1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/init/main.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
11 */
12
13 #define DEBUG /* Enable initcall_debug */
14
15 #include <linux/types.h>
16 #include <linux/export.h>
17 #include <linux/extable.h>
18 #include <linux/module.h>
19 #include <linux/proc_fs.h>
20 #include <linux/binfmts.h>
21 #include <linux/kernel.h>
22 #include <linux/syscalls.h>
23 #include <linux/stackprotector.h>
24 #include <linux/string.h>
25 #include <linux/ctype.h>
26 #include <linux/delay.h>
27 #include <linux/ioport.h>
28 #include <linux/init.h>
29 #include <linux/initrd.h>
30 #include <linux/memblock.h>
31 #include <linux/acpi.h>
32 #include <linux/bootconfig.h>
33 #include <linux/console.h>
34 #include <linux/nmi.h>
35 #include <linux/percpu.h>
36 #include <linux/kmod.h>
37 #include <linux/kprobes.h>
38 #include <linux/kmsan.h>
39 #include <linux/vmalloc.h>
40 #include <linux/kernel_stat.h>
41 #include <linux/start_kernel.h>
42 #include <linux/security.h>
43 #include <linux/smp.h>
44 #include <linux/profile.h>
45 #include <linux/kfence.h>
46 #include <linux/rcupdate.h>
47 #include <linux/srcu.h>
48 #include <linux/moduleparam.h>
49 #include <linux/kallsyms.h>
50 #include <linux/buildid.h>
51 #include <linux/writeback.h>
52 #include <linux/cpu.h>
53 #include <linux/cpuset.h>
54 #include <linux/memcontrol.h>
55 #include <linux/cgroup.h>
56 #include <linux/tick.h>
57 #include <linux/sched/isolation.h>
58 #include <linux/interrupt.h>
59 #include <linux/taskstats_kern.h>
60 #include <linux/delayacct.h>
61 #include <linux/unistd.h>
62 #include <linux/utsname.h>
63 #include <linux/rmap.h>
64 #include <linux/mempolicy.h>
65 #include <linux/key.h>
66 #include <linux/debug_locks.h>
67 #include <linux/debugobjects.h>
68 #include <linux/lockdep.h>
69 #include <linux/kmemleak.h>
70 #include <linux/padata.h>
71 #include <linux/pid_namespace.h>
72 #include <linux/device/driver.h>
73 #include <linux/kthread.h>
74 #include <linux/sched.h>
75 #include <linux/sched/init.h>
76 #include <linux/signal.h>
77 #include <linux/idr.h>
78 #include <linux/kgdb.h>
79 #include <linux/ftrace.h>
80 #include <linux/async.h>
81 #include <linux/shmem_fs.h>
82 #include <linux/slab.h>
83 #include <linux/perf_event.h>
84 #include <linux/ptrace.h>
85 #include <linux/pti.h>
86 #include <linux/blkdev.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/moduleloader.h>
93 #include <linux/list.h>
94 #include <linux/integrity.h>
95 #include <linux/proc_ns.h>
96 #include <linux/io.h>
97 #include <linux/cache.h>
98 #include <linux/rodata_test.h>
99 #include <linux/jump_label.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
103 #include <linux/randomize_kstack.h>
104 #include <linux/pidfs.h>
105 #include <linux/ptdump.h>
106 #include <net/net_namespace.h>
107
108 #include <asm/io.h>
109 #include <asm/setup.h>
110 #include <asm/sections.h>
111 #include <asm/cacheflush.h>
112
113 #define CREATE_TRACE_POINTS
114 #include <trace/events/initcall.h>
115
116 #include <kunit/test.h>
117
118 static int kernel_init(void *);
119
120 /*
121 * Debug helper: via this flag we know that we are in 'early bootup code'
122 * where only the boot processor is running with IRQ disabled. This means
123 * two things - IRQ must not be enabled before the flag is cleared and some
124 * operations which are not allowed with IRQ disabled are allowed while the
125 * flag is set.
126 */
127 bool early_boot_irqs_disabled __read_mostly;
128
129 enum system_states system_state __read_mostly;
130 EXPORT_SYMBOL(system_state);
131
132 /*
133 * Boot command-line arguments
134 */
135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
137
138 /* Default late time init is NULL. archs can override this later. */
139 void (*__initdata late_time_init)(void);
140
141 /* Untouched command line saved by arch-specific code. */
142 char __initdata boot_command_line[COMMAND_LINE_SIZE];
143 /* Untouched saved command line (eg. for /proc) */
144 char *saved_command_line __ro_after_init;
145 unsigned int saved_command_line_len __ro_after_init;
146 /* Command line for parameter parsing */
147 static char *static_command_line;
148 /* Untouched extra command line */
149 static char *extra_command_line;
150 /* Extra init arguments */
151 static char *extra_init_args;
152
153 #ifdef CONFIG_BOOT_CONFIG
154 /* Is bootconfig on command line? */
155 static bool bootconfig_found;
156 static size_t initargs_offs;
157 #else
158 # define bootconfig_found false
159 # define initargs_offs 0
160 #endif
161
162 static char *execute_command;
163 static char *ramdisk_execute_command = "/init";
164
165 /*
166 * Used to generate warnings if static_key manipulation functions are used
167 * before jump_label_init is called.
168 */
169 bool static_key_initialized __read_mostly;
170 EXPORT_SYMBOL_GPL(static_key_initialized);
171
172 /*
173 * If set, this is an indication to the drivers that reset the underlying
174 * device before going ahead with the initialization otherwise driver might
175 * rely on the BIOS and skip the reset operation.
176 *
177 * This is useful if kernel is booting in an unreliable environment.
178 * For ex. kdump situation where previous kernel has crashed, BIOS has been
179 * skipped and devices will be in unknown state.
180 */
181 unsigned int reset_devices;
182 EXPORT_SYMBOL(reset_devices);
183
set_reset_devices(char * str)184 static int __init set_reset_devices(char *str)
185 {
186 reset_devices = 1;
187 return 1;
188 }
189
190 __setup("reset_devices", set_reset_devices);
191
192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194 static const char *panic_later, *panic_param;
195
obsolete_checksetup(char * line)196 static bool __init obsolete_checksetup(char *line)
197 {
198 const struct obs_kernel_param *p;
199 bool had_early_param = false;
200
201 p = __setup_start;
202 do {
203 int n = strlen(p->str);
204 if (parameqn(line, p->str, n)) {
205 if (p->early) {
206 /* Already done in parse_early_param?
207 * (Needs exact match on param part).
208 * Keep iterating, as we can have early
209 * params and __setups of same names 8( */
210 if (line[n] == '\0' || line[n] == '=')
211 had_early_param = true;
212 } else if (!p->setup_func) {
213 pr_warn("Parameter %s is obsolete, ignored\n",
214 p->str);
215 return true;
216 } else if (p->setup_func(line + n))
217 return true;
218 }
219 p++;
220 } while (p < __setup_end);
221
222 return had_early_param;
223 }
224
225 /*
226 * This should be approx 2 Bo*oMips to start (note initial shift), and will
227 * still work even if initially too large, it will just take slightly longer
228 */
229 unsigned long loops_per_jiffy = (1<<12);
230 EXPORT_SYMBOL(loops_per_jiffy);
231
debug_kernel(char * str)232 static int __init debug_kernel(char *str)
233 {
234 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
235 return 0;
236 }
237
quiet_kernel(char * str)238 static int __init quiet_kernel(char *str)
239 {
240 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
241 return 0;
242 }
243
244 early_param("debug", debug_kernel);
245 early_param("quiet", quiet_kernel);
246
loglevel(char * str)247 static int __init loglevel(char *str)
248 {
249 int newlevel;
250
251 /*
252 * Only update loglevel value when a correct setting was passed,
253 * to prevent blind crashes (when loglevel being set to 0) that
254 * are quite hard to debug
255 */
256 if (get_option(&str, &newlevel)) {
257 console_loglevel = newlevel;
258 return 0;
259 }
260
261 return -EINVAL;
262 }
263
264 early_param("loglevel", loglevel);
265
266 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(size_t * _size)267 static void * __init get_boot_config_from_initrd(size_t *_size)
268 {
269 u32 size, csum;
270 char *data;
271 u32 *hdr;
272 int i;
273
274 if (!initrd_end)
275 return NULL;
276
277 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
278 /*
279 * Since Grub may align the size of initrd to 4, we must
280 * check the preceding 3 bytes as well.
281 */
282 for (i = 0; i < 4; i++) {
283 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
284 goto found;
285 data--;
286 }
287 return NULL;
288
289 found:
290 hdr = (u32 *)(data - 8);
291 size = le32_to_cpu(hdr[0]);
292 csum = le32_to_cpu(hdr[1]);
293
294 data = ((void *)hdr) - size;
295 if ((unsigned long)data < initrd_start) {
296 pr_err("bootconfig size %d is greater than initrd size %ld\n",
297 size, initrd_end - initrd_start);
298 return NULL;
299 }
300
301 if (xbc_calc_checksum(data, size) != csum) {
302 pr_err("bootconfig checksum failed\n");
303 return NULL;
304 }
305
306 /* Remove bootconfig from initramfs/initrd */
307 initrd_end = (unsigned long)data;
308 if (_size)
309 *_size = size;
310
311 return data;
312 }
313 #else
get_boot_config_from_initrd(size_t * _size)314 static void * __init get_boot_config_from_initrd(size_t *_size)
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, *q;
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 /*
350 * For prettier and more readable /proc/cmdline, only
351 * quote the value when necessary, i.e. when it contains
352 * whitespace.
353 */
354 q = strpbrk(val, " \t\r\n") ? "\"" : "";
355 ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
356 xbc_namebuf, q, val, q);
357 if (ret < 0)
358 return ret;
359 buf += ret;
360 }
361 }
362
363 return buf - (end - size);
364 }
365 #undef rest
366
367 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)368 static char * __init xbc_make_cmdline(const char *key)
369 {
370 struct xbc_node *root;
371 char *new_cmdline;
372 int ret, len = 0;
373
374 root = xbc_find_node(key);
375 if (!root)
376 return NULL;
377
378 /* Count required buffer size */
379 len = xbc_snprint_cmdline(NULL, 0, root);
380 if (len <= 0)
381 return NULL;
382
383 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
384 if (!new_cmdline) {
385 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
386 return NULL;
387 }
388
389 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
390 if (ret < 0 || ret > len) {
391 pr_err("Failed to print extra kernel cmdline.\n");
392 memblock_free(new_cmdline, len + 1);
393 return NULL;
394 }
395
396 return new_cmdline;
397 }
398
bootconfig_params(char * param,char * val,const char * unused,void * arg)399 static int __init bootconfig_params(char *param, char *val,
400 const char *unused, void *arg)
401 {
402 if (strcmp(param, "bootconfig") == 0) {
403 bootconfig_found = true;
404 }
405 return 0;
406 }
407
warn_bootconfig(char * str)408 static int __init warn_bootconfig(char *str)
409 {
410 /* The 'bootconfig' has been handled by bootconfig_params(). */
411 return 0;
412 }
413
setup_boot_config(void)414 static void __init setup_boot_config(void)
415 {
416 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
417 const char *msg, *data;
418 int pos, ret;
419 size_t size;
420 char *err;
421
422 /* Cut out the bootconfig data even if we have no bootconfig option */
423 data = get_boot_config_from_initrd(&size);
424 /* If there is no bootconfig in initrd, try embedded one. */
425 if (!data)
426 data = xbc_get_embedded_bootconfig(&size);
427
428 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
429 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
430 bootconfig_params);
431
432 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
433 return;
434
435 /* parse_args() stops at the next param of '--' and returns an address */
436 if (err)
437 initargs_offs = err - tmp_cmdline;
438
439 if (!data) {
440 /* If user intended to use bootconfig, show an error level message */
441 if (bootconfig_found)
442 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
443 else
444 pr_info("No bootconfig data provided, so skipping bootconfig");
445 return;
446 }
447
448 if (size >= XBC_DATA_MAX) {
449 pr_err("bootconfig size %ld greater than max size %d\n",
450 (long)size, XBC_DATA_MAX);
451 return;
452 }
453
454 ret = xbc_init(data, size, &msg, &pos);
455 if (ret < 0) {
456 if (pos < 0)
457 pr_err("Failed to init bootconfig: %s.\n", msg);
458 else
459 pr_err("Failed to parse bootconfig: %s at %d.\n",
460 msg, pos);
461 } else {
462 xbc_get_info(&ret, NULL);
463 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
464 /* keys starting with "kernel." are passed via cmdline */
465 extra_command_line = xbc_make_cmdline("kernel");
466 /* Also, "init." keys are init arguments */
467 extra_init_args = xbc_make_cmdline("init");
468 }
469 return;
470 }
471
exit_boot_config(void)472 static void __init exit_boot_config(void)
473 {
474 xbc_exit();
475 }
476
477 #else /* !CONFIG_BOOT_CONFIG */
478
setup_boot_config(void)479 static void __init setup_boot_config(void)
480 {
481 /* Remove bootconfig data from initrd */
482 get_boot_config_from_initrd(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
491 #define exit_boot_config() do {} while (0)
492
493 #endif /* CONFIG_BOOT_CONFIG */
494
495 early_param("bootconfig", warn_bootconfig);
496
cmdline_has_extra_options(void)497 bool __init cmdline_has_extra_options(void)
498 {
499 return extra_command_line || extra_init_args;
500 }
501
502 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)503 static void __init repair_env_string(char *param, char *val)
504 {
505 if (val) {
506 /* param=val or param="val"? */
507 if (val == param+strlen(param)+1)
508 val[-1] = '=';
509 else if (val == param+strlen(param)+2) {
510 val[-2] = '=';
511 memmove(val-1, val, strlen(val)+1);
512 } else
513 BUG();
514 }
515 }
516
517 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)518 static int __init set_init_arg(char *param, char *val,
519 const char *unused, void *arg)
520 {
521 unsigned int i;
522
523 if (panic_later)
524 return 0;
525
526 repair_env_string(param, val);
527
528 for (i = 0; argv_init[i]; i++) {
529 if (i == MAX_INIT_ARGS) {
530 panic_later = "init";
531 panic_param = param;
532 return 0;
533 }
534 }
535 argv_init[i] = param;
536 return 0;
537 }
538
539 /*
540 * Unknown boot options get handed to init, unless they look like
541 * unused parameters (modprobe will find them in /proc/cmdline).
542 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)543 static int __init unknown_bootoption(char *param, char *val,
544 const char *unused, void *arg)
545 {
546 size_t len = strlen(param);
547
548 /* Handle params aliased to sysctls */
549 if (sysctl_is_alias(param))
550 return 0;
551
552 repair_env_string(param, val);
553
554 /* Handle obsolete-style parameters */
555 if (obsolete_checksetup(param))
556 return 0;
557
558 /* Unused module parameter. */
559 if (strnchr(param, len, '.'))
560 return 0;
561
562 if (panic_later)
563 return 0;
564
565 if (val) {
566 /* Environment option */
567 unsigned int i;
568 for (i = 0; envp_init[i]; i++) {
569 if (i == MAX_INIT_ENVS) {
570 panic_later = "env";
571 panic_param = param;
572 }
573 if (!strncmp(param, envp_init[i], len+1))
574 break;
575 }
576 envp_init[i] = param;
577 } else {
578 /* Command line option */
579 unsigned int i;
580 for (i = 0; argv_init[i]; i++) {
581 if (i == MAX_INIT_ARGS) {
582 panic_later = "init";
583 panic_param = param;
584 }
585 }
586 argv_init[i] = param;
587 }
588 return 0;
589 }
590
init_setup(char * str)591 static int __init init_setup(char *str)
592 {
593 unsigned int i;
594
595 execute_command = str;
596 /*
597 * In case LILO is going to boot us with default command line,
598 * it prepends "auto" before the whole cmdline which makes
599 * the shell think it should execute a script with such name.
600 * So we ignore all arguments entered _before_ init=... [MJ]
601 */
602 for (i = 1; i < MAX_INIT_ARGS; i++)
603 argv_init[i] = NULL;
604 return 1;
605 }
606 __setup("init=", init_setup);
607
rdinit_setup(char * str)608 static int __init rdinit_setup(char *str)
609 {
610 unsigned int i;
611
612 ramdisk_execute_command = str;
613 /* See "auto" comment in init_setup */
614 for (i = 1; i < MAX_INIT_ARGS; i++)
615 argv_init[i] = NULL;
616 return 1;
617 }
618 __setup("rdinit=", rdinit_setup);
619
620 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)621 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)622 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
623 #endif
624
625 /*
626 * We need to store the untouched command line for future reference.
627 * We also need to store the touched command line since the parameter
628 * parsing is performed in place, and we should allow a component to
629 * store reference of name/value for future reference.
630 */
setup_command_line(char * command_line)631 static void __init setup_command_line(char *command_line)
632 {
633 size_t len, xlen = 0, ilen = 0;
634
635 if (extra_command_line)
636 xlen = strlen(extra_command_line);
637 if (extra_init_args) {
638 extra_init_args = strim(extra_init_args); /* remove trailing space */
639 ilen = strlen(extra_init_args) + 4; /* for " -- " */
640 }
641
642 len = xlen + strlen(boot_command_line) + ilen + 1;
643
644 saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
645
646 len = xlen + strlen(command_line) + 1;
647
648 static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
649
650 if (xlen) {
651 /*
652 * We have to put extra_command_line before boot command
653 * lines because there could be dashes (separator of init
654 * command line) in the command lines.
655 */
656 strcpy(saved_command_line, extra_command_line);
657 strcpy(static_command_line, extra_command_line);
658 }
659 strcpy(saved_command_line + xlen, boot_command_line);
660 strcpy(static_command_line + xlen, command_line);
661
662 if (ilen) {
663 /*
664 * Append supplemental init boot args to saved_command_line
665 * so that user can check what command line options passed
666 * to init.
667 * The order should always be
668 * " -- "[bootconfig init-param][cmdline init-param]
669 */
670 if (initargs_offs) {
671 len = xlen + initargs_offs;
672 strcpy(saved_command_line + len, extra_init_args);
673 len += ilen - 4; /* strlen(extra_init_args) */
674 strcpy(saved_command_line + len,
675 boot_command_line + initargs_offs - 1);
676 } else {
677 len = strlen(saved_command_line);
678 strcpy(saved_command_line + len, " -- ");
679 len += 4;
680 strcpy(saved_command_line + len, extra_init_args);
681 }
682 }
683
684 saved_command_line_len = strlen(saved_command_line);
685 }
686
687 /*
688 * We need to finalize in a non-__init function or else race conditions
689 * between the root thread and the init thread may cause start_kernel to
690 * be reaped by free_initmem before the root thread has proceeded to
691 * cpu_idle.
692 *
693 * gcc-3.4 accidentally inlines this function, so use noinline.
694 */
695
696 static __initdata DECLARE_COMPLETION(kthreadd_done);
697
rest_init(void)698 static noinline void __ref __noreturn rest_init(void)
699 {
700 struct task_struct *tsk;
701 int pid;
702
703 rcu_scheduler_starting();
704 /*
705 * We need to spawn init first so that it obtains pid 1, however
706 * the init task will end up wanting to create kthreads, which, if
707 * we schedule it before we create kthreadd, will OOPS.
708 */
709 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
710 /*
711 * Pin init on the boot CPU. Task migration is not properly working
712 * until sched_init_smp() has been run. It will set the allowed
713 * CPUs for init to the non isolated CPUs.
714 */
715 rcu_read_lock();
716 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
717 tsk->flags |= PF_NO_SETAFFINITY;
718 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
719 rcu_read_unlock();
720
721 numa_default_policy();
722 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
723 rcu_read_lock();
724 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
725 rcu_read_unlock();
726
727 /*
728 * Enable might_sleep() and smp_processor_id() checks.
729 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
730 * kernel_thread() would trigger might_sleep() splats. With
731 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
732 * already, but it's stuck on the kthreadd_done completion.
733 */
734 system_state = SYSTEM_SCHEDULING;
735
736 complete(&kthreadd_done);
737
738 /*
739 * The boot idle thread must execute schedule()
740 * at least once to get things moving:
741 */
742 schedule_preempt_disabled();
743 /* Call into cpu_idle with preempt disabled */
744 cpu_startup_entry(CPUHP_ONLINE);
745 }
746
747 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)748 static int __init do_early_param(char *param, char *val,
749 const char *unused, void *arg)
750 {
751 const struct obs_kernel_param *p;
752
753 for (p = __setup_start; p < __setup_end; p++) {
754 if (p->early && parameq(param, p->str)) {
755 if (p->setup_func(val) != 0)
756 pr_warn("Malformed early option '%s'\n", param);
757 }
758 }
759 /* We accept everything at this stage. */
760 return 0;
761 }
762
parse_early_options(char * cmdline)763 void __init parse_early_options(char *cmdline)
764 {
765 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
766 do_early_param);
767 }
768
769 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)770 void __init parse_early_param(void)
771 {
772 static int done __initdata;
773 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
774
775 if (done)
776 return;
777
778 /* All fall through to do_early_param. */
779 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
780 parse_early_options(tmp_cmdline);
781 done = 1;
782 }
783
arch_post_acpi_subsys_init(void)784 void __init __weak arch_post_acpi_subsys_init(void) { }
785
smp_setup_processor_id(void)786 void __init __weak smp_setup_processor_id(void)
787 {
788 }
789
smp_prepare_boot_cpu(void)790 void __init __weak smp_prepare_boot_cpu(void)
791 {
792 }
793
794 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)795 void __init __weak thread_stack_cache_init(void)
796 {
797 }
798 #endif
799
poking_init(void)800 void __init __weak poking_init(void) { }
801
pgtable_cache_init(void)802 void __init __weak pgtable_cache_init(void) { }
803
trap_init(void)804 void __init __weak trap_init(void) { }
805
806 bool initcall_debug;
807 core_param(initcall_debug, initcall_debug, bool, 0644);
808
809 #ifdef TRACEPOINTS_ENABLED
810 static void __init initcall_debug_enable(void);
811 #else
initcall_debug_enable(void)812 static inline void initcall_debug_enable(void)
813 {
814 }
815 #endif
816
817 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
818 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
819 randomize_kstack_offset);
820 DEFINE_PER_CPU(u32, kstack_offset);
821
early_randomize_kstack_offset(char * buf)822 static int __init early_randomize_kstack_offset(char *buf)
823 {
824 int ret;
825 bool bool_result;
826
827 ret = kstrtobool(buf, &bool_result);
828 if (ret)
829 return ret;
830
831 if (bool_result)
832 static_branch_enable(&randomize_kstack_offset);
833 else
834 static_branch_disable(&randomize_kstack_offset);
835 return 0;
836 }
837 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
838 #endif
839
print_unknown_bootoptions(void)840 static void __init print_unknown_bootoptions(void)
841 {
842 char *unknown_options;
843 char *end;
844 const char *const *p;
845 size_t len;
846
847 if (panic_later || (!argv_init[1] && !envp_init[2]))
848 return;
849
850 /*
851 * Determine how many options we have to print out, plus a space
852 * before each
853 */
854 len = 1; /* null terminator */
855 for (p = &argv_init[1]; *p; p++) {
856 len++;
857 len += strlen(*p);
858 }
859 for (p = &envp_init[2]; *p; p++) {
860 len++;
861 len += strlen(*p);
862 }
863
864 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
865 if (!unknown_options) {
866 pr_err("%s: Failed to allocate %zu bytes\n",
867 __func__, len);
868 return;
869 }
870 end = unknown_options;
871
872 for (p = &argv_init[1]; *p; p++)
873 end += sprintf(end, " %s", *p);
874 for (p = &envp_init[2]; *p; p++)
875 end += sprintf(end, " %s", *p);
876
877 /* Start at unknown_options[1] to skip the initial space */
878 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
879 &unknown_options[1]);
880 memblock_free(unknown_options, len);
881 }
882
early_numa_node_init(void)883 static void __init early_numa_node_init(void)
884 {
885 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
886 #ifndef cpu_to_node
887 int cpu;
888
889 /* The early_cpu_to_node() should be ready here. */
890 for_each_possible_cpu(cpu)
891 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
892 #endif
893 #endif
894 }
895
896 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)897 void start_kernel(void)
898 {
899 char *command_line;
900 char *after_dashes;
901
902 set_task_stack_end_magic(&init_task);
903 smp_setup_processor_id();
904 debug_objects_early_init();
905 init_vmlinux_build_id();
906
907 cgroup_init_early();
908
909 local_irq_disable();
910 early_boot_irqs_disabled = true;
911
912 /*
913 * Interrupts are still disabled. Do necessary setups, then
914 * enable them.
915 */
916 boot_cpu_init();
917 page_address_init();
918 pr_notice("%s", linux_banner);
919 setup_arch(&command_line);
920 /* Static keys and static calls are needed by LSMs */
921 jump_label_init();
922 static_call_init();
923 early_security_init();
924 setup_boot_config();
925 setup_command_line(command_line);
926 setup_nr_cpu_ids();
927 setup_per_cpu_areas();
928 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
929 early_numa_node_init();
930 boot_cpu_hotplug_init();
931
932 pr_notice("Kernel command line: %s\n", saved_command_line);
933 /* parameters may set static keys */
934 parse_early_param();
935 after_dashes = parse_args("Booting kernel",
936 static_command_line, __start___param,
937 __stop___param - __start___param,
938 -1, -1, NULL, &unknown_bootoption);
939 print_unknown_bootoptions();
940 if (!IS_ERR_OR_NULL(after_dashes))
941 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
942 NULL, set_init_arg);
943 if (extra_init_args)
944 parse_args("Setting extra init args", extra_init_args,
945 NULL, 0, -1, -1, NULL, set_init_arg);
946
947 /* Architectural and non-timekeeping rng init, before allocator init */
948 random_init_early(command_line);
949
950 /*
951 * These use large bootmem allocations and must precede
952 * initalization of page allocator
953 */
954 setup_log_buf(0);
955 vfs_caches_init_early();
956 sort_main_extable();
957 trap_init();
958 mm_core_init();
959 poking_init();
960 ftrace_init();
961
962 /* trace_printk can be enabled here */
963 early_trace_init();
964
965 /*
966 * Set up the scheduler prior starting any interrupts (such as the
967 * timer interrupt). Full topology setup happens at smp_init()
968 * time - but meanwhile we still have a functioning scheduler.
969 */
970 sched_init();
971
972 if (WARN(!irqs_disabled(),
973 "Interrupts were enabled *very* early, fixing it\n"))
974 local_irq_disable();
975 radix_tree_init();
976 maple_tree_init();
977
978 /*
979 * Set up housekeeping before setting up workqueues to allow the unbound
980 * workqueue to take non-housekeeping into account.
981 */
982 housekeeping_init();
983
984 /*
985 * Allow workqueue creation and work item queueing/cancelling
986 * early. Work item execution depends on kthreads and starts after
987 * workqueue_init().
988 */
989 workqueue_init_early();
990
991 rcu_init();
992 kvfree_rcu_init();
993
994 /* Trace events are available after this */
995 trace_init();
996
997 if (initcall_debug)
998 initcall_debug_enable();
999
1000 context_tracking_init();
1001 /* init some links before init_ISA_irqs() */
1002 early_irq_init();
1003 init_IRQ();
1004 tick_init();
1005 rcu_init_nohz();
1006 timers_init();
1007 srcu_init();
1008 hrtimers_init();
1009 softirq_init();
1010 timekeeping_init();
1011 time_init();
1012
1013 /* This must be after timekeeping is initialized */
1014 random_init();
1015
1016 /* These make use of the fully initialized rng */
1017 kfence_init();
1018 boot_init_stack_canary();
1019
1020 perf_event_init();
1021 profile_init();
1022 call_function_init();
1023 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1024
1025 early_boot_irqs_disabled = false;
1026 local_irq_enable();
1027
1028 kmem_cache_init_late();
1029
1030 /*
1031 * HACK ALERT! This is early. We're enabling the console before
1032 * we've done PCI setups etc, and console_init() must be aware of
1033 * this. But we do want output early, in case something goes wrong.
1034 */
1035 console_init();
1036 if (panic_later)
1037 panic("Too many boot %s vars at `%s'", panic_later,
1038 panic_param);
1039
1040 lockdep_init();
1041
1042 /*
1043 * Need to run this when irqs are enabled, because it wants
1044 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1045 * too:
1046 */
1047 locking_selftest();
1048
1049 #ifdef CONFIG_BLK_DEV_INITRD
1050 if (initrd_start && !initrd_below_start_ok &&
1051 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1052 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1053 page_to_pfn(virt_to_page((void *)initrd_start)),
1054 min_low_pfn);
1055 initrd_start = 0;
1056 }
1057 #endif
1058 setup_per_cpu_pageset();
1059 numa_policy_init();
1060 acpi_early_init();
1061 if (late_time_init)
1062 late_time_init();
1063 sched_clock_init();
1064 calibrate_delay();
1065
1066 arch_cpu_finalize_init();
1067
1068 pid_idr_init();
1069 anon_vma_init();
1070 thread_stack_cache_init();
1071 cred_init();
1072 fork_init();
1073 proc_caches_init();
1074 uts_ns_init();
1075 key_init();
1076 security_init();
1077 dbg_late_init();
1078 net_ns_init();
1079 vfs_caches_init();
1080 pagecache_init();
1081 signals_init();
1082 seq_file_init();
1083 proc_root_init();
1084 nsfs_init();
1085 pidfs_init();
1086 cpuset_init();
1087 mem_cgroup_init();
1088 cgroup_init();
1089 taskstats_init_early();
1090 delayacct_init();
1091
1092 acpi_subsystem_init();
1093 arch_post_acpi_subsys_init();
1094 kcsan_init();
1095
1096 /* Do the rest non-__init'ed, we're now alive */
1097 rest_init();
1098
1099 /*
1100 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1101 * and older.
1102 */
1103 #if !__has_attribute(__no_stack_protector__)
1104 prevent_tail_call_optimization();
1105 #endif
1106 }
1107
1108 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1109 static void __init do_ctors(void)
1110 {
1111 /*
1112 * For UML, the constructors have already been called by the
1113 * normal setup code as it's just a normal ELF binary, so we
1114 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1115 * even on UML for modules.
1116 */
1117 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1118 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1119
1120 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1121 (*fn)();
1122 #endif
1123 }
1124
1125 #ifdef CONFIG_KALLSYMS
1126 struct blacklist_entry {
1127 struct list_head next;
1128 char *buf;
1129 };
1130
1131 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1132
initcall_blacklist(char * str)1133 static int __init initcall_blacklist(char *str)
1134 {
1135 char *str_entry;
1136 struct blacklist_entry *entry;
1137
1138 /* str argument is a comma-separated list of functions */
1139 do {
1140 str_entry = strsep(&str, ",");
1141 if (str_entry) {
1142 pr_debug("blacklisting initcall %s\n", str_entry);
1143 entry = memblock_alloc_or_panic(sizeof(*entry),
1144 SMP_CACHE_BYTES);
1145 entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1,
1146 SMP_CACHE_BYTES);
1147 strcpy(entry->buf, str_entry);
1148 list_add(&entry->next, &blacklisted_initcalls);
1149 }
1150 } while (str_entry);
1151
1152 return 1;
1153 }
1154
initcall_blacklisted(initcall_t fn)1155 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1156 {
1157 struct blacklist_entry *entry;
1158 char fn_name[KSYM_SYMBOL_LEN];
1159 unsigned long addr;
1160
1161 if (list_empty(&blacklisted_initcalls))
1162 return false;
1163
1164 addr = (unsigned long) dereference_function_descriptor(fn);
1165 sprint_symbol_no_offset(fn_name, addr);
1166
1167 /*
1168 * fn will be "function_name [module_name]" where [module_name] is not
1169 * displayed for built-in init functions. Strip off the [module_name].
1170 */
1171 strreplace(fn_name, ' ', '\0');
1172
1173 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1174 if (!strcmp(fn_name, entry->buf)) {
1175 pr_debug("initcall %s blacklisted\n", fn_name);
1176 return true;
1177 }
1178 }
1179
1180 return false;
1181 }
1182 #else
initcall_blacklist(char * str)1183 static int __init initcall_blacklist(char *str)
1184 {
1185 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1186 return 0;
1187 }
1188
initcall_blacklisted(initcall_t fn)1189 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1190 {
1191 return false;
1192 }
1193 #endif
1194 __setup("initcall_blacklist=", initcall_blacklist);
1195
1196 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1197 trace_initcall_start_cb(void *data, initcall_t fn)
1198 {
1199 ktime_t *calltime = data;
1200
1201 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1202 *calltime = ktime_get();
1203 }
1204
1205 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1206 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1207 {
1208 ktime_t rettime, *calltime = data;
1209
1210 rettime = ktime_get();
1211 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1212 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1213 }
1214
1215 static __init_or_module void
trace_initcall_level_cb(void * data,const char * level)1216 trace_initcall_level_cb(void *data, const char *level)
1217 {
1218 printk(KERN_DEBUG "entering initcall level: %s\n", level);
1219 }
1220
1221 static ktime_t initcall_calltime;
1222
1223 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1224 static void __init initcall_debug_enable(void)
1225 {
1226 int ret;
1227
1228 ret = register_trace_initcall_start(trace_initcall_start_cb,
1229 &initcall_calltime);
1230 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1231 &initcall_calltime);
1232 ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL);
1233 WARN(ret, "Failed to register initcall tracepoints\n");
1234 }
1235 # define do_trace_initcall_start trace_initcall_start
1236 # define do_trace_initcall_finish trace_initcall_finish
1237 # define do_trace_initcall_level trace_initcall_level
1238 #else
do_trace_initcall_start(initcall_t fn)1239 static inline void do_trace_initcall_start(initcall_t fn)
1240 {
1241 if (!initcall_debug)
1242 return;
1243 trace_initcall_start_cb(&initcall_calltime, fn);
1244 }
do_trace_initcall_finish(initcall_t fn,int ret)1245 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1246 {
1247 if (!initcall_debug)
1248 return;
1249 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1250 }
do_trace_initcall_level(const char * level)1251 static inline void do_trace_initcall_level(const char *level)
1252 {
1253 if (!initcall_debug)
1254 return;
1255 trace_initcall_level_cb(NULL, level);
1256 }
1257 #endif /* !TRACEPOINTS_ENABLED */
1258
do_one_initcall(initcall_t fn)1259 int __init_or_module do_one_initcall(initcall_t fn)
1260 {
1261 int count = preempt_count();
1262 char msgbuf[64];
1263 int ret;
1264
1265 if (initcall_blacklisted(fn))
1266 return -EPERM;
1267
1268 do_trace_initcall_start(fn);
1269 ret = fn();
1270 do_trace_initcall_finish(fn, ret);
1271
1272 msgbuf[0] = 0;
1273
1274 if (preempt_count() != count) {
1275 sprintf(msgbuf, "preemption imbalance ");
1276 preempt_count_set(count);
1277 }
1278 if (irqs_disabled()) {
1279 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1280 local_irq_enable();
1281 }
1282 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1283
1284 add_latent_entropy();
1285 return ret;
1286 }
1287
1288
1289 static initcall_entry_t *initcall_levels[] __initdata = {
1290 __initcall0_start,
1291 __initcall1_start,
1292 __initcall2_start,
1293 __initcall3_start,
1294 __initcall4_start,
1295 __initcall5_start,
1296 __initcall6_start,
1297 __initcall7_start,
1298 __initcall_end,
1299 };
1300
1301 /* Keep these in sync with initcalls in include/linux/init.h */
1302 static const char *initcall_level_names[] __initdata = {
1303 "pure",
1304 "core",
1305 "postcore",
1306 "arch",
1307 "subsys",
1308 "fs",
1309 "device",
1310 "late",
1311 };
1312
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1313 static int __init ignore_unknown_bootoption(char *param, char *val,
1314 const char *unused, void *arg)
1315 {
1316 return 0;
1317 }
1318
do_initcall_level(int level,char * command_line)1319 static void __init do_initcall_level(int level, char *command_line)
1320 {
1321 initcall_entry_t *fn;
1322
1323 parse_args(initcall_level_names[level],
1324 command_line, __start___param,
1325 __stop___param - __start___param,
1326 level, level,
1327 NULL, ignore_unknown_bootoption);
1328
1329 do_trace_initcall_level(initcall_level_names[level]);
1330 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1331 do_one_initcall(initcall_from_entry(fn));
1332 }
1333
do_initcalls(void)1334 static void __init do_initcalls(void)
1335 {
1336 int level;
1337 size_t len = saved_command_line_len + 1;
1338 char *command_line;
1339
1340 command_line = kzalloc(len, GFP_KERNEL);
1341 if (!command_line)
1342 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1343
1344 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1345 /* Parser modifies command_line, restore it each time */
1346 strcpy(command_line, saved_command_line);
1347 do_initcall_level(level, command_line);
1348 }
1349
1350 kfree(command_line);
1351 }
1352
1353 /*
1354 * Ok, the machine is now initialized. None of the devices
1355 * have been touched yet, but the CPU subsystem is up and
1356 * running, and memory and process management works.
1357 *
1358 * Now we can finally start doing some real work..
1359 */
do_basic_setup(void)1360 static void __init do_basic_setup(void)
1361 {
1362 cpuset_init_smp();
1363 driver_init();
1364 init_irq_proc();
1365 do_ctors();
1366 do_initcalls();
1367 }
1368
do_pre_smp_initcalls(void)1369 static void __init do_pre_smp_initcalls(void)
1370 {
1371 initcall_entry_t *fn;
1372
1373 do_trace_initcall_level("early");
1374 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1375 do_one_initcall(initcall_from_entry(fn));
1376 }
1377
run_init_process(const char * init_filename)1378 static int run_init_process(const char *init_filename)
1379 {
1380 const char *const *p;
1381
1382 argv_init[0] = init_filename;
1383 pr_info("Run %s as init process\n", init_filename);
1384 pr_debug(" with arguments:\n");
1385 for (p = argv_init; *p; p++)
1386 pr_debug(" %s\n", *p);
1387 pr_debug(" with environment:\n");
1388 for (p = envp_init; *p; p++)
1389 pr_debug(" %s\n", *p);
1390 return kernel_execve(init_filename, argv_init, envp_init);
1391 }
1392
try_to_run_init_process(const char * init_filename)1393 static int try_to_run_init_process(const char *init_filename)
1394 {
1395 int ret;
1396
1397 ret = run_init_process(init_filename);
1398
1399 if (ret && ret != -ENOENT) {
1400 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1401 init_filename, ret);
1402 }
1403
1404 return ret;
1405 }
1406
1407 static noinline void __init kernel_init_freeable(void);
1408
1409 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1410 bool rodata_enabled __ro_after_init = true;
1411
1412 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1413 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1414 #endif
1415
set_debug_rodata(char * str)1416 static int __init set_debug_rodata(char *str)
1417 {
1418 if (arch_parse_debug_rodata(str))
1419 return 0;
1420
1421 if (str && !strcmp(str, "on"))
1422 rodata_enabled = true;
1423 else if (str && !strcmp(str, "off"))
1424 rodata_enabled = false;
1425 else
1426 pr_warn("Invalid option string for rodata: '%s'\n", str);
1427 return 0;
1428 }
1429 early_param("rodata", set_debug_rodata);
1430 #endif
1431
mark_readonly(void)1432 static void mark_readonly(void)
1433 {
1434 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1435 /*
1436 * load_module() results in W+X mappings, which are cleaned
1437 * up with init_free_wq. Let's make sure that queued work is
1438 * flushed so that we don't hit false positives looking for
1439 * insecure pages which are W+X.
1440 */
1441 flush_module_init_free_work();
1442 jump_label_init_ro();
1443 mark_rodata_ro();
1444 debug_checkwx();
1445 rodata_test();
1446 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1447 pr_info("Kernel memory protection disabled.\n");
1448 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1449 pr_warn("Kernel memory protection not selected by kernel config.\n");
1450 } else {
1451 pr_warn("This architecture does not have kernel memory protection.\n");
1452 }
1453 }
1454
free_initmem(void)1455 void __weak free_initmem(void)
1456 {
1457 free_initmem_default(POISON_FREE_INITMEM);
1458 }
1459
kernel_init(void * unused)1460 static int __ref kernel_init(void *unused)
1461 {
1462 int ret;
1463
1464 /*
1465 * Wait until kthreadd is all set-up.
1466 */
1467 wait_for_completion(&kthreadd_done);
1468
1469 kernel_init_freeable();
1470 /* need to finish all async __init code before freeing the memory */
1471 async_synchronize_full();
1472
1473 system_state = SYSTEM_FREEING_INITMEM;
1474 kprobe_free_init_mem();
1475 ftrace_free_init_mem();
1476 kgdb_free_init_mem();
1477 exit_boot_config();
1478 free_initmem();
1479 mark_readonly();
1480
1481 /*
1482 * Kernel mappings are now finalized - update the userspace page-table
1483 * to finalize PTI.
1484 */
1485 pti_finalize();
1486
1487 system_state = SYSTEM_RUNNING;
1488 numa_default_policy();
1489
1490 rcu_end_inkernel_boot();
1491
1492 do_sysctl_args();
1493
1494 if (ramdisk_execute_command) {
1495 ret = run_init_process(ramdisk_execute_command);
1496 if (!ret)
1497 return 0;
1498 pr_err("Failed to execute %s (error %d)\n",
1499 ramdisk_execute_command, ret);
1500 }
1501
1502 /*
1503 * We try each of these until one succeeds.
1504 *
1505 * The Bourne shell can be used instead of init if we are
1506 * trying to recover a really broken machine.
1507 */
1508 if (execute_command) {
1509 ret = run_init_process(execute_command);
1510 if (!ret)
1511 return 0;
1512 panic("Requested init %s failed (error %d).",
1513 execute_command, ret);
1514 }
1515
1516 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1517 ret = run_init_process(CONFIG_DEFAULT_INIT);
1518 if (ret)
1519 pr_err("Default init %s failed (error %d)\n",
1520 CONFIG_DEFAULT_INIT, ret);
1521 else
1522 return 0;
1523 }
1524
1525 if (!try_to_run_init_process("/sbin/init") ||
1526 !try_to_run_init_process("/etc/init") ||
1527 !try_to_run_init_process("/bin/init") ||
1528 !try_to_run_init_process("/bin/sh"))
1529 return 0;
1530
1531 panic("No working init found. Try passing init= option to kernel. "
1532 "See Linux Documentation/admin-guide/init.rst for guidance.");
1533 }
1534
1535 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1536 void __init console_on_rootfs(void)
1537 {
1538 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1539
1540 if (IS_ERR(file)) {
1541 pr_err("Warning: unable to open an initial console.\n");
1542 return;
1543 }
1544 init_dup(file);
1545 init_dup(file);
1546 init_dup(file);
1547 fput(file);
1548 }
1549
kernel_init_freeable(void)1550 static noinline void __init kernel_init_freeable(void)
1551 {
1552 /* Now the scheduler is fully set up and can do blocking allocations */
1553 gfp_allowed_mask = __GFP_BITS_MASK;
1554
1555 /*
1556 * init can allocate pages on any node
1557 */
1558 set_mems_allowed(node_states[N_MEMORY]);
1559
1560 cad_pid = get_pid(task_pid(current));
1561
1562 smp_prepare_cpus(setup_max_cpus);
1563
1564 workqueue_init();
1565
1566 init_mm_internals();
1567
1568 do_pre_smp_initcalls();
1569 lockup_detector_init();
1570
1571 smp_init();
1572 sched_init_smp();
1573
1574 workqueue_init_topology();
1575 async_init();
1576 padata_init();
1577 page_alloc_init_late();
1578
1579 do_basic_setup();
1580
1581 kunit_run_all_tests();
1582
1583 wait_for_initramfs();
1584 console_on_rootfs();
1585
1586 /*
1587 * check if there is an early userspace init. If yes, let it do all
1588 * the work
1589 */
1590 if (init_eaccess(ramdisk_execute_command) != 0) {
1591 ramdisk_execute_command = NULL;
1592 prepare_namespace();
1593 }
1594
1595 /*
1596 * Ok, we have completed the initial bootup, and
1597 * we're essentially up and running. Get rid of the
1598 * initmem segments and start the user-mode stuff..
1599 *
1600 * rootfs is available now, try loading the public keys
1601 * and default modules
1602 */
1603
1604 integrity_load_keys();
1605 }
1606