1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/init/main.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
11 */
12
13 #define DEBUG /* Enable initcall_debug */
14
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/cgroup.h>
54 #include <linux/efi.h>
55 #include <linux/tick.h>
56 #include <linux/sched/isolation.h>
57 #include <linux/interrupt.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/delayacct.h>
60 #include <linux/unistd.h>
61 #include <linux/utsname.h>
62 #include <linux/rmap.h>
63 #include <linux/mempolicy.h>
64 #include <linux/key.h>
65 #include <linux/debug_locks.h>
66 #include <linux/debugobjects.h>
67 #include <linux/lockdep.h>
68 #include <linux/kmemleak.h>
69 #include <linux/padata.h>
70 #include <linux/pid_namespace.h>
71 #include <linux/device/driver.h>
72 #include <linux/kthread.h>
73 #include <linux/sched.h>
74 #include <linux/sched/init.h>
75 #include <linux/signal.h>
76 #include <linux/idr.h>
77 #include <linux/kgdb.h>
78 #include <linux/ftrace.h>
79 #include <linux/async.h>
80 #include <linux/shmem_fs.h>
81 #include <linux/slab.h>
82 #include <linux/perf_event.h>
83 #include <linux/ptrace.h>
84 #include <linux/pti.h>
85 #include <linux/blkdev.h>
86 #include <linux/sched/clock.h>
87 #include <linux/sched/task.h>
88 #include <linux/sched/task_stack.h>
89 #include <linux/context_tracking.h>
90 #include <linux/random.h>
91 #include <linux/moduleloader.h>
92 #include <linux/list.h>
93 #include <linux/integrity.h>
94 #include <linux/proc_ns.h>
95 #include <linux/io.h>
96 #include <linux/cache.h>
97 #include <linux/rodata_test.h>
98 #include <linux/jump_label.h>
99 #include <linux/kcsan.h>
100 #include <linux/init_syscalls.h>
101 #include <linux/stackdepot.h>
102 #include <linux/randomize_kstack.h>
103 #include <linux/pidfs.h>
104 #include <linux/ptdump.h>
105 #include <net/net_namespace.h>
106
107 #include <asm/io.h>
108 #include <asm/setup.h>
109 #include <asm/sections.h>
110 #include <asm/cacheflush.h>
111
112 #define CREATE_TRACE_POINTS
113 #include <trace/events/initcall.h>
114
115 #include <kunit/test.h>
116
117 static int kernel_init(void *);
118
119 /*
120 * Debug helper: via this flag we know that we are in 'early bootup code'
121 * where only the boot processor is running with IRQ disabled. This means
122 * two things - IRQ must not be enabled before the flag is cleared and some
123 * operations which are not allowed with IRQ disabled are allowed while the
124 * flag is set.
125 */
126 bool early_boot_irqs_disabled __read_mostly;
127
128 enum system_states system_state __read_mostly;
129 EXPORT_SYMBOL(system_state);
130
131 /*
132 * Boot command-line arguments
133 */
134 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
135 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
136
137 /* Default late time init is NULL. archs can override this later. */
138 void (*__initdata late_time_init)(void);
139
140 /* Untouched command line saved by arch-specific code. */
141 char __initdata boot_command_line[COMMAND_LINE_SIZE];
142 /* Untouched saved command line (eg. for /proc) */
143 char *saved_command_line __ro_after_init;
144 unsigned int saved_command_line_len __ro_after_init;
145 /* Command line for parameter parsing */
146 static char *static_command_line;
147 /* Untouched extra command line */
148 static char *extra_command_line;
149 /* Extra init arguments */
150 static char *extra_init_args;
151
152 #ifdef CONFIG_BOOT_CONFIG
153 /* Is bootconfig on command line? */
154 static bool bootconfig_found;
155 static size_t initargs_offs;
156 #else
157 # define bootconfig_found false
158 # define initargs_offs 0
159 #endif
160
161 static char *execute_command;
162 static char *ramdisk_execute_command = "/init";
163
164 /*
165 * Used to generate warnings if static_key manipulation functions are used
166 * before jump_label_init is called.
167 */
168 bool static_key_initialized __read_mostly;
169 EXPORT_SYMBOL_GPL(static_key_initialized);
170
171 /*
172 * If set, this is an indication to the drivers that reset the underlying
173 * device before going ahead with the initialization otherwise driver might
174 * rely on the BIOS and skip the reset operation.
175 *
176 * This is useful if kernel is booting in an unreliable environment.
177 * For ex. kdump situation where previous kernel has crashed, BIOS has been
178 * skipped and devices will be in unknown state.
179 */
180 unsigned int reset_devices;
181 EXPORT_SYMBOL(reset_devices);
182
set_reset_devices(char * str)183 static int __init set_reset_devices(char *str)
184 {
185 reset_devices = 1;
186 return 1;
187 }
188
189 __setup("reset_devices", set_reset_devices);
190
191 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
192 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
193 static const char *panic_later, *panic_param;
194
obsolete_checksetup(char * line)195 static bool __init obsolete_checksetup(char *line)
196 {
197 const struct obs_kernel_param *p;
198 bool had_early_param = false;
199
200 p = __setup_start;
201 do {
202 int n = strlen(p->str);
203 if (parameqn(line, p->str, n)) {
204 if (p->early) {
205 /* Already done in parse_early_param?
206 * (Needs exact match on param part).
207 * Keep iterating, as we can have early
208 * params and __setups of same names 8( */
209 if (line[n] == '\0' || line[n] == '=')
210 had_early_param = true;
211 } else if (!p->setup_func) {
212 pr_warn("Parameter %s is obsolete, ignored\n",
213 p->str);
214 return true;
215 } else if (p->setup_func(line + n))
216 return true;
217 }
218 p++;
219 } while (p < __setup_end);
220
221 return had_early_param;
222 }
223
224 /*
225 * This should be approx 2 Bo*oMips to start (note initial shift), and will
226 * still work even if initially too large, it will just take slightly longer
227 */
228 unsigned long loops_per_jiffy = (1<<12);
229 EXPORT_SYMBOL(loops_per_jiffy);
230
debug_kernel(char * str)231 static int __init debug_kernel(char *str)
232 {
233 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
234 return 0;
235 }
236
quiet_kernel(char * str)237 static int __init quiet_kernel(char *str)
238 {
239 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
240 return 0;
241 }
242
243 early_param("debug", debug_kernel);
244 early_param("quiet", quiet_kernel);
245
loglevel(char * str)246 static int __init loglevel(char *str)
247 {
248 int newlevel;
249
250 /*
251 * Only update loglevel value when a correct setting was passed,
252 * to prevent blind crashes (when loglevel being set to 0) that
253 * are quite hard to debug
254 */
255 if (get_option(&str, &newlevel)) {
256 console_loglevel = newlevel;
257 return 0;
258 }
259
260 return -EINVAL;
261 }
262
263 early_param("loglevel", loglevel);
264
265 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(size_t * _size)266 static void * __init get_boot_config_from_initrd(size_t *_size)
267 {
268 u32 size, csum;
269 char *data;
270 u32 *hdr;
271 int i;
272
273 if (!initrd_end)
274 return NULL;
275
276 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
277 /*
278 * Since Grub may align the size of initrd to 4, we must
279 * check the preceding 3 bytes as well.
280 */
281 for (i = 0; i < 4; i++) {
282 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
283 goto found;
284 data--;
285 }
286 return NULL;
287
288 found:
289 hdr = (u32 *)(data - 8);
290 size = le32_to_cpu(hdr[0]);
291 csum = le32_to_cpu(hdr[1]);
292
293 data = ((void *)hdr) - size;
294 if ((unsigned long)data < initrd_start) {
295 pr_err("bootconfig size %d is greater than initrd size %ld\n",
296 size, initrd_end - initrd_start);
297 return NULL;
298 }
299
300 if (xbc_calc_checksum(data, size) != csum) {
301 pr_err("bootconfig checksum failed\n");
302 return NULL;
303 }
304
305 /* Remove bootconfig from initramfs/initrd */
306 initrd_end = (unsigned long)data;
307 if (_size)
308 *_size = size;
309
310 return data;
311 }
312 #else
get_boot_config_from_initrd(size_t * _size)313 static void * __init get_boot_config_from_initrd(size_t *_size)
314 {
315 return NULL;
316 }
317 #endif
318
319 #ifdef CONFIG_BOOT_CONFIG
320
321 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
322
323 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
324
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)325 static int __init xbc_snprint_cmdline(char *buf, size_t size,
326 struct xbc_node *root)
327 {
328 struct xbc_node *knode, *vnode;
329 char *end = buf + size;
330 const char *val, *q;
331 int ret;
332
333 xbc_node_for_each_key_value(root, knode, val) {
334 ret = xbc_node_compose_key_after(root, knode,
335 xbc_namebuf, XBC_KEYLEN_MAX);
336 if (ret < 0)
337 return ret;
338
339 vnode = xbc_node_get_child(knode);
340 if (!vnode) {
341 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
342 if (ret < 0)
343 return ret;
344 buf += ret;
345 continue;
346 }
347 xbc_array_for_each_value(vnode, val) {
348 /*
349 * For prettier and more readable /proc/cmdline, only
350 * quote the value when necessary, i.e. when it contains
351 * whitespace.
352 */
353 q = strpbrk(val, " \t\r\n") ? "\"" : "";
354 ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
355 xbc_namebuf, q, val, q);
356 if (ret < 0)
357 return ret;
358 buf += ret;
359 }
360 }
361
362 return buf - (end - size);
363 }
364 #undef rest
365
366 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)367 static char * __init xbc_make_cmdline(const char *key)
368 {
369 struct xbc_node *root;
370 char *new_cmdline;
371 int ret, len = 0;
372
373 root = xbc_find_node(key);
374 if (!root)
375 return NULL;
376
377 /* Count required buffer size */
378 len = xbc_snprint_cmdline(NULL, 0, root);
379 if (len <= 0)
380 return NULL;
381
382 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
383 if (!new_cmdline) {
384 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
385 return NULL;
386 }
387
388 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
389 if (ret < 0 || ret > len) {
390 pr_err("Failed to print extra kernel cmdline.\n");
391 memblock_free(new_cmdline, len + 1);
392 return NULL;
393 }
394
395 return new_cmdline;
396 }
397
bootconfig_params(char * param,char * val,const char * unused,void * arg)398 static int __init bootconfig_params(char *param, char *val,
399 const char *unused, void *arg)
400 {
401 if (strcmp(param, "bootconfig") == 0) {
402 bootconfig_found = true;
403 }
404 return 0;
405 }
406
warn_bootconfig(char * str)407 static int __init warn_bootconfig(char *str)
408 {
409 /* The 'bootconfig' has been handled by bootconfig_params(). */
410 return 0;
411 }
412
setup_boot_config(void)413 static void __init setup_boot_config(void)
414 {
415 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
416 const char *msg, *data;
417 int pos, ret;
418 size_t size;
419 char *err;
420
421 /* Cut out the bootconfig data even if we have no bootconfig option */
422 data = get_boot_config_from_initrd(&size);
423 /* If there is no bootconfig in initrd, try embedded one. */
424 if (!data)
425 data = xbc_get_embedded_bootconfig(&size);
426
427 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
428 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
429 bootconfig_params);
430
431 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
432 return;
433
434 /* parse_args() stops at the next param of '--' and returns an address */
435 if (err)
436 initargs_offs = err - tmp_cmdline;
437
438 if (!data) {
439 /* If user intended to use bootconfig, show an error level message */
440 if (bootconfig_found)
441 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
442 else
443 pr_info("No bootconfig data provided, so skipping bootconfig");
444 return;
445 }
446
447 if (size >= XBC_DATA_MAX) {
448 pr_err("bootconfig size %ld greater than max size %d\n",
449 (long)size, XBC_DATA_MAX);
450 return;
451 }
452
453 ret = xbc_init(data, size, &msg, &pos);
454 if (ret < 0) {
455 if (pos < 0)
456 pr_err("Failed to init bootconfig: %s.\n", msg);
457 else
458 pr_err("Failed to parse bootconfig: %s at %d.\n",
459 msg, pos);
460 } else {
461 xbc_get_info(&ret, NULL);
462 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
463 /* keys starting with "kernel." are passed via cmdline */
464 extra_command_line = xbc_make_cmdline("kernel");
465 /* Also, "init." keys are init arguments */
466 extra_init_args = xbc_make_cmdline("init");
467 }
468 return;
469 }
470
exit_boot_config(void)471 static void __init exit_boot_config(void)
472 {
473 xbc_exit();
474 }
475
476 #else /* !CONFIG_BOOT_CONFIG */
477
setup_boot_config(void)478 static void __init setup_boot_config(void)
479 {
480 /* Remove bootconfig data from initrd */
481 get_boot_config_from_initrd(NULL);
482 }
483
warn_bootconfig(char * str)484 static int __init warn_bootconfig(char *str)
485 {
486 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
487 return 0;
488 }
489
490 #define exit_boot_config() do {} while (0)
491
492 #endif /* CONFIG_BOOT_CONFIG */
493
494 early_param("bootconfig", warn_bootconfig);
495
cmdline_has_extra_options(void)496 bool __init cmdline_has_extra_options(void)
497 {
498 return extra_command_line || extra_init_args;
499 }
500
501 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)502 static void __init repair_env_string(char *param, char *val)
503 {
504 if (val) {
505 /* param=val or param="val"? */
506 if (val == param+strlen(param)+1)
507 val[-1] = '=';
508 else if (val == param+strlen(param)+2) {
509 val[-2] = '=';
510 memmove(val-1, val, strlen(val)+1);
511 } else
512 BUG();
513 }
514 }
515
516 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)517 static int __init set_init_arg(char *param, char *val,
518 const char *unused, void *arg)
519 {
520 unsigned int i;
521
522 if (panic_later)
523 return 0;
524
525 repair_env_string(param, val);
526
527 for (i = 0; argv_init[i]; i++) {
528 if (i == MAX_INIT_ARGS) {
529 panic_later = "init";
530 panic_param = param;
531 return 0;
532 }
533 }
534 argv_init[i] = param;
535 return 0;
536 }
537
538 /*
539 * Unknown boot options get handed to init, unless they look like
540 * unused parameters (modprobe will find them in /proc/cmdline).
541 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)542 static int __init unknown_bootoption(char *param, char *val,
543 const char *unused, void *arg)
544 {
545 size_t len = strlen(param);
546
547 /* Handle params aliased to sysctls */
548 if (sysctl_is_alias(param))
549 return 0;
550
551 repair_env_string(param, val);
552
553 /* Handle obsolete-style parameters */
554 if (obsolete_checksetup(param))
555 return 0;
556
557 /* Unused module parameter. */
558 if (strnchr(param, len, '.'))
559 return 0;
560
561 if (panic_later)
562 return 0;
563
564 if (val) {
565 /* Environment option */
566 unsigned int i;
567 for (i = 0; envp_init[i]; i++) {
568 if (i == MAX_INIT_ENVS) {
569 panic_later = "env";
570 panic_param = param;
571 }
572 if (!strncmp(param, envp_init[i], len+1))
573 break;
574 }
575 envp_init[i] = param;
576 } else {
577 /* Command line option */
578 unsigned int i;
579 for (i = 0; argv_init[i]; i++) {
580 if (i == MAX_INIT_ARGS) {
581 panic_later = "init";
582 panic_param = param;
583 }
584 }
585 argv_init[i] = param;
586 }
587 return 0;
588 }
589
init_setup(char * str)590 static int __init init_setup(char *str)
591 {
592 unsigned int i;
593
594 execute_command = str;
595 /*
596 * In case LILO is going to boot us with default command line,
597 * it prepends "auto" before the whole cmdline which makes
598 * the shell think it should execute a script with such name.
599 * So we ignore all arguments entered _before_ init=... [MJ]
600 */
601 for (i = 1; i < MAX_INIT_ARGS; i++)
602 argv_init[i] = NULL;
603 return 1;
604 }
605 __setup("init=", init_setup);
606
rdinit_setup(char * str)607 static int __init rdinit_setup(char *str)
608 {
609 unsigned int i;
610
611 ramdisk_execute_command = str;
612 /* See "auto" comment in init_setup */
613 for (i = 1; i < MAX_INIT_ARGS; i++)
614 argv_init[i] = NULL;
615 return 1;
616 }
617 __setup("rdinit=", rdinit_setup);
618
619 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)620 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)621 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
622 #endif
623
624 /*
625 * We need to store the untouched command line for future reference.
626 * We also need to store the touched command line since the parameter
627 * parsing is performed in place, and we should allow a component to
628 * store reference of name/value for future reference.
629 */
setup_command_line(char * command_line)630 static void __init setup_command_line(char *command_line)
631 {
632 size_t len, xlen = 0, ilen = 0;
633
634 if (extra_command_line)
635 xlen = strlen(extra_command_line);
636 if (extra_init_args) {
637 extra_init_args = strim(extra_init_args); /* remove trailing space */
638 ilen = strlen(extra_init_args) + 4; /* for " -- " */
639 }
640
641 len = xlen + strlen(boot_command_line) + ilen + 1;
642
643 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
644 if (!saved_command_line)
645 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
646
647 len = xlen + strlen(command_line) + 1;
648
649 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
650 if (!static_command_line)
651 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
652
653 if (xlen) {
654 /*
655 * We have to put extra_command_line before boot command
656 * lines because there could be dashes (separator of init
657 * command line) in the command lines.
658 */
659 strcpy(saved_command_line, extra_command_line);
660 strcpy(static_command_line, extra_command_line);
661 }
662 strcpy(saved_command_line + xlen, boot_command_line);
663 strcpy(static_command_line + xlen, command_line);
664
665 if (ilen) {
666 /*
667 * Append supplemental init boot args to saved_command_line
668 * so that user can check what command line options passed
669 * to init.
670 * The order should always be
671 * " -- "[bootconfig init-param][cmdline init-param]
672 */
673 if (initargs_offs) {
674 len = xlen + initargs_offs;
675 strcpy(saved_command_line + len, extra_init_args);
676 len += ilen - 4; /* strlen(extra_init_args) */
677 strcpy(saved_command_line + len,
678 boot_command_line + initargs_offs - 1);
679 } else {
680 len = strlen(saved_command_line);
681 strcpy(saved_command_line + len, " -- ");
682 len += 4;
683 strcpy(saved_command_line + len, extra_init_args);
684 }
685 }
686
687 saved_command_line_len = strlen(saved_command_line);
688 }
689
690 /*
691 * We need to finalize in a non-__init function or else race conditions
692 * between the root thread and the init thread may cause start_kernel to
693 * be reaped by free_initmem before the root thread has proceeded to
694 * cpu_idle.
695 *
696 * gcc-3.4 accidentally inlines this function, so use noinline.
697 */
698
699 static __initdata DECLARE_COMPLETION(kthreadd_done);
700
rest_init(void)701 static noinline void __ref __noreturn rest_init(void)
702 {
703 struct task_struct *tsk;
704 int pid;
705
706 rcu_scheduler_starting();
707 /*
708 * We need to spawn init first so that it obtains pid 1, however
709 * the init task will end up wanting to create kthreads, which, if
710 * we schedule it before we create kthreadd, will OOPS.
711 */
712 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
713 /*
714 * Pin init on the boot CPU. Task migration is not properly working
715 * until sched_init_smp() has been run. It will set the allowed
716 * CPUs for init to the non isolated CPUs.
717 */
718 rcu_read_lock();
719 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
720 tsk->flags |= PF_NO_SETAFFINITY;
721 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
722 rcu_read_unlock();
723
724 numa_default_policy();
725 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
726 rcu_read_lock();
727 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
728 rcu_read_unlock();
729
730 /*
731 * Enable might_sleep() and smp_processor_id() checks.
732 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
733 * kernel_thread() would trigger might_sleep() splats. With
734 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
735 * already, but it's stuck on the kthreadd_done completion.
736 */
737 system_state = SYSTEM_SCHEDULING;
738
739 complete(&kthreadd_done);
740
741 /*
742 * The boot idle thread must execute schedule()
743 * at least once to get things moving:
744 */
745 schedule_preempt_disabled();
746 /* Call into cpu_idle with preempt disabled */
747 cpu_startup_entry(CPUHP_ONLINE);
748 }
749
750 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)751 static int __init do_early_param(char *param, char *val,
752 const char *unused, void *arg)
753 {
754 const struct obs_kernel_param *p;
755
756 for (p = __setup_start; p < __setup_end; p++) {
757 if ((p->early && parameq(param, p->str)) ||
758 (strcmp(param, "console") == 0 &&
759 strcmp(p->str, "earlycon") == 0)
760 ) {
761 if (p->setup_func(val) != 0)
762 pr_warn("Malformed early option '%s'\n", param);
763 }
764 }
765 /* We accept everything at this stage. */
766 return 0;
767 }
768
parse_early_options(char * cmdline)769 void __init parse_early_options(char *cmdline)
770 {
771 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
772 do_early_param);
773 }
774
775 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)776 void __init parse_early_param(void)
777 {
778 static int done __initdata;
779 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
780
781 if (done)
782 return;
783
784 /* All fall through to do_early_param. */
785 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
786 parse_early_options(tmp_cmdline);
787 done = 1;
788 }
789
arch_post_acpi_subsys_init(void)790 void __init __weak arch_post_acpi_subsys_init(void) { }
791
smp_setup_processor_id(void)792 void __init __weak smp_setup_processor_id(void)
793 {
794 }
795
smp_prepare_boot_cpu(void)796 void __init __weak smp_prepare_boot_cpu(void)
797 {
798 }
799
800 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)801 void __init __weak thread_stack_cache_init(void)
802 {
803 }
804 #endif
805
poking_init(void)806 void __init __weak poking_init(void) { }
807
pgtable_cache_init(void)808 void __init __weak pgtable_cache_init(void) { }
809
trap_init(void)810 void __init __weak trap_init(void) { }
811
812 bool initcall_debug;
813 core_param(initcall_debug, initcall_debug, bool, 0644);
814
815 #ifdef TRACEPOINTS_ENABLED
816 static void __init initcall_debug_enable(void);
817 #else
initcall_debug_enable(void)818 static inline void initcall_debug_enable(void)
819 {
820 }
821 #endif
822
823 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
824 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
825 randomize_kstack_offset);
826 DEFINE_PER_CPU(u32, kstack_offset);
827
early_randomize_kstack_offset(char * buf)828 static int __init early_randomize_kstack_offset(char *buf)
829 {
830 int ret;
831 bool bool_result;
832
833 ret = kstrtobool(buf, &bool_result);
834 if (ret)
835 return ret;
836
837 if (bool_result)
838 static_branch_enable(&randomize_kstack_offset);
839 else
840 static_branch_disable(&randomize_kstack_offset);
841 return 0;
842 }
843 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
844 #endif
845
print_unknown_bootoptions(void)846 static void __init print_unknown_bootoptions(void)
847 {
848 char *unknown_options;
849 char *end;
850 const char *const *p;
851 size_t len;
852
853 if (panic_later || (!argv_init[1] && !envp_init[2]))
854 return;
855
856 /*
857 * Determine how many options we have to print out, plus a space
858 * before each
859 */
860 len = 1; /* null terminator */
861 for (p = &argv_init[1]; *p; p++) {
862 len++;
863 len += strlen(*p);
864 }
865 for (p = &envp_init[2]; *p; p++) {
866 len++;
867 len += strlen(*p);
868 }
869
870 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
871 if (!unknown_options) {
872 pr_err("%s: Failed to allocate %zu bytes\n",
873 __func__, len);
874 return;
875 }
876 end = unknown_options;
877
878 for (p = &argv_init[1]; *p; p++)
879 end += sprintf(end, " %s", *p);
880 for (p = &envp_init[2]; *p; p++)
881 end += sprintf(end, " %s", *p);
882
883 /* Start at unknown_options[1] to skip the initial space */
884 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
885 &unknown_options[1]);
886 memblock_free(unknown_options, len);
887 }
888
early_numa_node_init(void)889 static void __init early_numa_node_init(void)
890 {
891 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
892 #ifndef cpu_to_node
893 int cpu;
894
895 /* The early_cpu_to_node() should be ready here. */
896 for_each_possible_cpu(cpu)
897 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
898 #endif
899 #endif
900 }
901
902 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)903 void start_kernel(void)
904 {
905 char *command_line;
906 char *after_dashes;
907
908 set_task_stack_end_magic(&init_task);
909 smp_setup_processor_id();
910 debug_objects_early_init();
911 init_vmlinux_build_id();
912
913 cgroup_init_early();
914
915 local_irq_disable();
916 early_boot_irqs_disabled = true;
917
918 /*
919 * Interrupts are still disabled. Do necessary setups, then
920 * enable them.
921 */
922 boot_cpu_init();
923 page_address_init();
924 pr_notice("%s", linux_banner);
925 early_security_init();
926 setup_arch(&command_line);
927 setup_boot_config();
928 setup_command_line(command_line);
929 setup_nr_cpu_ids();
930 setup_per_cpu_areas();
931 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
932 early_numa_node_init();
933 boot_cpu_hotplug_init();
934
935 pr_notice("Kernel command line: %s\n", saved_command_line);
936 /* parameters may set static keys */
937 jump_label_init();
938 parse_early_param();
939 after_dashes = parse_args("Booting kernel",
940 static_command_line, __start___param,
941 __stop___param - __start___param,
942 -1, -1, NULL, &unknown_bootoption);
943 print_unknown_bootoptions();
944 if (!IS_ERR_OR_NULL(after_dashes))
945 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
946 NULL, set_init_arg);
947 if (extra_init_args)
948 parse_args("Setting extra init args", extra_init_args,
949 NULL, 0, -1, -1, NULL, set_init_arg);
950
951 /* Architectural and non-timekeeping rng init, before allocator init */
952 random_init_early(command_line);
953
954 /*
955 * These use large bootmem allocations and must precede
956 * initalization of page allocator
957 */
958 setup_log_buf(0);
959 vfs_caches_init_early();
960 sort_main_extable();
961 trap_init();
962 mm_core_init();
963 poking_init();
964 ftrace_init();
965
966 /* trace_printk can be enabled here */
967 early_trace_init();
968
969 /*
970 * Set up the scheduler prior starting any interrupts (such as the
971 * timer interrupt). Full topology setup happens at smp_init()
972 * time - but meanwhile we still have a functioning scheduler.
973 */
974 sched_init();
975
976 if (WARN(!irqs_disabled(),
977 "Interrupts were enabled *very* early, fixing it\n"))
978 local_irq_disable();
979 radix_tree_init();
980 maple_tree_init();
981
982 /*
983 * Set up housekeeping before setting up workqueues to allow the unbound
984 * workqueue to take non-housekeeping into account.
985 */
986 housekeeping_init();
987
988 /*
989 * Allow workqueue creation and work item queueing/cancelling
990 * early. Work item execution depends on kthreads and starts after
991 * workqueue_init().
992 */
993 workqueue_init_early();
994
995 rcu_init();
996
997 /* Trace events are available after this */
998 trace_init();
999
1000 if (initcall_debug)
1001 initcall_debug_enable();
1002
1003 context_tracking_init();
1004 /* init some links before init_ISA_irqs() */
1005 early_irq_init();
1006 init_IRQ();
1007 tick_init();
1008 rcu_init_nohz();
1009 init_timers();
1010 srcu_init();
1011 hrtimers_init();
1012 softirq_init();
1013 timekeeping_init();
1014 time_init();
1015
1016 /* This must be after timekeeping is initialized */
1017 random_init();
1018
1019 /* These make use of the fully initialized rng */
1020 kfence_init();
1021 boot_init_stack_canary();
1022
1023 perf_event_init();
1024 profile_init();
1025 call_function_init();
1026 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1027
1028 early_boot_irqs_disabled = false;
1029 local_irq_enable();
1030
1031 kmem_cache_init_late();
1032
1033 /*
1034 * HACK ALERT! This is early. We're enabling the console before
1035 * we've done PCI setups etc, and console_init() must be aware of
1036 * this. But we do want output early, in case something goes wrong.
1037 */
1038 console_init();
1039 if (panic_later)
1040 panic("Too many boot %s vars at `%s'", panic_later,
1041 panic_param);
1042
1043 lockdep_init();
1044
1045 /*
1046 * Need to run this when irqs are enabled, because it wants
1047 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1048 * too:
1049 */
1050 locking_selftest();
1051
1052 #ifdef CONFIG_BLK_DEV_INITRD
1053 if (initrd_start && !initrd_below_start_ok &&
1054 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1055 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1056 page_to_pfn(virt_to_page((void *)initrd_start)),
1057 min_low_pfn);
1058 initrd_start = 0;
1059 }
1060 #endif
1061 setup_per_cpu_pageset();
1062 numa_policy_init();
1063 acpi_early_init();
1064 if (late_time_init)
1065 late_time_init();
1066 sched_clock_init();
1067 calibrate_delay();
1068
1069 arch_cpu_finalize_init();
1070
1071 pid_idr_init();
1072 anon_vma_init();
1073 #ifdef CONFIG_X86
1074 if (efi_enabled(EFI_RUNTIME_SERVICES))
1075 efi_enter_virtual_mode();
1076 #endif
1077 thread_stack_cache_init();
1078 cred_init();
1079 fork_init();
1080 proc_caches_init();
1081 uts_ns_init();
1082 key_init();
1083 security_init();
1084 dbg_late_init();
1085 net_ns_init();
1086 vfs_caches_init();
1087 pagecache_init();
1088 signals_init();
1089 seq_file_init();
1090 proc_root_init();
1091 nsfs_init();
1092 pidfs_init();
1093 cpuset_init();
1094 cgroup_init();
1095 taskstats_init_early();
1096 delayacct_init();
1097
1098 acpi_subsystem_init();
1099 arch_post_acpi_subsys_init();
1100 kcsan_init();
1101
1102 /* Do the rest non-__init'ed, we're now alive */
1103 rest_init();
1104
1105 /*
1106 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1107 * and older.
1108 */
1109 #if !__has_attribute(__no_stack_protector__)
1110 prevent_tail_call_optimization();
1111 #endif
1112 }
1113
1114 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1115 static void __init do_ctors(void)
1116 {
1117 /*
1118 * For UML, the constructors have already been called by the
1119 * normal setup code as it's just a normal ELF binary, so we
1120 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1121 * even on UML for modules.
1122 */
1123 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1124 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1125
1126 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1127 (*fn)();
1128 #endif
1129 }
1130
1131 #ifdef CONFIG_KALLSYMS
1132 struct blacklist_entry {
1133 struct list_head next;
1134 char *buf;
1135 };
1136
1137 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1138
initcall_blacklist(char * str)1139 static int __init initcall_blacklist(char *str)
1140 {
1141 char *str_entry;
1142 struct blacklist_entry *entry;
1143
1144 /* str argument is a comma-separated list of functions */
1145 do {
1146 str_entry = strsep(&str, ",");
1147 if (str_entry) {
1148 pr_debug("blacklisting initcall %s\n", str_entry);
1149 entry = memblock_alloc(sizeof(*entry),
1150 SMP_CACHE_BYTES);
1151 if (!entry)
1152 panic("%s: Failed to allocate %zu bytes\n",
1153 __func__, sizeof(*entry));
1154 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1155 SMP_CACHE_BYTES);
1156 if (!entry->buf)
1157 panic("%s: Failed to allocate %zu bytes\n",
1158 __func__, strlen(str_entry) + 1);
1159 strcpy(entry->buf, str_entry);
1160 list_add(&entry->next, &blacklisted_initcalls);
1161 }
1162 } while (str_entry);
1163
1164 return 1;
1165 }
1166
initcall_blacklisted(initcall_t fn)1167 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1168 {
1169 struct blacklist_entry *entry;
1170 char fn_name[KSYM_SYMBOL_LEN];
1171 unsigned long addr;
1172
1173 if (list_empty(&blacklisted_initcalls))
1174 return false;
1175
1176 addr = (unsigned long) dereference_function_descriptor(fn);
1177 sprint_symbol_no_offset(fn_name, addr);
1178
1179 /*
1180 * fn will be "function_name [module_name]" where [module_name] is not
1181 * displayed for built-in init functions. Strip off the [module_name].
1182 */
1183 strreplace(fn_name, ' ', '\0');
1184
1185 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1186 if (!strcmp(fn_name, entry->buf)) {
1187 pr_debug("initcall %s blacklisted\n", fn_name);
1188 return true;
1189 }
1190 }
1191
1192 return false;
1193 }
1194 #else
initcall_blacklist(char * str)1195 static int __init initcall_blacklist(char *str)
1196 {
1197 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1198 return 0;
1199 }
1200
initcall_blacklisted(initcall_t fn)1201 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1202 {
1203 return false;
1204 }
1205 #endif
1206 __setup("initcall_blacklist=", initcall_blacklist);
1207
1208 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1209 trace_initcall_start_cb(void *data, initcall_t fn)
1210 {
1211 ktime_t *calltime = data;
1212
1213 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1214 *calltime = ktime_get();
1215 }
1216
1217 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1218 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1219 {
1220 ktime_t rettime, *calltime = data;
1221
1222 rettime = ktime_get();
1223 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1224 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1225 }
1226
1227 static ktime_t initcall_calltime;
1228
1229 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1230 static void __init initcall_debug_enable(void)
1231 {
1232 int ret;
1233
1234 ret = register_trace_initcall_start(trace_initcall_start_cb,
1235 &initcall_calltime);
1236 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1237 &initcall_calltime);
1238 WARN(ret, "Failed to register initcall tracepoints\n");
1239 }
1240 # define do_trace_initcall_start trace_initcall_start
1241 # define do_trace_initcall_finish trace_initcall_finish
1242 #else
do_trace_initcall_start(initcall_t fn)1243 static inline void do_trace_initcall_start(initcall_t fn)
1244 {
1245 if (!initcall_debug)
1246 return;
1247 trace_initcall_start_cb(&initcall_calltime, fn);
1248 }
do_trace_initcall_finish(initcall_t fn,int ret)1249 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1250 {
1251 if (!initcall_debug)
1252 return;
1253 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1254 }
1255 #endif /* !TRACEPOINTS_ENABLED */
1256
do_one_initcall(initcall_t fn)1257 int __init_or_module do_one_initcall(initcall_t fn)
1258 {
1259 int count = preempt_count();
1260 char msgbuf[64];
1261 int ret;
1262
1263 if (initcall_blacklisted(fn))
1264 return -EPERM;
1265
1266 do_trace_initcall_start(fn);
1267 ret = fn();
1268 do_trace_initcall_finish(fn, ret);
1269
1270 msgbuf[0] = 0;
1271
1272 if (preempt_count() != count) {
1273 sprintf(msgbuf, "preemption imbalance ");
1274 preempt_count_set(count);
1275 }
1276 if (irqs_disabled()) {
1277 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1278 local_irq_enable();
1279 }
1280 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1281
1282 add_latent_entropy();
1283 return ret;
1284 }
1285
1286
1287 static initcall_entry_t *initcall_levels[] __initdata = {
1288 __initcall0_start,
1289 __initcall1_start,
1290 __initcall2_start,
1291 __initcall3_start,
1292 __initcall4_start,
1293 __initcall5_start,
1294 __initcall6_start,
1295 __initcall7_start,
1296 __initcall_end,
1297 };
1298
1299 /* Keep these in sync with initcalls in include/linux/init.h */
1300 static const char *initcall_level_names[] __initdata = {
1301 "pure",
1302 "core",
1303 "postcore",
1304 "arch",
1305 "subsys",
1306 "fs",
1307 "device",
1308 "late",
1309 };
1310
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1311 static int __init ignore_unknown_bootoption(char *param, char *val,
1312 const char *unused, void *arg)
1313 {
1314 return 0;
1315 }
1316
do_initcall_level(int level,char * command_line)1317 static void __init do_initcall_level(int level, char *command_line)
1318 {
1319 initcall_entry_t *fn;
1320
1321 parse_args(initcall_level_names[level],
1322 command_line, __start___param,
1323 __stop___param - __start___param,
1324 level, level,
1325 NULL, ignore_unknown_bootoption);
1326
1327 trace_initcall_level(initcall_level_names[level]);
1328 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1329 do_one_initcall(initcall_from_entry(fn));
1330 }
1331
do_initcalls(void)1332 static void __init do_initcalls(void)
1333 {
1334 int level;
1335 size_t len = saved_command_line_len + 1;
1336 char *command_line;
1337
1338 command_line = kzalloc(len, GFP_KERNEL);
1339 if (!command_line)
1340 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1341
1342 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1343 /* Parser modifies command_line, restore it each time */
1344 strcpy(command_line, saved_command_line);
1345 do_initcall_level(level, command_line);
1346 }
1347
1348 kfree(command_line);
1349 }
1350
1351 /*
1352 * Ok, the machine is now initialized. None of the devices
1353 * have been touched yet, but the CPU subsystem is up and
1354 * running, and memory and process management works.
1355 *
1356 * Now we can finally start doing some real work..
1357 */
do_basic_setup(void)1358 static void __init do_basic_setup(void)
1359 {
1360 cpuset_init_smp();
1361 driver_init();
1362 init_irq_proc();
1363 do_ctors();
1364 do_initcalls();
1365 }
1366
do_pre_smp_initcalls(void)1367 static void __init do_pre_smp_initcalls(void)
1368 {
1369 initcall_entry_t *fn;
1370
1371 trace_initcall_level("early");
1372 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1373 do_one_initcall(initcall_from_entry(fn));
1374 }
1375
run_init_process(const char * init_filename)1376 static int run_init_process(const char *init_filename)
1377 {
1378 const char *const *p;
1379
1380 argv_init[0] = init_filename;
1381 pr_info("Run %s as init process\n", init_filename);
1382 pr_debug(" with arguments:\n");
1383 for (p = argv_init; *p; p++)
1384 pr_debug(" %s\n", *p);
1385 pr_debug(" with environment:\n");
1386 for (p = envp_init; *p; p++)
1387 pr_debug(" %s\n", *p);
1388 return kernel_execve(init_filename, argv_init, envp_init);
1389 }
1390
try_to_run_init_process(const char * init_filename)1391 static int try_to_run_init_process(const char *init_filename)
1392 {
1393 int ret;
1394
1395 ret = run_init_process(init_filename);
1396
1397 if (ret && ret != -ENOENT) {
1398 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1399 init_filename, ret);
1400 }
1401
1402 return ret;
1403 }
1404
1405 static noinline void __init kernel_init_freeable(void);
1406
1407 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1408 bool rodata_enabled __ro_after_init = true;
1409
1410 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1411 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1412 #endif
1413
set_debug_rodata(char * str)1414 static int __init set_debug_rodata(char *str)
1415 {
1416 if (arch_parse_debug_rodata(str))
1417 return 0;
1418
1419 if (str && !strcmp(str, "on"))
1420 rodata_enabled = true;
1421 else if (str && !strcmp(str, "off"))
1422 rodata_enabled = false;
1423 else
1424 pr_warn("Invalid option string for rodata: '%s'\n", str);
1425 return 0;
1426 }
1427 early_param("rodata", set_debug_rodata);
1428 #endif
1429
mark_readonly(void)1430 static void mark_readonly(void)
1431 {
1432 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1433 /*
1434 * load_module() results in W+X mappings, which are cleaned
1435 * up with init_free_wq. Let's make sure that queued work is
1436 * flushed so that we don't hit false positives looking for
1437 * insecure pages which are W+X.
1438 */
1439 flush_module_init_free_work();
1440 jump_label_init_ro();
1441 mark_rodata_ro();
1442 debug_checkwx();
1443 rodata_test();
1444 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1445 pr_info("Kernel memory protection disabled.\n");
1446 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1447 pr_warn("Kernel memory protection not selected by kernel config.\n");
1448 } else {
1449 pr_warn("This architecture does not have kernel memory protection.\n");
1450 }
1451 }
1452
free_initmem(void)1453 void __weak free_initmem(void)
1454 {
1455 free_initmem_default(POISON_FREE_INITMEM);
1456 }
1457
kernel_init(void * unused)1458 static int __ref kernel_init(void *unused)
1459 {
1460 int ret;
1461
1462 /*
1463 * Wait until kthreadd is all set-up.
1464 */
1465 wait_for_completion(&kthreadd_done);
1466
1467 kernel_init_freeable();
1468 /* need to finish all async __init code before freeing the memory */
1469 async_synchronize_full();
1470
1471 system_state = SYSTEM_FREEING_INITMEM;
1472 kprobe_free_init_mem();
1473 ftrace_free_init_mem();
1474 kgdb_free_init_mem();
1475 exit_boot_config();
1476 free_initmem();
1477 mark_readonly();
1478
1479 /*
1480 * Kernel mappings are now finalized - update the userspace page-table
1481 * to finalize PTI.
1482 */
1483 pti_finalize();
1484
1485 system_state = SYSTEM_RUNNING;
1486 numa_default_policy();
1487
1488 rcu_end_inkernel_boot();
1489
1490 do_sysctl_args();
1491
1492 if (ramdisk_execute_command) {
1493 ret = run_init_process(ramdisk_execute_command);
1494 if (!ret)
1495 return 0;
1496 pr_err("Failed to execute %s (error %d)\n",
1497 ramdisk_execute_command, ret);
1498 }
1499
1500 /*
1501 * We try each of these until one succeeds.
1502 *
1503 * The Bourne shell can be used instead of init if we are
1504 * trying to recover a really broken machine.
1505 */
1506 if (execute_command) {
1507 ret = run_init_process(execute_command);
1508 if (!ret)
1509 return 0;
1510 panic("Requested init %s failed (error %d).",
1511 execute_command, ret);
1512 }
1513
1514 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1515 ret = run_init_process(CONFIG_DEFAULT_INIT);
1516 if (ret)
1517 pr_err("Default init %s failed (error %d)\n",
1518 CONFIG_DEFAULT_INIT, ret);
1519 else
1520 return 0;
1521 }
1522
1523 if (!try_to_run_init_process("/sbin/init") ||
1524 !try_to_run_init_process("/etc/init") ||
1525 !try_to_run_init_process("/bin/init") ||
1526 !try_to_run_init_process("/bin/sh"))
1527 return 0;
1528
1529 panic("No working init found. Try passing init= option to kernel. "
1530 "See Linux Documentation/admin-guide/init.rst for guidance.");
1531 }
1532
1533 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1534 void __init console_on_rootfs(void)
1535 {
1536 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1537
1538 if (IS_ERR(file)) {
1539 pr_err("Warning: unable to open an initial console.\n");
1540 return;
1541 }
1542 init_dup(file);
1543 init_dup(file);
1544 init_dup(file);
1545 fput(file);
1546 }
1547
kernel_init_freeable(void)1548 static noinline void __init kernel_init_freeable(void)
1549 {
1550 /* Now the scheduler is fully set up and can do blocking allocations */
1551 gfp_allowed_mask = __GFP_BITS_MASK;
1552
1553 /*
1554 * init can allocate pages on any node
1555 */
1556 set_mems_allowed(node_states[N_MEMORY]);
1557
1558 cad_pid = get_pid(task_pid(current));
1559
1560 smp_prepare_cpus(setup_max_cpus);
1561
1562 workqueue_init();
1563
1564 init_mm_internals();
1565
1566 rcu_init_tasks_generic();
1567 do_pre_smp_initcalls();
1568 lockup_detector_init();
1569
1570 smp_init();
1571 sched_init_smp();
1572
1573 workqueue_init_topology();
1574 async_init();
1575 padata_init();
1576 page_alloc_init_late();
1577
1578 do_basic_setup();
1579
1580 kunit_run_all_tests();
1581
1582 wait_for_initramfs();
1583 console_on_rootfs();
1584
1585 /*
1586 * check if there is an early userspace init. If yes, let it do all
1587 * the work
1588 */
1589 if (init_eaccess(ramdisk_execute_command) != 0) {
1590 ramdisk_execute_command = NULL;
1591 prepare_namespace();
1592 }
1593
1594 /*
1595 * Ok, we have completed the initial bootup, and
1596 * we're essentially up and running. Get rid of the
1597 * initmem segments and start the user-mode stuff..
1598 *
1599 * rootfs is available now, try loading the public keys
1600 * and default modules
1601 */
1602
1603 integrity_load_keys();
1604 }
1605