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