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 <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 * Well-known bootloader identifiers:
550 * 1. LILO/Grub pass "BOOT_IMAGE=...";
551 * 2. kexec/kdump (kexec-tools) pass "kexec".
552 */
553 const char *bootloader[] = { "BOOT_IMAGE=", "kexec", NULL };
554
555 /* Handle params aliased to sysctls */
556 if (sysctl_is_alias(param))
557 return 0;
558
559 repair_env_string(param, val);
560
561 /* Handle bootloader identifier */
562 for (int i = 0; bootloader[i]; i++) {
563 if (strstarts(param, bootloader[i]))
564 return 0;
565 }
566
567 /* Handle obsolete-style parameters */
568 if (obsolete_checksetup(param))
569 return 0;
570
571 /* Unused module parameter. */
572 if (strnchr(param, len, '.'))
573 return 0;
574
575 if (panic_later)
576 return 0;
577
578 if (val) {
579 /* Environment option */
580 unsigned int i;
581 for (i = 0; envp_init[i]; i++) {
582 if (i == MAX_INIT_ENVS) {
583 panic_later = "env";
584 panic_param = param;
585 }
586 if (!strncmp(param, envp_init[i], len+1))
587 break;
588 }
589 envp_init[i] = param;
590 } else {
591 /* Command line option */
592 unsigned int i;
593 for (i = 0; argv_init[i]; i++) {
594 if (i == MAX_INIT_ARGS) {
595 panic_later = "init";
596 panic_param = param;
597 }
598 }
599 argv_init[i] = param;
600 }
601 return 0;
602 }
603
init_setup(char * str)604 static int __init init_setup(char *str)
605 {
606 unsigned int i;
607
608 execute_command = str;
609 /*
610 * In case LILO is going to boot us with default command line,
611 * it prepends "auto" before the whole cmdline which makes
612 * the shell think it should execute a script with such name.
613 * So we ignore all arguments entered _before_ init=... [MJ]
614 */
615 for (i = 1; i < MAX_INIT_ARGS; i++)
616 argv_init[i] = NULL;
617 return 1;
618 }
619 __setup("init=", init_setup);
620
rdinit_setup(char * str)621 static int __init rdinit_setup(char *str)
622 {
623 unsigned int i;
624
625 ramdisk_execute_command = str;
626 /* See "auto" comment in init_setup */
627 for (i = 1; i < MAX_INIT_ARGS; i++)
628 argv_init[i] = NULL;
629 return 1;
630 }
631 __setup("rdinit=", rdinit_setup);
632
633 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)634 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)635 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
636 #endif
637
638 /*
639 * We need to store the untouched command line for future reference.
640 * We also need to store the touched command line since the parameter
641 * parsing is performed in place, and we should allow a component to
642 * store reference of name/value for future reference.
643 */
setup_command_line(char * command_line)644 static void __init setup_command_line(char *command_line)
645 {
646 size_t len, xlen = 0, ilen = 0;
647
648 if (extra_command_line)
649 xlen = strlen(extra_command_line);
650 if (extra_init_args) {
651 extra_init_args = strim(extra_init_args); /* remove trailing space */
652 ilen = strlen(extra_init_args) + 4; /* for " -- " */
653 }
654
655 len = xlen + strlen(boot_command_line) + ilen + 1;
656
657 saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
658
659 len = xlen + strlen(command_line) + 1;
660
661 static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
662
663 if (xlen) {
664 /*
665 * We have to put extra_command_line before boot command
666 * lines because there could be dashes (separator of init
667 * command line) in the command lines.
668 */
669 strcpy(saved_command_line, extra_command_line);
670 strcpy(static_command_line, extra_command_line);
671 }
672 strcpy(saved_command_line + xlen, boot_command_line);
673 strcpy(static_command_line + xlen, command_line);
674
675 if (ilen) {
676 /*
677 * Append supplemental init boot args to saved_command_line
678 * so that user can check what command line options passed
679 * to init.
680 * The order should always be
681 * " -- "[bootconfig init-param][cmdline init-param]
682 */
683 if (initargs_offs) {
684 len = xlen + initargs_offs;
685 strcpy(saved_command_line + len, extra_init_args);
686 len += ilen - 4; /* strlen(extra_init_args) */
687 strcpy(saved_command_line + len,
688 boot_command_line + initargs_offs - 1);
689 } else {
690 len = strlen(saved_command_line);
691 strcpy(saved_command_line + len, " -- ");
692 len += 4;
693 strcpy(saved_command_line + len, extra_init_args);
694 }
695 }
696
697 saved_command_line_len = strlen(saved_command_line);
698 }
699
700 /*
701 * We need to finalize in a non-__init function or else race conditions
702 * between the root thread and the init thread may cause start_kernel to
703 * be reaped by free_initmem before the root thread has proceeded to
704 * cpu_idle.
705 *
706 * gcc-3.4 accidentally inlines this function, so use noinline.
707 */
708
709 static __initdata DECLARE_COMPLETION(kthreadd_done);
710
rest_init(void)711 static noinline void __ref __noreturn rest_init(void)
712 {
713 struct task_struct *tsk;
714 int pid;
715
716 rcu_scheduler_starting();
717 /*
718 * We need to spawn init first so that it obtains pid 1, however
719 * the init task will end up wanting to create kthreads, which, if
720 * we schedule it before we create kthreadd, will OOPS.
721 */
722 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
723 /*
724 * Pin init on the boot CPU. Task migration is not properly working
725 * until sched_init_smp() has been run. It will set the allowed
726 * CPUs for init to the non isolated CPUs.
727 */
728 rcu_read_lock();
729 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
730 tsk->flags |= PF_NO_SETAFFINITY;
731 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
732 rcu_read_unlock();
733
734 numa_default_policy();
735 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
736 rcu_read_lock();
737 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
738 rcu_read_unlock();
739
740 /*
741 * Enable might_sleep() and smp_processor_id() checks.
742 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
743 * kernel_thread() would trigger might_sleep() splats. With
744 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
745 * already, but it's stuck on the kthreadd_done completion.
746 */
747 system_state = SYSTEM_SCHEDULING;
748
749 complete(&kthreadd_done);
750
751 /*
752 * The boot idle thread must execute schedule()
753 * at least once to get things moving:
754 */
755 schedule_preempt_disabled();
756 /* Call into cpu_idle with preempt disabled */
757 cpu_startup_entry(CPUHP_ONLINE);
758 }
759
760 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)761 static int __init do_early_param(char *param, char *val,
762 const char *unused, void *arg)
763 {
764 const struct obs_kernel_param *p;
765
766 for (p = __setup_start; p < __setup_end; p++) {
767 if (p->early && parameq(param, p->str)) {
768 if (p->setup_func(val) != 0)
769 pr_warn("Malformed early option '%s'\n", param);
770 }
771 }
772 /* We accept everything at this stage. */
773 return 0;
774 }
775
parse_early_options(char * cmdline)776 void __init parse_early_options(char *cmdline)
777 {
778 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
779 do_early_param);
780 }
781
782 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)783 void __init parse_early_param(void)
784 {
785 static int done __initdata;
786 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
787
788 if (done)
789 return;
790
791 /* All fall through to do_early_param. */
792 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
793 parse_early_options(tmp_cmdline);
794 done = 1;
795 }
796
arch_post_acpi_subsys_init(void)797 void __init __weak arch_post_acpi_subsys_init(void) { }
798
smp_setup_processor_id(void)799 void __init __weak smp_setup_processor_id(void)
800 {
801 }
802
smp_prepare_boot_cpu(void)803 void __init __weak smp_prepare_boot_cpu(void)
804 {
805 }
806
807 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)808 void __init __weak thread_stack_cache_init(void)
809 {
810 }
811 #endif
812
poking_init(void)813 void __init __weak poking_init(void) { }
814
pgtable_cache_init(void)815 void __init __weak pgtable_cache_init(void) { }
816
trap_init(void)817 void __init __weak trap_init(void) { }
818
819 bool initcall_debug;
820 core_param(initcall_debug, initcall_debug, bool, 0644);
821
822 #ifdef TRACEPOINTS_ENABLED
823 static void __init initcall_debug_enable(void);
824 #else
initcall_debug_enable(void)825 static inline void initcall_debug_enable(void)
826 {
827 }
828 #endif
829
830 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
831 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
832 randomize_kstack_offset);
833 DEFINE_PER_CPU(u32, kstack_offset);
834
early_randomize_kstack_offset(char * buf)835 static int __init early_randomize_kstack_offset(char *buf)
836 {
837 int ret;
838 bool bool_result;
839
840 ret = kstrtobool(buf, &bool_result);
841 if (ret)
842 return ret;
843
844 if (bool_result)
845 static_branch_enable(&randomize_kstack_offset);
846 else
847 static_branch_disable(&randomize_kstack_offset);
848 return 0;
849 }
850 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
851 #endif
852
print_unknown_bootoptions(void)853 static void __init print_unknown_bootoptions(void)
854 {
855 char *unknown_options;
856 char *end;
857 const char *const *p;
858 size_t len;
859
860 if (panic_later || (!argv_init[1] && !envp_init[2]))
861 return;
862
863 /*
864 * Determine how many options we have to print out, plus a space
865 * before each
866 */
867 len = 1; /* null terminator */
868 for (p = &argv_init[1]; *p; p++) {
869 len++;
870 len += strlen(*p);
871 }
872 for (p = &envp_init[2]; *p; p++) {
873 len++;
874 len += strlen(*p);
875 }
876
877 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
878 if (!unknown_options) {
879 pr_err("%s: Failed to allocate %zu bytes\n",
880 __func__, len);
881 return;
882 }
883 end = unknown_options;
884
885 for (p = &argv_init[1]; *p; p++)
886 end += sprintf(end, " %s", *p);
887 for (p = &envp_init[2]; *p; p++)
888 end += sprintf(end, " %s", *p);
889
890 /* Start at unknown_options[1] to skip the initial space */
891 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
892 &unknown_options[1]);
893 memblock_free(unknown_options, len);
894 }
895
early_numa_node_init(void)896 static void __init early_numa_node_init(void)
897 {
898 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
899 #ifndef cpu_to_node
900 int cpu;
901
902 /* The early_cpu_to_node() should be ready here. */
903 for_each_possible_cpu(cpu)
904 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
905 #endif
906 #endif
907 }
908
909 #define KERNEL_CMDLINE_PREFIX "Kernel command line: "
910 #define KERNEL_CMDLINE_PREFIX_LEN (sizeof(KERNEL_CMDLINE_PREFIX) - 1)
911 #define KERNEL_CMDLINE_CONTINUATION " \\"
912 #define KERNEL_CMDLINE_CONTINUATION_LEN (sizeof(KERNEL_CMDLINE_CONTINUATION) - 1)
913
914 #define MIN_CMDLINE_LOG_WRAP_IDEAL_LEN (KERNEL_CMDLINE_PREFIX_LEN + \
915 KERNEL_CMDLINE_CONTINUATION_LEN)
916 #define CMDLINE_LOG_WRAP_IDEAL_LEN (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN > \
917 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN ? \
918 CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN : \
919 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN)
920
921 #define IDEAL_CMDLINE_LEN (CMDLINE_LOG_WRAP_IDEAL_LEN - KERNEL_CMDLINE_PREFIX_LEN)
922 #define IDEAL_CMDLINE_SPLIT_LEN (IDEAL_CMDLINE_LEN - KERNEL_CMDLINE_CONTINUATION_LEN)
923
924 /**
925 * print_kernel_cmdline() - Print the kernel cmdline with wrapping.
926 * @cmdline: The cmdline to print.
927 *
928 * Print the kernel command line, trying to wrap based on the Kconfig knob
929 * CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN.
930 *
931 * Wrapping is based on spaces, ignoring quotes. All lines are prefixed
932 * with "Kernel command line: " and lines that are not the last line have
933 * a " \" suffix added to them. The prefix and suffix count towards the
934 * line length for wrapping purposes. The ideal length will be exceeded
935 * if no appropriate place to wrap is found.
936 *
937 * Example output if CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN is 40:
938 * Kernel command line: loglevel=7 \
939 * Kernel command line: init=/sbin/init \
940 * Kernel command line: root=PARTUUID=8c3efc1a-768b-6642-8d0c-89eb782f19f0/PARTNROFF=1 \
941 * Kernel command line: rootwait ro \
942 * Kernel command line: my_quoted_arg="The \
943 * Kernel command line: quick brown fox \
944 * Kernel command line: jumps over the \
945 * Kernel command line: lazy dog."
946 */
print_kernel_cmdline(const char * cmdline)947 static void __init print_kernel_cmdline(const char *cmdline)
948 {
949 size_t len;
950
951 /* Config option of 0 or anything longer than the max disables wrapping */
952 if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN == 0 ||
953 IDEAL_CMDLINE_LEN >= COMMAND_LINE_SIZE - 1) {
954 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline);
955 return;
956 }
957
958 len = strlen(cmdline);
959 while (len > IDEAL_CMDLINE_LEN) {
960 const char *first_space;
961 const char *prev_cutoff;
962 const char *cutoff;
963 int to_print;
964 size_t used;
965
966 /* Find the last ' ' that wouldn't make the line too long */
967 prev_cutoff = NULL;
968 cutoff = cmdline;
969 while (true) {
970 cutoff = strchr(cutoff + 1, ' ');
971 if (!cutoff || cutoff - cmdline > IDEAL_CMDLINE_SPLIT_LEN)
972 break;
973 prev_cutoff = cutoff;
974 }
975 if (prev_cutoff)
976 cutoff = prev_cutoff;
977 else if (!cutoff)
978 break;
979
980 /* Find the beginning and end of the string of spaces */
981 first_space = cutoff;
982 while (first_space > cmdline && first_space[-1] == ' ')
983 first_space--;
984 to_print = first_space - cmdline;
985 while (*cutoff == ' ')
986 cutoff++;
987 used = cutoff - cmdline;
988
989 /* If the whole string is used, break and do the final printout */
990 if (len == used)
991 break;
992
993 if (to_print)
994 pr_notice("%s%.*s%s\n", KERNEL_CMDLINE_PREFIX,
995 to_print, cmdline, KERNEL_CMDLINE_CONTINUATION);
996
997 len -= used;
998 cmdline += used;
999 }
1000 if (len)
1001 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline);
1002 }
1003
1004 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)1005 void start_kernel(void)
1006 {
1007 char *command_line;
1008 char *after_dashes;
1009
1010 set_task_stack_end_magic(&init_task);
1011 smp_setup_processor_id();
1012 debug_objects_early_init();
1013 init_vmlinux_build_id();
1014
1015 cgroup_init_early();
1016
1017 local_irq_disable();
1018 early_boot_irqs_disabled = true;
1019
1020 /*
1021 * Interrupts are still disabled. Do necessary setups, then
1022 * enable them.
1023 */
1024 boot_cpu_init();
1025 page_address_init();
1026 pr_notice("%s", linux_banner);
1027 setup_arch(&command_line);
1028 /* Static keys and static calls are needed by LSMs */
1029 jump_label_init();
1030 static_call_init();
1031 early_security_init();
1032 setup_boot_config();
1033 setup_command_line(command_line);
1034 setup_nr_cpu_ids();
1035 setup_per_cpu_areas();
1036 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
1037 early_numa_node_init();
1038 boot_cpu_hotplug_init();
1039
1040 print_kernel_cmdline(saved_command_line);
1041 /* parameters may set static keys */
1042 parse_early_param();
1043 after_dashes = parse_args("Booting kernel",
1044 static_command_line, __start___param,
1045 __stop___param - __start___param,
1046 -1, -1, NULL, &unknown_bootoption);
1047 print_unknown_bootoptions();
1048 if (!IS_ERR_OR_NULL(after_dashes))
1049 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
1050 NULL, set_init_arg);
1051 if (extra_init_args)
1052 parse_args("Setting extra init args", extra_init_args,
1053 NULL, 0, -1, -1, NULL, set_init_arg);
1054
1055 /* Architectural and non-timekeeping rng init, before allocator init */
1056 random_init_early(command_line);
1057
1058 /*
1059 * These use large bootmem allocations and must precede
1060 * initalization of page allocator
1061 */
1062 setup_log_buf(0);
1063 vfs_caches_init_early();
1064 sort_main_extable();
1065 trap_init();
1066 mm_core_init();
1067 maple_tree_init();
1068 poking_init();
1069 ftrace_init();
1070
1071 /* trace_printk can be enabled here */
1072 early_trace_init();
1073
1074 /*
1075 * Set up the scheduler prior starting any interrupts (such as the
1076 * timer interrupt). Full topology setup happens at smp_init()
1077 * time - but meanwhile we still have a functioning scheduler.
1078 */
1079 sched_init();
1080
1081 if (WARN(!irqs_disabled(),
1082 "Interrupts were enabled *very* early, fixing it\n"))
1083 local_irq_disable();
1084 radix_tree_init();
1085
1086 /*
1087 * Set up housekeeping before setting up workqueues to allow the unbound
1088 * workqueue to take non-housekeeping into account.
1089 */
1090 housekeeping_init();
1091
1092 /*
1093 * Allow workqueue creation and work item queueing/cancelling
1094 * early. Work item execution depends on kthreads and starts after
1095 * workqueue_init().
1096 */
1097 workqueue_init_early();
1098
1099 rcu_init();
1100 kvfree_rcu_init();
1101
1102 /* Trace events are available after this */
1103 trace_init();
1104
1105 if (initcall_debug)
1106 initcall_debug_enable();
1107
1108 context_tracking_init();
1109 /* init some links before init_ISA_irqs() */
1110 early_irq_init();
1111 init_IRQ();
1112 tick_init();
1113 rcu_init_nohz();
1114 timers_init();
1115 srcu_init();
1116 hrtimers_init();
1117 softirq_init();
1118 timekeeping_init();
1119 time_init();
1120
1121 /* This must be after timekeeping is initialized */
1122 random_init();
1123
1124 /* These make use of the fully initialized rng */
1125 kfence_init();
1126 boot_init_stack_canary();
1127
1128 perf_event_init();
1129 profile_init();
1130 call_function_init();
1131 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1132
1133 early_boot_irqs_disabled = false;
1134 local_irq_enable();
1135
1136 kmem_cache_init_late();
1137
1138 /*
1139 * HACK ALERT! This is early. We're enabling the console before
1140 * we've done PCI setups etc, and console_init() must be aware of
1141 * this. But we do want output early, in case something goes wrong.
1142 */
1143 console_init();
1144 if (panic_later)
1145 panic("Too many boot %s vars at `%s'", panic_later,
1146 panic_param);
1147
1148 lockdep_init();
1149
1150 /*
1151 * Need to run this when irqs are enabled, because it wants
1152 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1153 * too:
1154 */
1155 locking_selftest();
1156
1157 #ifdef CONFIG_BLK_DEV_INITRD
1158 if (initrd_start && !initrd_below_start_ok &&
1159 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1160 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1161 page_to_pfn(virt_to_page((void *)initrd_start)),
1162 min_low_pfn);
1163 initrd_start = 0;
1164 }
1165 #endif
1166 setup_per_cpu_pageset();
1167 numa_policy_init();
1168 acpi_early_init();
1169 if (late_time_init)
1170 late_time_init();
1171 sched_clock_init();
1172 calibrate_delay();
1173
1174 arch_cpu_finalize_init();
1175
1176 pid_idr_init();
1177 anon_vma_init();
1178 thread_stack_cache_init();
1179 cred_init();
1180 fork_init();
1181 proc_caches_init();
1182 uts_ns_init();
1183 time_ns_init();
1184 key_init();
1185 security_init();
1186 dbg_late_init();
1187 net_ns_init();
1188 vfs_caches_init();
1189 pagecache_init();
1190 signals_init();
1191 seq_file_init();
1192 proc_root_init();
1193 nsfs_init();
1194 pidfs_init();
1195 cpuset_init();
1196 mem_cgroup_init();
1197 cgroup_init();
1198 taskstats_init_early();
1199 delayacct_init();
1200
1201 acpi_subsystem_init();
1202 arch_post_acpi_subsys_init();
1203 kcsan_init();
1204
1205 /* Do the rest non-__init'ed, we're now alive */
1206 rest_init();
1207
1208 /*
1209 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1210 * and older.
1211 */
1212 #if !__has_attribute(__no_stack_protector__)
1213 prevent_tail_call_optimization();
1214 #endif
1215 }
1216
1217 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1218 static void __init do_ctors(void)
1219 {
1220 /*
1221 * For UML, the constructors have already been called by the
1222 * normal setup code as it's just a normal ELF binary, so we
1223 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1224 * even on UML for modules.
1225 */
1226 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1227 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1228
1229 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1230 (*fn)();
1231 #endif
1232 }
1233
1234 #ifdef CONFIG_KALLSYMS
1235 struct blacklist_entry {
1236 struct list_head next;
1237 char *buf;
1238 };
1239
1240 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1241
initcall_blacklist(char * str)1242 static int __init initcall_blacklist(char *str)
1243 {
1244 char *str_entry;
1245 struct blacklist_entry *entry;
1246
1247 /* str argument is a comma-separated list of functions */
1248 do {
1249 str_entry = strsep(&str, ",");
1250 if (str_entry) {
1251 pr_debug("blacklisting initcall %s\n", str_entry);
1252 entry = memblock_alloc_or_panic(sizeof(*entry),
1253 SMP_CACHE_BYTES);
1254 entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1,
1255 SMP_CACHE_BYTES);
1256 strcpy(entry->buf, str_entry);
1257 list_add(&entry->next, &blacklisted_initcalls);
1258 }
1259 } while (str_entry);
1260
1261 return 1;
1262 }
1263
initcall_blacklisted(initcall_t fn)1264 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1265 {
1266 struct blacklist_entry *entry;
1267 char fn_name[KSYM_SYMBOL_LEN];
1268 unsigned long addr;
1269
1270 if (list_empty(&blacklisted_initcalls))
1271 return false;
1272
1273 addr = (unsigned long) dereference_function_descriptor(fn);
1274 sprint_symbol_no_offset(fn_name, addr);
1275
1276 /*
1277 * fn will be "function_name [module_name]" where [module_name] is not
1278 * displayed for built-in init functions. Strip off the [module_name].
1279 */
1280 strreplace(fn_name, ' ', '\0');
1281
1282 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1283 if (!strcmp(fn_name, entry->buf)) {
1284 pr_debug("initcall %s blacklisted\n", fn_name);
1285 return true;
1286 }
1287 }
1288
1289 return false;
1290 }
1291 #else
initcall_blacklist(char * str)1292 static int __init initcall_blacklist(char *str)
1293 {
1294 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1295 return 0;
1296 }
1297
initcall_blacklisted(initcall_t fn)1298 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1299 {
1300 return false;
1301 }
1302 #endif
1303 __setup("initcall_blacklist=", initcall_blacklist);
1304
1305 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1306 trace_initcall_start_cb(void *data, initcall_t fn)
1307 {
1308 ktime_t *calltime = data;
1309
1310 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1311 *calltime = ktime_get();
1312 }
1313
1314 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1315 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1316 {
1317 ktime_t rettime, *calltime = data;
1318
1319 rettime = ktime_get();
1320 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1321 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1322 }
1323
1324 static __init_or_module void
trace_initcall_level_cb(void * data,const char * level)1325 trace_initcall_level_cb(void *data, const char *level)
1326 {
1327 printk(KERN_DEBUG "entering initcall level: %s\n", level);
1328 }
1329
1330 static ktime_t initcall_calltime;
1331
1332 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1333 static void __init initcall_debug_enable(void)
1334 {
1335 int ret;
1336
1337 ret = register_trace_initcall_start(trace_initcall_start_cb,
1338 &initcall_calltime);
1339 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1340 &initcall_calltime);
1341 ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL);
1342 WARN(ret, "Failed to register initcall tracepoints\n");
1343 }
1344 # define do_trace_initcall_start trace_initcall_start
1345 # define do_trace_initcall_finish trace_initcall_finish
1346 # define do_trace_initcall_level trace_initcall_level
1347 #else
do_trace_initcall_start(initcall_t fn)1348 static inline void do_trace_initcall_start(initcall_t fn)
1349 {
1350 if (!initcall_debug)
1351 return;
1352 trace_initcall_start_cb(&initcall_calltime, fn);
1353 }
do_trace_initcall_finish(initcall_t fn,int ret)1354 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1355 {
1356 if (!initcall_debug)
1357 return;
1358 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1359 }
do_trace_initcall_level(const char * level)1360 static inline void do_trace_initcall_level(const char *level)
1361 {
1362 if (!initcall_debug)
1363 return;
1364 trace_initcall_level_cb(NULL, level);
1365 }
1366 #endif /* !TRACEPOINTS_ENABLED */
1367
do_one_initcall(initcall_t fn)1368 int __init_or_module do_one_initcall(initcall_t fn)
1369 {
1370 int count = preempt_count();
1371 char msgbuf[64];
1372 int ret;
1373
1374 if (initcall_blacklisted(fn))
1375 return -EPERM;
1376
1377 do_trace_initcall_start(fn);
1378 ret = fn();
1379 do_trace_initcall_finish(fn, ret);
1380
1381 msgbuf[0] = 0;
1382
1383 if (preempt_count() != count) {
1384 sprintf(msgbuf, "preemption imbalance ");
1385 preempt_count_set(count);
1386 }
1387 if (irqs_disabled()) {
1388 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1389 local_irq_enable();
1390 }
1391 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1392
1393 add_latent_entropy();
1394 return ret;
1395 }
1396
1397
1398 static initcall_entry_t *initcall_levels[] __initdata = {
1399 __initcall0_start,
1400 __initcall1_start,
1401 __initcall2_start,
1402 __initcall3_start,
1403 __initcall4_start,
1404 __initcall5_start,
1405 __initcall6_start,
1406 __initcall7_start,
1407 __initcall_end,
1408 };
1409
1410 /* Keep these in sync with initcalls in include/linux/init.h */
1411 static const char *initcall_level_names[] __initdata = {
1412 "pure",
1413 "core",
1414 "postcore",
1415 "arch",
1416 "subsys",
1417 "fs",
1418 "device",
1419 "late",
1420 };
1421
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1422 static int __init ignore_unknown_bootoption(char *param, char *val,
1423 const char *unused, void *arg)
1424 {
1425 return 0;
1426 }
1427
do_initcall_level(int level,char * command_line)1428 static void __init do_initcall_level(int level, char *command_line)
1429 {
1430 initcall_entry_t *fn;
1431
1432 parse_args(initcall_level_names[level],
1433 command_line, __start___param,
1434 __stop___param - __start___param,
1435 level, level,
1436 NULL, ignore_unknown_bootoption);
1437
1438 do_trace_initcall_level(initcall_level_names[level]);
1439 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1440 do_one_initcall(initcall_from_entry(fn));
1441 }
1442
do_initcalls(void)1443 static void __init do_initcalls(void)
1444 {
1445 int level;
1446 size_t len = saved_command_line_len + 1;
1447 char *command_line;
1448
1449 command_line = kzalloc(len, GFP_KERNEL);
1450 if (!command_line)
1451 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1452
1453 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1454 /* Parser modifies command_line, restore it each time */
1455 strcpy(command_line, saved_command_line);
1456 do_initcall_level(level, command_line);
1457 }
1458
1459 kfree(command_line);
1460 }
1461
1462 /*
1463 * Ok, the machine is now initialized. None of the devices
1464 * have been touched yet, but the CPU subsystem is up and
1465 * running, and memory and process management works.
1466 *
1467 * Now we can finally start doing some real work..
1468 */
do_basic_setup(void)1469 static void __init do_basic_setup(void)
1470 {
1471 cpuset_init_smp();
1472 driver_init();
1473 init_irq_proc();
1474 do_ctors();
1475 do_initcalls();
1476 }
1477
do_pre_smp_initcalls(void)1478 static void __init do_pre_smp_initcalls(void)
1479 {
1480 initcall_entry_t *fn;
1481
1482 do_trace_initcall_level("early");
1483 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1484 do_one_initcall(initcall_from_entry(fn));
1485 }
1486
run_init_process(const char * init_filename)1487 static int run_init_process(const char *init_filename)
1488 {
1489 const char *const *p;
1490
1491 argv_init[0] = init_filename;
1492 pr_info("Run %s as init process\n", init_filename);
1493 pr_debug(" with arguments:\n");
1494 for (p = argv_init; *p; p++)
1495 pr_debug(" %s\n", *p);
1496 pr_debug(" with environment:\n");
1497 for (p = envp_init; *p; p++)
1498 pr_debug(" %s\n", *p);
1499 return kernel_execve(init_filename, argv_init, envp_init);
1500 }
1501
try_to_run_init_process(const char * init_filename)1502 static int try_to_run_init_process(const char *init_filename)
1503 {
1504 int ret;
1505
1506 ret = run_init_process(init_filename);
1507
1508 if (ret && ret != -ENOENT) {
1509 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1510 init_filename, ret);
1511 }
1512
1513 return ret;
1514 }
1515
1516 static noinline void __init kernel_init_freeable(void);
1517
1518 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1519 bool rodata_enabled __ro_after_init = true;
1520
1521 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1522 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1523 #endif
1524
set_debug_rodata(char * str)1525 static int __init set_debug_rodata(char *str)
1526 {
1527 if (arch_parse_debug_rodata(str))
1528 return 0;
1529
1530 if (str && !strcmp(str, "on"))
1531 rodata_enabled = true;
1532 else if (str && !strcmp(str, "off"))
1533 rodata_enabled = false;
1534 else
1535 pr_warn("Invalid option string for rodata: '%s'\n", str);
1536 return 0;
1537 }
1538 early_param("rodata", set_debug_rodata);
1539 #endif
1540
mark_readonly(void)1541 static void mark_readonly(void)
1542 {
1543 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1544 /*
1545 * load_module() results in W+X mappings, which are cleaned
1546 * up with init_free_wq. Let's make sure that queued work is
1547 * flushed so that we don't hit false positives looking for
1548 * insecure pages which are W+X.
1549 */
1550 flush_module_init_free_work();
1551 jump_label_init_ro();
1552 mark_rodata_ro();
1553 debug_checkwx();
1554 rodata_test();
1555 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1556 pr_info("Kernel memory protection disabled.\n");
1557 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1558 pr_warn("Kernel memory protection not selected by kernel config.\n");
1559 } else {
1560 pr_warn("This architecture does not have kernel memory protection.\n");
1561 }
1562 }
1563
free_initmem(void)1564 void __weak free_initmem(void)
1565 {
1566 free_initmem_default(POISON_FREE_INITMEM);
1567 }
1568
kernel_init(void * unused)1569 static int __ref kernel_init(void *unused)
1570 {
1571 int ret;
1572
1573 /*
1574 * Wait until kthreadd is all set-up.
1575 */
1576 wait_for_completion(&kthreadd_done);
1577
1578 kernel_init_freeable();
1579 /* need to finish all async __init code before freeing the memory */
1580 async_synchronize_full();
1581
1582 system_state = SYSTEM_FREEING_INITMEM;
1583 kprobe_free_init_mem();
1584 ftrace_free_init_mem();
1585 kgdb_free_init_mem();
1586 exit_boot_config();
1587 free_initmem();
1588 mark_readonly();
1589
1590 /*
1591 * Kernel mappings are now finalized - update the userspace page-table
1592 * to finalize PTI.
1593 */
1594 pti_finalize();
1595
1596 system_state = SYSTEM_RUNNING;
1597 numa_default_policy();
1598
1599 rcu_end_inkernel_boot();
1600
1601 do_sysctl_args();
1602
1603 if (ramdisk_execute_command) {
1604 ret = run_init_process(ramdisk_execute_command);
1605 if (!ret)
1606 return 0;
1607 pr_err("Failed to execute %s (error %d)\n",
1608 ramdisk_execute_command, ret);
1609 }
1610
1611 /*
1612 * We try each of these until one succeeds.
1613 *
1614 * The Bourne shell can be used instead of init if we are
1615 * trying to recover a really broken machine.
1616 */
1617 if (execute_command) {
1618 ret = run_init_process(execute_command);
1619 if (!ret)
1620 return 0;
1621 panic("Requested init %s failed (error %d).",
1622 execute_command, ret);
1623 }
1624
1625 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1626 ret = run_init_process(CONFIG_DEFAULT_INIT);
1627 if (ret)
1628 pr_err("Default init %s failed (error %d)\n",
1629 CONFIG_DEFAULT_INIT, ret);
1630 else
1631 return 0;
1632 }
1633
1634 if (!try_to_run_init_process("/sbin/init") ||
1635 !try_to_run_init_process("/etc/init") ||
1636 !try_to_run_init_process("/bin/init") ||
1637 !try_to_run_init_process("/bin/sh"))
1638 return 0;
1639
1640 panic("No working init found. Try passing init= option to kernel. "
1641 "See Linux Documentation/admin-guide/init.rst for guidance.");
1642 }
1643
1644 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1645 void __init console_on_rootfs(void)
1646 {
1647 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1648
1649 if (IS_ERR(file)) {
1650 pr_err("Warning: unable to open an initial console.\n");
1651 return;
1652 }
1653 init_dup(file);
1654 init_dup(file);
1655 init_dup(file);
1656 fput(file);
1657 }
1658
kernel_init_freeable(void)1659 static noinline void __init kernel_init_freeable(void)
1660 {
1661 /* Now the scheduler is fully set up and can do blocking allocations */
1662 gfp_allowed_mask = __GFP_BITS_MASK;
1663
1664 /*
1665 * init can allocate pages on any node
1666 */
1667 set_mems_allowed(node_states[N_MEMORY]);
1668
1669 cad_pid = get_pid(task_pid(current));
1670
1671 smp_prepare_cpus(setup_max_cpus);
1672
1673 workqueue_init();
1674
1675 init_mm_internals();
1676
1677 do_pre_smp_initcalls();
1678 lockup_detector_init();
1679
1680 smp_init();
1681 sched_init_smp();
1682
1683 workqueue_init_topology();
1684 async_init();
1685 padata_init();
1686 page_alloc_init_late();
1687
1688 do_basic_setup();
1689
1690 kunit_run_all_tests();
1691
1692 wait_for_initramfs();
1693 console_on_rootfs();
1694
1695 /*
1696 * check if there is an early userspace init. If yes, let it do all
1697 * the work
1698 */
1699 int ramdisk_command_access;
1700 ramdisk_command_access = init_eaccess(ramdisk_execute_command);
1701 if (ramdisk_command_access != 0) {
1702 pr_warn("check access for rdinit=%s failed: %i, ignoring\n",
1703 ramdisk_execute_command, ramdisk_command_access);
1704 ramdisk_execute_command = NULL;
1705 prepare_namespace();
1706 }
1707
1708 /*
1709 * Ok, we have completed the initial bootup, and
1710 * we're essentially up and running. Get rid of the
1711 * initmem segments and start the user-mode stuff..
1712 *
1713 * rootfs is available now, try loading the public keys
1714 * and default modules
1715 */
1716
1717 integrity_load_keys();
1718 }
1719