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/efi.h>
57 #include <linux/tick.h>
58 #include <linux/sched/isolation.h>
59 #include <linux/interrupt.h>
60 #include <linux/taskstats_kern.h>
61 #include <linux/delayacct.h>
62 #include <linux/unistd.h>
63 #include <linux/utsname.h>
64 #include <linux/rmap.h>
65 #include <linux/mempolicy.h>
66 #include <linux/key.h>
67 #include <linux/debug_locks.h>
68 #include <linux/debugobjects.h>
69 #include <linux/lockdep.h>
70 #include <linux/kmemleak.h>
71 #include <linux/padata.h>
72 #include <linux/pid_namespace.h>
73 #include <linux/device/driver.h>
74 #include <linux/kthread.h>
75 #include <linux/sched.h>
76 #include <linux/sched/init.h>
77 #include <linux/signal.h>
78 #include <linux/idr.h>
79 #include <linux/kgdb.h>
80 #include <linux/ftrace.h>
81 #include <linux/async.h>
82 #include <linux/shmem_fs.h>
83 #include <linux/slab.h>
84 #include <linux/perf_event.h>
85 #include <linux/ptrace.h>
86 #include <linux/pti.h>
87 #include <linux/blkdev.h>
88 #include <linux/sched/clock.h>
89 #include <linux/sched/task.h>
90 #include <linux/sched/task_stack.h>
91 #include <linux/context_tracking.h>
92 #include <linux/random.h>
93 #include <linux/moduleloader.h>
94 #include <linux/list.h>
95 #include <linux/integrity.h>
96 #include <linux/proc_ns.h>
97 #include <linux/io.h>
98 #include <linux/cache.h>
99 #include <linux/rodata_test.h>
100 #include <linux/jump_label.h>
101 #include <linux/kcsan.h>
102 #include <linux/init_syscalls.h>
103 #include <linux/stackdepot.h>
104 #include <linux/randomize_kstack.h>
105 #include <linux/pidfs.h>
106 #include <linux/ptdump.h>
107 #include <net/net_namespace.h>
108
109 #include <asm/io.h>
110 #include <asm/setup.h>
111 #include <asm/sections.h>
112 #include <asm/cacheflush.h>
113
114 #define CREATE_TRACE_POINTS
115 #include <trace/events/initcall.h>
116
117 #include <kunit/test.h>
118
119 static int kernel_init(void *);
120
121 /*
122 * Debug helper: via this flag we know that we are in 'early bootup code'
123 * where only the boot processor is running with IRQ disabled. This means
124 * two things - IRQ must not be enabled before the flag is cleared and some
125 * operations which are not allowed with IRQ disabled are allowed while the
126 * flag is set.
127 */
128 bool early_boot_irqs_disabled __read_mostly;
129
130 enum system_states system_state __read_mostly;
131 EXPORT_SYMBOL(system_state);
132
133 /*
134 * Boot command-line arguments
135 */
136 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
137 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
138
139 /* Default late time init is NULL. archs can override this later. */
140 void (*__initdata late_time_init)(void);
141
142 /* Untouched command line saved by arch-specific code. */
143 char __initdata boot_command_line[COMMAND_LINE_SIZE];
144 /* Untouched saved command line (eg. for /proc) */
145 char *saved_command_line __ro_after_init;
146 unsigned int saved_command_line_len __ro_after_init;
147 /* Command line for parameter parsing */
148 static char *static_command_line;
149 /* Untouched extra command line */
150 static char *extra_command_line;
151 /* Extra init arguments */
152 static char *extra_init_args;
153
154 #ifdef CONFIG_BOOT_CONFIG
155 /* Is bootconfig on command line? */
156 static bool bootconfig_found;
157 static size_t initargs_offs;
158 #else
159 # define bootconfig_found false
160 # define initargs_offs 0
161 #endif
162
163 static char *execute_command;
164 static char *ramdisk_execute_command = "/init";
165
166 /*
167 * Used to generate warnings if static_key manipulation functions are used
168 * before jump_label_init is called.
169 */
170 bool static_key_initialized __read_mostly;
171 EXPORT_SYMBOL_GPL(static_key_initialized);
172
173 /*
174 * If set, this is an indication to the drivers that reset the underlying
175 * device before going ahead with the initialization otherwise driver might
176 * rely on the BIOS and skip the reset operation.
177 *
178 * This is useful if kernel is booting in an unreliable environment.
179 * For ex. kdump situation where previous kernel has crashed, BIOS has been
180 * skipped and devices will be in unknown state.
181 */
182 unsigned int reset_devices;
183 EXPORT_SYMBOL(reset_devices);
184
set_reset_devices(char * str)185 static int __init set_reset_devices(char *str)
186 {
187 reset_devices = 1;
188 return 1;
189 }
190
191 __setup("reset_devices", set_reset_devices);
192
193 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
194 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
195 static const char *panic_later, *panic_param;
196
obsolete_checksetup(char * line)197 static bool __init obsolete_checksetup(char *line)
198 {
199 const struct obs_kernel_param *p;
200 bool had_early_param = false;
201
202 p = __setup_start;
203 do {
204 int n = strlen(p->str);
205 if (parameqn(line, p->str, n)) {
206 if (p->early) {
207 /* Already done in parse_early_param?
208 * (Needs exact match on param part).
209 * Keep iterating, as we can have early
210 * params and __setups of same names 8( */
211 if (line[n] == '\0' || line[n] == '=')
212 had_early_param = true;
213 } else if (!p->setup_func) {
214 pr_warn("Parameter %s is obsolete, ignored\n",
215 p->str);
216 return true;
217 } else if (p->setup_func(line + n))
218 return true;
219 }
220 p++;
221 } while (p < __setup_end);
222
223 return had_early_param;
224 }
225
226 /*
227 * This should be approx 2 Bo*oMips to start (note initial shift), and will
228 * still work even if initially too large, it will just take slightly longer
229 */
230 unsigned long loops_per_jiffy = (1<<12);
231 EXPORT_SYMBOL(loops_per_jiffy);
232
debug_kernel(char * str)233 static int __init debug_kernel(char *str)
234 {
235 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
236 return 0;
237 }
238
quiet_kernel(char * str)239 static int __init quiet_kernel(char *str)
240 {
241 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
242 return 0;
243 }
244
245 early_param("debug", debug_kernel);
246 early_param("quiet", quiet_kernel);
247
loglevel(char * str)248 static int __init loglevel(char *str)
249 {
250 int newlevel;
251
252 /*
253 * Only update loglevel value when a correct setting was passed,
254 * to prevent blind crashes (when loglevel being set to 0) that
255 * are quite hard to debug
256 */
257 if (get_option(&str, &newlevel)) {
258 console_loglevel = newlevel;
259 return 0;
260 }
261
262 return -EINVAL;
263 }
264
265 early_param("loglevel", loglevel);
266
267 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(size_t * _size)268 static void * __init get_boot_config_from_initrd(size_t *_size)
269 {
270 u32 size, csum;
271 char *data;
272 u32 *hdr;
273 int i;
274
275 if (!initrd_end)
276 return NULL;
277
278 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
279 /*
280 * Since Grub may align the size of initrd to 4, we must
281 * check the preceding 3 bytes as well.
282 */
283 for (i = 0; i < 4; i++) {
284 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
285 goto found;
286 data--;
287 }
288 return NULL;
289
290 found:
291 hdr = (u32 *)(data - 8);
292 size = le32_to_cpu(hdr[0]);
293 csum = le32_to_cpu(hdr[1]);
294
295 data = ((void *)hdr) - size;
296 if ((unsigned long)data < initrd_start) {
297 pr_err("bootconfig size %d is greater than initrd size %ld\n",
298 size, initrd_end - initrd_start);
299 return NULL;
300 }
301
302 if (xbc_calc_checksum(data, size) != csum) {
303 pr_err("bootconfig checksum failed\n");
304 return NULL;
305 }
306
307 /* Remove bootconfig from initramfs/initrd */
308 initrd_end = (unsigned long)data;
309 if (_size)
310 *_size = size;
311
312 return data;
313 }
314 #else
get_boot_config_from_initrd(size_t * _size)315 static void * __init get_boot_config_from_initrd(size_t *_size)
316 {
317 return NULL;
318 }
319 #endif
320
321 #ifdef CONFIG_BOOT_CONFIG
322
323 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
324
325 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
326
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)327 static int __init xbc_snprint_cmdline(char *buf, size_t size,
328 struct xbc_node *root)
329 {
330 struct xbc_node *knode, *vnode;
331 char *end = buf + size;
332 const char *val, *q;
333 int ret;
334
335 xbc_node_for_each_key_value(root, knode, val) {
336 ret = xbc_node_compose_key_after(root, knode,
337 xbc_namebuf, XBC_KEYLEN_MAX);
338 if (ret < 0)
339 return ret;
340
341 vnode = xbc_node_get_child(knode);
342 if (!vnode) {
343 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
344 if (ret < 0)
345 return ret;
346 buf += ret;
347 continue;
348 }
349 xbc_array_for_each_value(vnode, val) {
350 /*
351 * For prettier and more readable /proc/cmdline, only
352 * quote the value when necessary, i.e. when it contains
353 * whitespace.
354 */
355 q = strpbrk(val, " \t\r\n") ? "\"" : "";
356 ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
357 xbc_namebuf, q, val, q);
358 if (ret < 0)
359 return ret;
360 buf += ret;
361 }
362 }
363
364 return buf - (end - size);
365 }
366 #undef rest
367
368 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)369 static char * __init xbc_make_cmdline(const char *key)
370 {
371 struct xbc_node *root;
372 char *new_cmdline;
373 int ret, len = 0;
374
375 root = xbc_find_node(key);
376 if (!root)
377 return NULL;
378
379 /* Count required buffer size */
380 len = xbc_snprint_cmdline(NULL, 0, root);
381 if (len <= 0)
382 return NULL;
383
384 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
385 if (!new_cmdline) {
386 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
387 return NULL;
388 }
389
390 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
391 if (ret < 0 || ret > len) {
392 pr_err("Failed to print extra kernel cmdline.\n");
393 memblock_free(new_cmdline, len + 1);
394 return NULL;
395 }
396
397 return new_cmdline;
398 }
399
bootconfig_params(char * param,char * val,const char * unused,void * arg)400 static int __init bootconfig_params(char *param, char *val,
401 const char *unused, void *arg)
402 {
403 if (strcmp(param, "bootconfig") == 0) {
404 bootconfig_found = true;
405 }
406 return 0;
407 }
408
warn_bootconfig(char * str)409 static int __init warn_bootconfig(char *str)
410 {
411 /* The 'bootconfig' has been handled by bootconfig_params(). */
412 return 0;
413 }
414
setup_boot_config(void)415 static void __init setup_boot_config(void)
416 {
417 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
418 const char *msg, *data;
419 int pos, ret;
420 size_t size;
421 char *err;
422
423 /* Cut out the bootconfig data even if we have no bootconfig option */
424 data = get_boot_config_from_initrd(&size);
425 /* If there is no bootconfig in initrd, try embedded one. */
426 if (!data)
427 data = xbc_get_embedded_bootconfig(&size);
428
429 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
430 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
431 bootconfig_params);
432
433 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
434 return;
435
436 /* parse_args() stops at the next param of '--' and returns an address */
437 if (err)
438 initargs_offs = err - tmp_cmdline;
439
440 if (!data) {
441 /* If user intended to use bootconfig, show an error level message */
442 if (bootconfig_found)
443 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
444 else
445 pr_info("No bootconfig data provided, so skipping bootconfig");
446 return;
447 }
448
449 if (size >= XBC_DATA_MAX) {
450 pr_err("bootconfig size %ld greater than max size %d\n",
451 (long)size, XBC_DATA_MAX);
452 return;
453 }
454
455 ret = xbc_init(data, size, &msg, &pos);
456 if (ret < 0) {
457 if (pos < 0)
458 pr_err("Failed to init bootconfig: %s.\n", msg);
459 else
460 pr_err("Failed to parse bootconfig: %s at %d.\n",
461 msg, pos);
462 } else {
463 xbc_get_info(&ret, NULL);
464 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
465 /* keys starting with "kernel." are passed via cmdline */
466 extra_command_line = xbc_make_cmdline("kernel");
467 /* Also, "init." keys are init arguments */
468 extra_init_args = xbc_make_cmdline("init");
469 }
470 return;
471 }
472
exit_boot_config(void)473 static void __init exit_boot_config(void)
474 {
475 xbc_exit();
476 }
477
478 #else /* !CONFIG_BOOT_CONFIG */
479
setup_boot_config(void)480 static void __init setup_boot_config(void)
481 {
482 /* Remove bootconfig data from initrd */
483 get_boot_config_from_initrd(NULL);
484 }
485
warn_bootconfig(char * str)486 static int __init warn_bootconfig(char *str)
487 {
488 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
489 return 0;
490 }
491
492 #define exit_boot_config() do {} while (0)
493
494 #endif /* CONFIG_BOOT_CONFIG */
495
496 early_param("bootconfig", warn_bootconfig);
497
cmdline_has_extra_options(void)498 bool __init cmdline_has_extra_options(void)
499 {
500 return extra_command_line || extra_init_args;
501 }
502
503 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)504 static void __init repair_env_string(char *param, char *val)
505 {
506 if (val) {
507 /* param=val or param="val"? */
508 if (val == param+strlen(param)+1)
509 val[-1] = '=';
510 else if (val == param+strlen(param)+2) {
511 val[-2] = '=';
512 memmove(val-1, val, strlen(val)+1);
513 } else
514 BUG();
515 }
516 }
517
518 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)519 static int __init set_init_arg(char *param, char *val,
520 const char *unused, void *arg)
521 {
522 unsigned int i;
523
524 if (panic_later)
525 return 0;
526
527 repair_env_string(param, val);
528
529 for (i = 0; argv_init[i]; i++) {
530 if (i == MAX_INIT_ARGS) {
531 panic_later = "init";
532 panic_param = param;
533 return 0;
534 }
535 }
536 argv_init[i] = param;
537 return 0;
538 }
539
540 /*
541 * Unknown boot options get handed to init, unless they look like
542 * unused parameters (modprobe will find them in /proc/cmdline).
543 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)544 static int __init unknown_bootoption(char *param, char *val,
545 const char *unused, void *arg)
546 {
547 size_t len = strlen(param);
548
549 /* Handle params aliased to sysctls */
550 if (sysctl_is_alias(param))
551 return 0;
552
553 repair_env_string(param, val);
554
555 /* Handle obsolete-style parameters */
556 if (obsolete_checksetup(param))
557 return 0;
558
559 /* Unused module parameter. */
560 if (strnchr(param, len, '.'))
561 return 0;
562
563 if (panic_later)
564 return 0;
565
566 if (val) {
567 /* Environment option */
568 unsigned int i;
569 for (i = 0; envp_init[i]; i++) {
570 if (i == MAX_INIT_ENVS) {
571 panic_later = "env";
572 panic_param = param;
573 }
574 if (!strncmp(param, envp_init[i], len+1))
575 break;
576 }
577 envp_init[i] = param;
578 } else {
579 /* Command line option */
580 unsigned int i;
581 for (i = 0; argv_init[i]; i++) {
582 if (i == MAX_INIT_ARGS) {
583 panic_later = "init";
584 panic_param = param;
585 }
586 }
587 argv_init[i] = param;
588 }
589 return 0;
590 }
591
init_setup(char * str)592 static int __init init_setup(char *str)
593 {
594 unsigned int i;
595
596 execute_command = str;
597 /*
598 * In case LILO is going to boot us with default command line,
599 * it prepends "auto" before the whole cmdline which makes
600 * the shell think it should execute a script with such name.
601 * So we ignore all arguments entered _before_ init=... [MJ]
602 */
603 for (i = 1; i < MAX_INIT_ARGS; i++)
604 argv_init[i] = NULL;
605 return 1;
606 }
607 __setup("init=", init_setup);
608
rdinit_setup(char * str)609 static int __init rdinit_setup(char *str)
610 {
611 unsigned int i;
612
613 ramdisk_execute_command = str;
614 /* See "auto" comment in init_setup */
615 for (i = 1; i < MAX_INIT_ARGS; i++)
616 argv_init[i] = NULL;
617 return 1;
618 }
619 __setup("rdinit=", rdinit_setup);
620
621 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)622 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)623 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
624 #endif
625
626 /*
627 * We need to store the untouched command line for future reference.
628 * We also need to store the touched command line since the parameter
629 * parsing is performed in place, and we should allow a component to
630 * store reference of name/value for future reference.
631 */
setup_command_line(char * command_line)632 static void __init setup_command_line(char *command_line)
633 {
634 size_t len, xlen = 0, ilen = 0;
635
636 if (extra_command_line)
637 xlen = strlen(extra_command_line);
638 if (extra_init_args) {
639 extra_init_args = strim(extra_init_args); /* remove trailing space */
640 ilen = strlen(extra_init_args) + 4; /* for " -- " */
641 }
642
643 len = xlen + strlen(boot_command_line) + ilen + 1;
644
645 saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
646
647 len = xlen + strlen(command_line) + 1;
648
649 static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
650
651 if (xlen) {
652 /*
653 * We have to put extra_command_line before boot command
654 * lines because there could be dashes (separator of init
655 * command line) in the command lines.
656 */
657 strcpy(saved_command_line, extra_command_line);
658 strcpy(static_command_line, extra_command_line);
659 }
660 strcpy(saved_command_line + xlen, boot_command_line);
661 strcpy(static_command_line + xlen, command_line);
662
663 if (ilen) {
664 /*
665 * Append supplemental init boot args to saved_command_line
666 * so that user can check what command line options passed
667 * to init.
668 * The order should always be
669 * " -- "[bootconfig init-param][cmdline init-param]
670 */
671 if (initargs_offs) {
672 len = xlen + initargs_offs;
673 strcpy(saved_command_line + len, extra_init_args);
674 len += ilen - 4; /* strlen(extra_init_args) */
675 strcpy(saved_command_line + len,
676 boot_command_line + initargs_offs - 1);
677 } else {
678 len = strlen(saved_command_line);
679 strcpy(saved_command_line + len, " -- ");
680 len += 4;
681 strcpy(saved_command_line + len, extra_init_args);
682 }
683 }
684
685 saved_command_line_len = strlen(saved_command_line);
686 }
687
688 /*
689 * We need to finalize in a non-__init function or else race conditions
690 * between the root thread and the init thread may cause start_kernel to
691 * be reaped by free_initmem before the root thread has proceeded to
692 * cpu_idle.
693 *
694 * gcc-3.4 accidentally inlines this function, so use noinline.
695 */
696
697 static __initdata DECLARE_COMPLETION(kthreadd_done);
698
rest_init(void)699 static noinline void __ref __noreturn rest_init(void)
700 {
701 struct task_struct *tsk;
702 int pid;
703
704 rcu_scheduler_starting();
705 /*
706 * We need to spawn init first so that it obtains pid 1, however
707 * the init task will end up wanting to create kthreads, which, if
708 * we schedule it before we create kthreadd, will OOPS.
709 */
710 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
711 /*
712 * Pin init on the boot CPU. Task migration is not properly working
713 * until sched_init_smp() has been run. It will set the allowed
714 * CPUs for init to the non isolated CPUs.
715 */
716 rcu_read_lock();
717 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
718 tsk->flags |= PF_NO_SETAFFINITY;
719 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
720 rcu_read_unlock();
721
722 numa_default_policy();
723 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
724 rcu_read_lock();
725 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
726 rcu_read_unlock();
727
728 /*
729 * Enable might_sleep() and smp_processor_id() checks.
730 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
731 * kernel_thread() would trigger might_sleep() splats. With
732 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
733 * already, but it's stuck on the kthreadd_done completion.
734 */
735 system_state = SYSTEM_SCHEDULING;
736
737 complete(&kthreadd_done);
738
739 /*
740 * The boot idle thread must execute schedule()
741 * at least once to get things moving:
742 */
743 schedule_preempt_disabled();
744 /* Call into cpu_idle with preempt disabled */
745 cpu_startup_entry(CPUHP_ONLINE);
746 }
747
748 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)749 static int __init do_early_param(char *param, char *val,
750 const char *unused, void *arg)
751 {
752 const struct obs_kernel_param *p;
753
754 for (p = __setup_start; p < __setup_end; p++) {
755 if (p->early && parameq(param, p->str)) {
756 if (p->setup_func(val) != 0)
757 pr_warn("Malformed early option '%s'\n", param);
758 }
759 }
760 /* We accept everything at this stage. */
761 return 0;
762 }
763
parse_early_options(char * cmdline)764 void __init parse_early_options(char *cmdline)
765 {
766 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
767 do_early_param);
768 }
769
770 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)771 void __init parse_early_param(void)
772 {
773 static int done __initdata;
774 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
775
776 if (done)
777 return;
778
779 /* All fall through to do_early_param. */
780 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
781 parse_early_options(tmp_cmdline);
782 done = 1;
783 }
784
arch_post_acpi_subsys_init(void)785 void __init __weak arch_post_acpi_subsys_init(void) { }
786
smp_setup_processor_id(void)787 void __init __weak smp_setup_processor_id(void)
788 {
789 }
790
smp_prepare_boot_cpu(void)791 void __init __weak smp_prepare_boot_cpu(void)
792 {
793 }
794
795 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)796 void __init __weak thread_stack_cache_init(void)
797 {
798 }
799 #endif
800
poking_init(void)801 void __init __weak poking_init(void) { }
802
pgtable_cache_init(void)803 void __init __weak pgtable_cache_init(void) { }
804
trap_init(void)805 void __init __weak trap_init(void) { }
806
807 bool initcall_debug;
808 core_param(initcall_debug, initcall_debug, bool, 0644);
809
810 #ifdef TRACEPOINTS_ENABLED
811 static void __init initcall_debug_enable(void);
812 #else
initcall_debug_enable(void)813 static inline void initcall_debug_enable(void)
814 {
815 }
816 #endif
817
818 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
819 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
820 randomize_kstack_offset);
821 DEFINE_PER_CPU(u32, kstack_offset);
822
early_randomize_kstack_offset(char * buf)823 static int __init early_randomize_kstack_offset(char *buf)
824 {
825 int ret;
826 bool bool_result;
827
828 ret = kstrtobool(buf, &bool_result);
829 if (ret)
830 return ret;
831
832 if (bool_result)
833 static_branch_enable(&randomize_kstack_offset);
834 else
835 static_branch_disable(&randomize_kstack_offset);
836 return 0;
837 }
838 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
839 #endif
840
print_unknown_bootoptions(void)841 static void __init print_unknown_bootoptions(void)
842 {
843 char *unknown_options;
844 char *end;
845 const char *const *p;
846 size_t len;
847
848 if (panic_later || (!argv_init[1] && !envp_init[2]))
849 return;
850
851 /*
852 * Determine how many options we have to print out, plus a space
853 * before each
854 */
855 len = 1; /* null terminator */
856 for (p = &argv_init[1]; *p; p++) {
857 len++;
858 len += strlen(*p);
859 }
860 for (p = &envp_init[2]; *p; p++) {
861 len++;
862 len += strlen(*p);
863 }
864
865 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
866 if (!unknown_options) {
867 pr_err("%s: Failed to allocate %zu bytes\n",
868 __func__, len);
869 return;
870 }
871 end = unknown_options;
872
873 for (p = &argv_init[1]; *p; p++)
874 end += sprintf(end, " %s", *p);
875 for (p = &envp_init[2]; *p; p++)
876 end += sprintf(end, " %s", *p);
877
878 /* Start at unknown_options[1] to skip the initial space */
879 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
880 &unknown_options[1]);
881 memblock_free(unknown_options, len);
882 }
883
early_numa_node_init(void)884 static void __init early_numa_node_init(void)
885 {
886 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
887 #ifndef cpu_to_node
888 int cpu;
889
890 /* The early_cpu_to_node() should be ready here. */
891 for_each_possible_cpu(cpu)
892 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
893 #endif
894 #endif
895 }
896
897 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)898 void start_kernel(void)
899 {
900 char *command_line;
901 char *after_dashes;
902
903 set_task_stack_end_magic(&init_task);
904 smp_setup_processor_id();
905 debug_objects_early_init();
906 init_vmlinux_build_id();
907
908 cgroup_init_early();
909
910 local_irq_disable();
911 early_boot_irqs_disabled = true;
912
913 /*
914 * Interrupts are still disabled. Do necessary setups, then
915 * enable them.
916 */
917 boot_cpu_init();
918 page_address_init();
919 pr_notice("%s", linux_banner);
920 setup_arch(&command_line);
921 /* Static keys and static calls are needed by LSMs */
922 jump_label_init();
923 static_call_init();
924 early_security_init();
925 setup_boot_config();
926 setup_command_line(command_line);
927 setup_nr_cpu_ids();
928 setup_per_cpu_areas();
929 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
930 early_numa_node_init();
931 boot_cpu_hotplug_init();
932
933 pr_notice("Kernel command line: %s\n", saved_command_line);
934 /* parameters may set static keys */
935 parse_early_param();
936 after_dashes = parse_args("Booting kernel",
937 static_command_line, __start___param,
938 __stop___param - __start___param,
939 -1, -1, NULL, &unknown_bootoption);
940 print_unknown_bootoptions();
941 if (!IS_ERR_OR_NULL(after_dashes))
942 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
943 NULL, set_init_arg);
944 if (extra_init_args)
945 parse_args("Setting extra init args", extra_init_args,
946 NULL, 0, -1, -1, NULL, set_init_arg);
947
948 /* Architectural and non-timekeeping rng init, before allocator init */
949 random_init_early(command_line);
950
951 /*
952 * These use large bootmem allocations and must precede
953 * initalization of page allocator
954 */
955 setup_log_buf(0);
956 vfs_caches_init_early();
957 sort_main_extable();
958 trap_init();
959 mm_core_init();
960 poking_init();
961 ftrace_init();
962
963 /* trace_printk can be enabled here */
964 early_trace_init();
965
966 /*
967 * Set up the scheduler prior starting any interrupts (such as the
968 * timer interrupt). Full topology setup happens at smp_init()
969 * time - but meanwhile we still have a functioning scheduler.
970 */
971 sched_init();
972
973 if (WARN(!irqs_disabled(),
974 "Interrupts were enabled *very* early, fixing it\n"))
975 local_irq_disable();
976 radix_tree_init();
977 maple_tree_init();
978
979 /*
980 * Set up housekeeping before setting up workqueues to allow the unbound
981 * workqueue to take non-housekeeping into account.
982 */
983 housekeeping_init();
984
985 /*
986 * Allow workqueue creation and work item queueing/cancelling
987 * early. Work item execution depends on kthreads and starts after
988 * workqueue_init().
989 */
990 workqueue_init_early();
991
992 rcu_init();
993 kvfree_rcu_init();
994
995 /* Trace events are available after this */
996 trace_init();
997
998 if (initcall_debug)
999 initcall_debug_enable();
1000
1001 context_tracking_init();
1002 /* init some links before init_ISA_irqs() */
1003 early_irq_init();
1004 init_IRQ();
1005 tick_init();
1006 rcu_init_nohz();
1007 timers_init();
1008 srcu_init();
1009 hrtimers_init();
1010 softirq_init();
1011 timekeeping_init();
1012 time_init();
1013
1014 /* This must be after timekeeping is initialized */
1015 random_init();
1016
1017 /* These make use of the fully initialized rng */
1018 kfence_init();
1019 boot_init_stack_canary();
1020
1021 perf_event_init();
1022 profile_init();
1023 call_function_init();
1024 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1025
1026 early_boot_irqs_disabled = false;
1027 local_irq_enable();
1028
1029 kmem_cache_init_late();
1030
1031 /*
1032 * HACK ALERT! This is early. We're enabling the console before
1033 * we've done PCI setups etc, and console_init() must be aware of
1034 * this. But we do want output early, in case something goes wrong.
1035 */
1036 console_init();
1037 if (panic_later)
1038 panic("Too many boot %s vars at `%s'", panic_later,
1039 panic_param);
1040
1041 lockdep_init();
1042
1043 /*
1044 * Need to run this when irqs are enabled, because it wants
1045 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1046 * too:
1047 */
1048 locking_selftest();
1049
1050 #ifdef CONFIG_BLK_DEV_INITRD
1051 if (initrd_start && !initrd_below_start_ok &&
1052 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1053 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1054 page_to_pfn(virt_to_page((void *)initrd_start)),
1055 min_low_pfn);
1056 initrd_start = 0;
1057 }
1058 #endif
1059 setup_per_cpu_pageset();
1060 numa_policy_init();
1061 acpi_early_init();
1062 if (late_time_init)
1063 late_time_init();
1064 sched_clock_init();
1065 calibrate_delay();
1066
1067 arch_cpu_finalize_init();
1068
1069 pid_idr_init();
1070 anon_vma_init();
1071 #ifdef CONFIG_X86
1072 if (efi_enabled(EFI_RUNTIME_SERVICES))
1073 efi_enter_virtual_mode();
1074 #endif
1075 thread_stack_cache_init();
1076 cred_init();
1077 fork_init();
1078 proc_caches_init();
1079 uts_ns_init();
1080 key_init();
1081 security_init();
1082 dbg_late_init();
1083 net_ns_init();
1084 vfs_caches_init();
1085 pagecache_init();
1086 signals_init();
1087 seq_file_init();
1088 proc_root_init();
1089 nsfs_init();
1090 pidfs_init();
1091 cpuset_init();
1092 mem_cgroup_init();
1093 cgroup_init();
1094 taskstats_init_early();
1095 delayacct_init();
1096
1097 acpi_subsystem_init();
1098 arch_post_acpi_subsys_init();
1099 kcsan_init();
1100
1101 /* Do the rest non-__init'ed, we're now alive */
1102 rest_init();
1103
1104 /*
1105 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1106 * and older.
1107 */
1108 #if !__has_attribute(__no_stack_protector__)
1109 prevent_tail_call_optimization();
1110 #endif
1111 }
1112
1113 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1114 static void __init do_ctors(void)
1115 {
1116 /*
1117 * For UML, the constructors have already been called by the
1118 * normal setup code as it's just a normal ELF binary, so we
1119 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1120 * even on UML for modules.
1121 */
1122 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1123 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1124
1125 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1126 (*fn)();
1127 #endif
1128 }
1129
1130 #ifdef CONFIG_KALLSYMS
1131 struct blacklist_entry {
1132 struct list_head next;
1133 char *buf;
1134 };
1135
1136 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1137
initcall_blacklist(char * str)1138 static int __init initcall_blacklist(char *str)
1139 {
1140 char *str_entry;
1141 struct blacklist_entry *entry;
1142
1143 /* str argument is a comma-separated list of functions */
1144 do {
1145 str_entry = strsep(&str, ",");
1146 if (str_entry) {
1147 pr_debug("blacklisting initcall %s\n", str_entry);
1148 entry = memblock_alloc_or_panic(sizeof(*entry),
1149 SMP_CACHE_BYTES);
1150 entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1,
1151 SMP_CACHE_BYTES);
1152 strcpy(entry->buf, str_entry);
1153 list_add(&entry->next, &blacklisted_initcalls);
1154 }
1155 } while (str_entry);
1156
1157 return 1;
1158 }
1159
initcall_blacklisted(initcall_t fn)1160 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1161 {
1162 struct blacklist_entry *entry;
1163 char fn_name[KSYM_SYMBOL_LEN];
1164 unsigned long addr;
1165
1166 if (list_empty(&blacklisted_initcalls))
1167 return false;
1168
1169 addr = (unsigned long) dereference_function_descriptor(fn);
1170 sprint_symbol_no_offset(fn_name, addr);
1171
1172 /*
1173 * fn will be "function_name [module_name]" where [module_name] is not
1174 * displayed for built-in init functions. Strip off the [module_name].
1175 */
1176 strreplace(fn_name, ' ', '\0');
1177
1178 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1179 if (!strcmp(fn_name, entry->buf)) {
1180 pr_debug("initcall %s blacklisted\n", fn_name);
1181 return true;
1182 }
1183 }
1184
1185 return false;
1186 }
1187 #else
initcall_blacklist(char * str)1188 static int __init initcall_blacklist(char *str)
1189 {
1190 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1191 return 0;
1192 }
1193
initcall_blacklisted(initcall_t fn)1194 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1195 {
1196 return false;
1197 }
1198 #endif
1199 __setup("initcall_blacklist=", initcall_blacklist);
1200
1201 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1202 trace_initcall_start_cb(void *data, initcall_t fn)
1203 {
1204 ktime_t *calltime = data;
1205
1206 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1207 *calltime = ktime_get();
1208 }
1209
1210 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1211 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1212 {
1213 ktime_t rettime, *calltime = data;
1214
1215 rettime = ktime_get();
1216 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1217 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1218 }
1219
1220 static __init_or_module void
trace_initcall_level_cb(void * data,const char * level)1221 trace_initcall_level_cb(void *data, const char *level)
1222 {
1223 printk(KERN_DEBUG "entering initcall level: %s\n", level);
1224 }
1225
1226 static ktime_t initcall_calltime;
1227
1228 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1229 static void __init initcall_debug_enable(void)
1230 {
1231 int ret;
1232
1233 ret = register_trace_initcall_start(trace_initcall_start_cb,
1234 &initcall_calltime);
1235 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1236 &initcall_calltime);
1237 ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL);
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 # define do_trace_initcall_level trace_initcall_level
1243 #else
do_trace_initcall_start(initcall_t fn)1244 static inline void do_trace_initcall_start(initcall_t fn)
1245 {
1246 if (!initcall_debug)
1247 return;
1248 trace_initcall_start_cb(&initcall_calltime, fn);
1249 }
do_trace_initcall_finish(initcall_t fn,int ret)1250 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1251 {
1252 if (!initcall_debug)
1253 return;
1254 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1255 }
do_trace_initcall_level(const char * level)1256 static inline void do_trace_initcall_level(const char *level)
1257 {
1258 if (!initcall_debug)
1259 return;
1260 trace_initcall_level_cb(NULL, level);
1261 }
1262 #endif /* !TRACEPOINTS_ENABLED */
1263
do_one_initcall(initcall_t fn)1264 int __init_or_module do_one_initcall(initcall_t fn)
1265 {
1266 int count = preempt_count();
1267 char msgbuf[64];
1268 int ret;
1269
1270 if (initcall_blacklisted(fn))
1271 return -EPERM;
1272
1273 do_trace_initcall_start(fn);
1274 ret = fn();
1275 do_trace_initcall_finish(fn, ret);
1276
1277 msgbuf[0] = 0;
1278
1279 if (preempt_count() != count) {
1280 sprintf(msgbuf, "preemption imbalance ");
1281 preempt_count_set(count);
1282 }
1283 if (irqs_disabled()) {
1284 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1285 local_irq_enable();
1286 }
1287 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1288
1289 add_latent_entropy();
1290 return ret;
1291 }
1292
1293
1294 static initcall_entry_t *initcall_levels[] __initdata = {
1295 __initcall0_start,
1296 __initcall1_start,
1297 __initcall2_start,
1298 __initcall3_start,
1299 __initcall4_start,
1300 __initcall5_start,
1301 __initcall6_start,
1302 __initcall7_start,
1303 __initcall_end,
1304 };
1305
1306 /* Keep these in sync with initcalls in include/linux/init.h */
1307 static const char *initcall_level_names[] __initdata = {
1308 "pure",
1309 "core",
1310 "postcore",
1311 "arch",
1312 "subsys",
1313 "fs",
1314 "device",
1315 "late",
1316 };
1317
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1318 static int __init ignore_unknown_bootoption(char *param, char *val,
1319 const char *unused, void *arg)
1320 {
1321 return 0;
1322 }
1323
do_initcall_level(int level,char * command_line)1324 static void __init do_initcall_level(int level, char *command_line)
1325 {
1326 initcall_entry_t *fn;
1327
1328 parse_args(initcall_level_names[level],
1329 command_line, __start___param,
1330 __stop___param - __start___param,
1331 level, level,
1332 NULL, ignore_unknown_bootoption);
1333
1334 do_trace_initcall_level(initcall_level_names[level]);
1335 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1336 do_one_initcall(initcall_from_entry(fn));
1337 }
1338
do_initcalls(void)1339 static void __init do_initcalls(void)
1340 {
1341 int level;
1342 size_t len = saved_command_line_len + 1;
1343 char *command_line;
1344
1345 command_line = kzalloc(len, GFP_KERNEL);
1346 if (!command_line)
1347 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1348
1349 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1350 /* Parser modifies command_line, restore it each time */
1351 strcpy(command_line, saved_command_line);
1352 do_initcall_level(level, command_line);
1353 }
1354
1355 kfree(command_line);
1356 }
1357
1358 /*
1359 * Ok, the machine is now initialized. None of the devices
1360 * have been touched yet, but the CPU subsystem is up and
1361 * running, and memory and process management works.
1362 *
1363 * Now we can finally start doing some real work..
1364 */
do_basic_setup(void)1365 static void __init do_basic_setup(void)
1366 {
1367 cpuset_init_smp();
1368 driver_init();
1369 init_irq_proc();
1370 do_ctors();
1371 do_initcalls();
1372 }
1373
do_pre_smp_initcalls(void)1374 static void __init do_pre_smp_initcalls(void)
1375 {
1376 initcall_entry_t *fn;
1377
1378 do_trace_initcall_level("early");
1379 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1380 do_one_initcall(initcall_from_entry(fn));
1381 }
1382
run_init_process(const char * init_filename)1383 static int run_init_process(const char *init_filename)
1384 {
1385 const char *const *p;
1386
1387 argv_init[0] = init_filename;
1388 pr_info("Run %s as init process\n", init_filename);
1389 pr_debug(" with arguments:\n");
1390 for (p = argv_init; *p; p++)
1391 pr_debug(" %s\n", *p);
1392 pr_debug(" with environment:\n");
1393 for (p = envp_init; *p; p++)
1394 pr_debug(" %s\n", *p);
1395 return kernel_execve(init_filename, argv_init, envp_init);
1396 }
1397
try_to_run_init_process(const char * init_filename)1398 static int try_to_run_init_process(const char *init_filename)
1399 {
1400 int ret;
1401
1402 ret = run_init_process(init_filename);
1403
1404 if (ret && ret != -ENOENT) {
1405 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1406 init_filename, ret);
1407 }
1408
1409 return ret;
1410 }
1411
1412 static noinline void __init kernel_init_freeable(void);
1413
1414 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1415 bool rodata_enabled __ro_after_init = true;
1416
1417 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1418 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1419 #endif
1420
set_debug_rodata(char * str)1421 static int __init set_debug_rodata(char *str)
1422 {
1423 if (arch_parse_debug_rodata(str))
1424 return 0;
1425
1426 if (str && !strcmp(str, "on"))
1427 rodata_enabled = true;
1428 else if (str && !strcmp(str, "off"))
1429 rodata_enabled = false;
1430 else
1431 pr_warn("Invalid option string for rodata: '%s'\n", str);
1432 return 0;
1433 }
1434 early_param("rodata", set_debug_rodata);
1435 #endif
1436
mark_readonly(void)1437 static void mark_readonly(void)
1438 {
1439 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1440 /*
1441 * load_module() results in W+X mappings, which are cleaned
1442 * up with init_free_wq. Let's make sure that queued work is
1443 * flushed so that we don't hit false positives looking for
1444 * insecure pages which are W+X.
1445 */
1446 flush_module_init_free_work();
1447 jump_label_init_ro();
1448 mark_rodata_ro();
1449 debug_checkwx();
1450 rodata_test();
1451 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1452 pr_info("Kernel memory protection disabled.\n");
1453 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1454 pr_warn("Kernel memory protection not selected by kernel config.\n");
1455 } else {
1456 pr_warn("This architecture does not have kernel memory protection.\n");
1457 }
1458 }
1459
free_initmem(void)1460 void __weak free_initmem(void)
1461 {
1462 free_initmem_default(POISON_FREE_INITMEM);
1463 }
1464
kernel_init(void * unused)1465 static int __ref kernel_init(void *unused)
1466 {
1467 int ret;
1468
1469 /*
1470 * Wait until kthreadd is all set-up.
1471 */
1472 wait_for_completion(&kthreadd_done);
1473
1474 kernel_init_freeable();
1475 /* need to finish all async __init code before freeing the memory */
1476 async_synchronize_full();
1477
1478 system_state = SYSTEM_FREEING_INITMEM;
1479 kprobe_free_init_mem();
1480 ftrace_free_init_mem();
1481 kgdb_free_init_mem();
1482 exit_boot_config();
1483 free_initmem();
1484 mark_readonly();
1485
1486 /*
1487 * Kernel mappings are now finalized - update the userspace page-table
1488 * to finalize PTI.
1489 */
1490 pti_finalize();
1491
1492 system_state = SYSTEM_RUNNING;
1493 numa_default_policy();
1494
1495 rcu_end_inkernel_boot();
1496
1497 do_sysctl_args();
1498
1499 if (ramdisk_execute_command) {
1500 ret = run_init_process(ramdisk_execute_command);
1501 if (!ret)
1502 return 0;
1503 pr_err("Failed to execute %s (error %d)\n",
1504 ramdisk_execute_command, ret);
1505 }
1506
1507 /*
1508 * We try each of these until one succeeds.
1509 *
1510 * The Bourne shell can be used instead of init if we are
1511 * trying to recover a really broken machine.
1512 */
1513 if (execute_command) {
1514 ret = run_init_process(execute_command);
1515 if (!ret)
1516 return 0;
1517 panic("Requested init %s failed (error %d).",
1518 execute_command, ret);
1519 }
1520
1521 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1522 ret = run_init_process(CONFIG_DEFAULT_INIT);
1523 if (ret)
1524 pr_err("Default init %s failed (error %d)\n",
1525 CONFIG_DEFAULT_INIT, ret);
1526 else
1527 return 0;
1528 }
1529
1530 if (!try_to_run_init_process("/sbin/init") ||
1531 !try_to_run_init_process("/etc/init") ||
1532 !try_to_run_init_process("/bin/init") ||
1533 !try_to_run_init_process("/bin/sh"))
1534 return 0;
1535
1536 panic("No working init found. Try passing init= option to kernel. "
1537 "See Linux Documentation/admin-guide/init.rst for guidance.");
1538 }
1539
1540 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1541 void __init console_on_rootfs(void)
1542 {
1543 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1544
1545 if (IS_ERR(file)) {
1546 pr_err("Warning: unable to open an initial console.\n");
1547 return;
1548 }
1549 init_dup(file);
1550 init_dup(file);
1551 init_dup(file);
1552 fput(file);
1553 }
1554
kernel_init_freeable(void)1555 static noinline void __init kernel_init_freeable(void)
1556 {
1557 /* Now the scheduler is fully set up and can do blocking allocations */
1558 gfp_allowed_mask = __GFP_BITS_MASK;
1559
1560 /*
1561 * init can allocate pages on any node
1562 */
1563 set_mems_allowed(node_states[N_MEMORY]);
1564
1565 cad_pid = get_pid(task_pid(current));
1566
1567 smp_prepare_cpus(setup_max_cpus);
1568
1569 workqueue_init();
1570
1571 init_mm_internals();
1572
1573 do_pre_smp_initcalls();
1574 lockup_detector_init();
1575
1576 smp_init();
1577 sched_init_smp();
1578
1579 workqueue_init_topology();
1580 async_init();
1581 padata_init();
1582 page_alloc_init_late();
1583
1584 do_basic_setup();
1585
1586 kunit_run_all_tests();
1587
1588 wait_for_initramfs();
1589 console_on_rootfs();
1590
1591 /*
1592 * check if there is an early userspace init. If yes, let it do all
1593 * the work
1594 */
1595 if (init_eaccess(ramdisk_execute_command) != 0) {
1596 ramdisk_execute_command = NULL;
1597 prepare_namespace();
1598 }
1599
1600 /*
1601 * Ok, we have completed the initial bootup, and
1602 * we're essentially up and running. Get rid of the
1603 * initmem segments and start the user-mode stuff..
1604 *
1605 * rootfs is available now, try loading the public keys
1606 * and default modules
1607 */
1608
1609 integrity_load_keys();
1610 }
1611