 	 * - timekeeping_init() for ktime entropy used in random_init()
966 	 * - time_init() for making random_get_entropy() work on some platforms
967 	 * - random_init() to initialize the RNG from from early entropy sources
968 	 */
969 	random_init(command_line);
970 	boot_init_stack_canary();
971 
972 	perf_event_init();
973 	profile_init();
974 	call_function_init();
975 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
976 
977 	early_boot_irqs_disabled = false;
978 	local_irq_enable();
979 
980 	kmem_cache_init_late();
981 
982 	/*
983 	 * HACK ALERT! This is early. We're enabling the console before
984 	 * we've done PCI setups etc, and console_init() must be aware of
985 	 * this. But we do want output early, in case something goes wrong.
986 	 */
987 	console_init();
988 	if (panic_later)
989 		panic("Too many boot %s vars at `%s'", panic_later,
990 		      panic_param);
991 
992 	lockdep_init();
993 
994 	/*
995 	 * Need to run this when irqs are enabled, because it wants
996 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
997 	 * too:
998 	 */
999 	locking_selftest();
1000 
1001 	/*
1002 	 * This needs to be called before any devices perform DMA
1003 	 * operations that might use the SWIOTLB bounce buffers. It will
1004 	 * mark the bounce buffers as decrypted so that their usage will
1005 	 * not cause "plain-text" data to be decrypted when accessed.
1006 	 */
1007 	mem_encrypt_init();
1008 
1009 #ifdef CONFIG_BLK_DEV_INITRD
1010 	if (initrd_start && !initrd_below_start_ok &&
1011 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1012 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1013 		    page_to_pfn(virt_to_page((void *)initrd_start)),
1014 		    min_low_pfn);
1015 		initrd_start = 0;
1016 	}
1017 #endif
1018 	setup_per_cpu_pageset();
1019 	numa_policy_init();
1020 	acpi_early_init();
1021 	if (late_time_init)
1022 		late_time_init();
1023 	sched_clock_init();
1024 	calibrate_delay();
1025 	pid_idr_init();
1026 	anon_vma_init();
1027 #ifdef CONFIG_X86
1028 	if (efi_enabled(EFI_RUNTIME_SERVICES))
1029 		efi_enter_virtual_mode();
1030 #endif
1031 	thread_stack_cache_init();
1032 	cred_init();
1033 	fork_init();
1034 	proc_caches_init();
1035 	uts_ns_init();
1036 	buffer_init();
1037 	key_init();
1038 	security_init();
1039 	dbg_late_init();
1040 	vfs_caches_init();
1041 	pagecache_init();
1042 	signals_init();
1043 	seq_file_init();
1044 	proc_root_init();
1045 	nsfs_init();
1046 	cpuset_init();
1047 	cgroup_init();
1048 	taskstats_init_early();
1049 	delayacct_init();
1050 #ifdef CONFIG_RECLAIM_ACCT
1051 	reclaimacct_init();
1052 #endif
1053 
1054 	poking_init();
1055 	check_bugs();
1056 
1057 	acpi_subsystem_init();
1058 	arch_post_acpi_subsys_init();
1059 	sfi_init_late();
1060 	kcsan_init();
1061 
1062 	/* Do the rest non-__init'ed, we're now alive */
1063 	arch_call_rest_init();
1064 
1065 	prevent_tail_call_optimization();
1066 }
1067 
1068 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1069 static void __init do_ctors(void)
1070 {
1071 #ifdef CONFIG_CONSTRUCTORS
1072 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1073 
1074 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1075 		(*fn)();
1076 #endif
1077 }
1078 
1079 #ifdef CONFIG_KALLSYMS
1080 struct blacklist_entry {
1081 	struct list_head next;
1082 	char *buf;
1083 };
1084 
1085 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1086 
initcall_blacklist(char * str)1087 static int __init initcall_blacklist(char *str)
1088 {
1089 	char *str_entry;
1090 	struct blacklist_entry *entry;
1091 
1092 	/* str argument is a comma-separated list of functions */
1093 	do {
1094 		str_entry = strsep(&str, ",");
1095 		if (str_entry) {
1096 			pr_debug("blacklisting initcall %s\n", str_entry);
1097 			entry = memblock_alloc(sizeof(*entry),
1098 					       SMP_CACHE_BYTES);
1099 			if (!entry)
1100 				panic("%s: Failed to allocate %zu bytes\n",
1101 				      __func__, sizeof(*entry));
1102 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
1103 						    SMP_CACHE_BYTES);
1104 			if (!entry->buf)
1105 				panic("%s: Failed to allocate %zu bytes\n",
1106 				      __func__, strlen(str_entry) + 1);
1107 			strcpy(entry->buf, str_entry);
1108 			list_add(&entry->next, &blacklisted_initcalls);
1109 		}
1110 	} while (str_entry);
1111 
1112 	return 1;
1113 }
1114 
initcall_blacklisted(initcall_t fn)1115 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1116 {
1117 	struct blacklist_entry *entry;
1118 	char fn_name[KSYM_SYMBOL_LEN];
1119 	unsigned long addr;
1120 
1121 	if (list_empty(&blacklisted_initcalls))
1122 		return false;
1123 
1124 	addr = (unsigned long) dereference_function_descriptor(fn);
1125 	sprint_symbol_no_offset(fn_name, addr);
1126 
1127 	/*
1128 	 * fn will be "function_name [module_name]" where [module_name] is not
1129 	 * displayed for built-in init functions.  Strip off the [module_name].
1130 	 */
1131 	strreplace(fn_name, ' ', '\0');
1132 
1133 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
1134 		if (!strcmp(fn_name, entry->buf)) {
1135 			pr_debug("initcall %s blacklisted\n", fn_name);
1136 			return true;
1137 		}
1138 	}
1139 
1140 	return false;
1141 }
1142 #else
initcall_blacklist(char * str)1143 static int __init initcall_blacklist(char *str)
1144 {
1145 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1146 	return 0;
1147 }
1148 
initcall_blacklisted(initcall_t fn)1149 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1150 {
1151 	return false;
1152 }
1153 #endif
1154 __setup("initcall_blacklist=", initcall_blacklist);
1155 
1156 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1157 trace_initcall_start_cb(void *data, initcall_t fn)
1158 {
1159 	ktime_t *calltime = (ktime_t *)data;
1160 
1161 	printk(KERN_DEBUG "calling  %pS @ %i\n", fn, task_pid_nr(current));
1162 	*calltime = ktime_get();
1163 }
1164 
1165 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1166 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1167 {
1168 	ktime_t *calltime = (ktime_t *)data;
1169 	ktime_t delta, rettime;
1170 	unsigned long long duration;
1171 
1172 	rettime = ktime_get();
1173 	delta = ktime_sub(rettime, *calltime);
1174 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1175 	printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1176 		 fn, ret, duration);
1177 }
1178 
1179 static ktime_t initcall_calltime;
1180 
1181 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1182 static void __init initcall_debug_enable(void)
1183 {
1184 	int ret;
1185 
1186 	ret = register_trace_initcall_start(trace_initcall_start_cb,
1187 					    &initcall_calltime);
1188 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1189 					      &initcall_calltime);
1190 	WARN(ret, "Failed to register initcall tracepoints\n");
1191 }
1192 # define do_trace_initcall_start	trace_initcall_start
1193 # define do_trace_initcall_finish	trace_initcall_finish
1194 #else
do_trace_initcall_start(initcall_t fn)1195 static inline void do_trace_initcall_start(initcall_t fn)
1196 {
1197 	if (!initcall_debug)
1198 		return;
1199 	trace_initcall_start_cb(&initcall_calltime, fn);
1200 }
do_trace_initcall_finish(initcall_t fn,int ret)1201 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1202 {
1203 	if (!initcall_debug)
1204 		return;
1205 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1206 }
1207 #endif /* !TRACEPOINTS_ENABLED */
1208 
do_one_initcall(initcall_t fn)1209 int __init_or_module do_one_initcall(initcall_t fn)
1210 {
1211 	int count = preempt_count();
1212 	char msgbuf[64];
1213 	int ret;
1214 
1215 	if (initcall_blacklisted(fn))
1216 		return -EPERM;
1217 
1218 	do_trace_initcall_start(fn);
1219 	ret = fn();
1220 	do_trace_initcall_finish(fn, ret);
1221 
1222 	msgbuf[0] = 0;
1223 
1224 	if (preempt_count() != count) {
1225 		sprintf(msgbuf, "preemption imbalance ");
1226 		preempt_count_set(count);
1227 	}
1228 	if (irqs_disabled()) {
1229 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1230 		local_irq_enable();
1231 	}
1232 	WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1233 
1234 	add_latent_entropy();
1235 	return ret;
1236 }
1237 
1238 
1239 extern initcall_entry_t __initcall_start[];
1240 extern initcall_entry_t __initcall0_start[];
1241 extern initcall_entry_t __initcall1_start[];
1242 extern initcall_entry_t __initcall2_start[];
1243 extern initcall_entry_t __initcall3_start[];
1244 extern initcall_entry_t __initcall4_start[];
1245 extern initcall_entry_t __initcall5_start[];
1246 extern initcall_entry_t __initcall6_start[];
1247 extern initcall_entry_t __initcall7_start[];
1248 extern initcall_entry_t __initcall_end[];
1249 
1250 static initcall_entry_t *initcall_levels[] __initdata = {
1251 	__initcall0_start,
1252 	__initcall1_start,
1253 	__initcall2_start,
1254 	__initcall3_start,
1255 	__initcall4_start,
1256 	__initcall5_start,
1257 	__initcall6_start,
1258 	__initcall7_start,
1259 	__initcall_end,
1260 };
1261 
1262 /* Keep these in sync with initcalls in include/linux/init.h */
1263 static const char *initcall_level_names[] __initdata = {
1264 	"pure",
1265 	"core",
1266 	"postcore",
1267 	"arch",
1268 	"subsys",
1269 	"fs",
1270 	"device",
1271 	"late",
1272 };
1273 
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1274 static int __init ignore_unknown_bootoption(char *param, char *val,
1275 			       const char *unused, void *arg)
1276 {
1277 	return 0;
1278 }
1279 
do_initcall_level(int level,char * command_line)1280 static void __init do_initcall_level(int level, char *command_line)
1281 {
1282 	initcall_entry_t *fn;
1283 
1284 	parse_args(initcall_level_names[level],
1285 		   command_line, __start___param,
1286 		   __stop___param - __start___param,
1287 		   level, level,
1288 		   NULL, ignore_unknown_bootoption);
1289 
1290 	trace_initcall_level(initcall_level_names[level]);
1291 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1292 		do_one_initcall(initcall_from_entry(fn));
1293 }
1294 
do_initcalls(void)1295 static void __init do_initcalls(void)
1296 {
1297 	int level;
1298 	size_t len = strlen(saved_command_line) + 1;
1299 	char *command_line;
1300 
1301 	command_line = kzalloc(len, GFP_KERNEL);
1302 	if (!command_line)
1303 		panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1304 
1305 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1306 		/* Parser modifies command_line, restore it each time */
1307 		strcpy(command_line, saved_command_line);
1308 		do_initcall_level(level, command_line);
1309 	}
1310 
1311 	kfree(command_line);
1312 }
1313 
1314 /*
1315  * Ok, the machine is now initialized. None of the devices
1316  * have been touched yet, but the CPU subsystem is up and
1317  * running, and memory and process management works.
1318  *
1319  * Now we can finally start doing some real work..
1320  */
do_basic_setup(void)1321 static void __init do_basic_setup(void)
1322 {
1323 	cpuset_init_smp();
1324 	driver_init();
1325 	init_irq_proc();
1326 	do_ctors();
1327 	usermodehelper_enable();
1328 	do_initcalls();
1329 }
1330 
do_pre_smp_initcalls(void)1331 static void __init do_pre_smp_initcalls(void)
1332 {
1333 	initcall_entry_t *fn;
1334 
1335 	trace_initcall_level("early");
1336 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
1337 		do_one_initcall(initcall_from_entry(fn));
1338 }
1339 
run_init_process(const char * init_filename)1340 static int run_init_process(const char *init_filename)
1341 {
1342 	const char *const *p;
1343 
1344 	argv_init[0] = init_filename;
1345 	pr_info("Run %s as init process\n", init_filename);
1346 	pr_debug("  with arguments:\n");
1347 	for (p = argv_init; *p; p++)
1348 		pr_debug("    %s\n", *p);
1349 	pr_debug("  with environment:\n");
1350 	for (p = envp_init; *p; p++)
1351 		pr_debug("    %s\n", *p);
1352 	return kernel_execve(init_filename, argv_init, envp_init);
1353 }
1354 
try_to_run_init_process(const char * init_filename)1355 static int try_to_run_init_process(const char *init_filename)
1356 {
1357 	int ret;
1358 
1359 	ret = run_init_process(init_filename);
1360 
1361 	if (ret && ret != -ENOENT) {
1362 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1363 		       init_filename, ret);
1364 	}
1365 
1366 	return ret;
1367 }
1368 
1369 static noinline void __init kernel_init_freeable(void);
1370 
1371 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1372 bool rodata_enabled __ro_after_init = true;
set_debug_rodata(char * str)1373 static int __init set_debug_rodata(char *str)
1374 {
1375 	if (strtobool(str, &rodata_enabled))
1376 		pr_warn("Invalid option string for rodata: '%s'\n", str);
1377 	return 1;
1378 }
1379 __setup("rodata=", set_debug_rodata);
1380 #endif
1381 
1382 #ifdef CONFIG_STRICT_KERNEL_RWX
mark_readonly(void)1383 static void mark_readonly(void)
1384 {
1385 	if (rodata_enabled) {
1386 		/*
1387 		 * load_module() results in W+X mappings, which are cleaned
1388 		 * up with call_rcu().  Let's make sure that queued work is
1389 		 * flushed so that we don't hit false positives looking for
1390 		 * insecure pages which are W+X.
1391 		 */
1392 		rcu_barrier();
1393 		mark_rodata_ro();
1394 		rodata_test();
1395 	} else
1396 		pr_info("Kernel memory protection disabled.\n");
1397 }
1398 #elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
mark_readonly(void)1399 static inline void mark_readonly(void)
1400 {
1401 	pr_warn("Kernel memory protection not selected by kernel config.\n");
1402 }
1403 #else
mark_readonly(void)1404 static inline void mark_readonly(void)
1405 {
1406 	pr_warn("This architecture does not have kernel memory protection.\n");
1407 }
1408 #endif
1409 
free_initmem(void)1410 void __weak free_initmem(void)
1411 {
1412 	free_initmem_default(POISON_FREE_INITMEM);
1413 }
1414 
kernel_init(void * unused)1415 static int __ref kernel_init(void *unused)
1416 {
1417 	int ret;
1418 
1419 	kernel_init_freeable();
1420 	/* need to finish all async __init code before freeing the memory */
1421 	async_synchronize_full();
1422 	kprobe_free_init_mem();
1423 	ftrace_free_init_mem();
1424 	kgdb_free_init_mem();
1425 	free_initmem();
1426 	mark_readonly();
1427 
1428 	/*
1429 	 * Kernel mappings are now finalized - update the userspace page-table
1430 	 * to finalize PTI.
1431 	 */
1432 	pti_finalize();
1433 
1434 	system_state = SYSTEM_RUNNING;
1435 	numa_default_policy();
1436 
1437 	rcu_end_inkernel_boot();
1438 
1439 	do_sysctl_args();
1440 
1441 	if (ramdisk_execute_command) {
1442 		ret = run_init_process(ramdisk_execute_command);
1443 		if (!ret)
1444 			return 0;
1445 		pr_err("Failed to execute %s (error %d)\n",
1446 		       ramdisk_execute_command, ret);
1447 	}
1448 
1449 	/*
1450 	 * We try each of these until one succeeds.
1451 	 *
1452 	 * The Bourne shell can be used instead of init if we are
1453 	 * trying to recover a really broken machine.
1454 	 */
1455 	if (execute_command) {
1456 		ret = run_init_process(execute_command);
1457 		if (!ret)
1458 			return 0;
1459 		panic("Requested init %s failed (error %d).",
1460 		      execute_command, ret);
1461 	}
1462 
1463 	if (CONFIG_DEFAULT_INIT[0] != '\0') {
1464 		ret = run_init_process(CONFIG_DEFAULT_INIT);
1465 		if (ret)
1466 			pr_err("Default init %s failed (error %d)\n",
1467 			       CONFIG_DEFAULT_INIT, ret);
1468 		else
1469 			return 0;
1470 	}
1471 
1472 	if (!try_to_run_init_process("/sbin/init") ||
1473 	    !try_to_run_init_process("/etc/init") ||
1474 	    !try_to_run_init_process("/bin/init") ||
1475 	    !try_to_run_init_process("/bin/sh"))
1476 		return 0;
1477 
1478 	panic("No working init found.  Try passing init= option to kernel. "
1479 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1480 }
1481 
1482 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1483 void __init console_on_rootfs(void)
1484 {
1485 	struct file *file = filp_open("/dev/console", O_RDWR, 0);
1486 
1487 	if (IS_ERR(file)) {
1488 		pr_err("Warning: unable to open an initial console.\n");
1489 		return;
1490 	}
1491 	init_dup(file);
1492 	init_dup(file);
1493 	init_dup(file);
1494 	fput(file);
1495 }
1496 
kernel_init_freeable(void)1497 static noinline void __init kernel_init_freeable(void)
1498 {
1499 	/*
1500 	 * Wait until kthreadd is all set-up.
1501 	 */
1502 	wait_for_completion(&kthreadd_done);
1503 
1504 	/* Now the scheduler is fully set up and can do blocking allocations */
1505 	gfp_allowed_mask = __GFP_BITS_MASK;
1506 
1507 	/*
1508 	 * init can allocate pages on any node
1509 	 */
1510 	set_mems_allowed(node_states[N_MEMORY]);
1511 
1512 	cad_pid = get_pid(task_pid(current));
1513 
1514 	smp_prepare_cpus(setup_max_cpus);
1515 
1516 	workqueue_init();
1517 
1518 	init_mm_internals();
1519 
1520 	rcu_init_tasks_generic();
1521 	do_pre_smp_initcalls();
1522 	lockup_detector_init();
1523 
1524 	smp_init();
1525 	sched_init_smp();
1526 
1527 	padata_init();
1528 	page_alloc_init_late();
1529 	/* Initialize page ext after all struct pages are initialized. */
1530 	page_ext_init();
1531 
1532 	do_basic_setup();
1533 
1534 	kunit_run_all_tests();
1535 
1536 	console_on_rootfs();
1537 
1538 	/*
1539 	 * check if there is an early userspace init.  If yes, let it do all
1540 	 * the work
1541 	 */
1542 	if (init_eaccess(ramdisk_execute_command) != 0) {
1543 		ramdisk_execute_command = NULL;
1544 		prepare_namespace();
1545 	}
1546 
1547 	/*
1548 	 * Ok, we have completed the initial bootup, and
1549 	 * we're essentially up and running. Get rid of the
1550 	 * initmem segments and start the user-mode stuff..
1551 	 *
1552 	 * rootfs is available now, try loading the public keys
1553 	 * and default modules
1554 	 */
1555 
1556 	integrity_load_keys();
1557 }
1558