1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/init/main.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
11 */
12
13 #define DEBUG /* Enable initcall_debug */
14
15 #include <linux/types.h>
16 #include <linux/extable.h>
17 #include <linux/module.h>
18 #include <linux/proc_fs.h>
19 #include <linux/binfmts.h>
20 #include <linux/kernel.h>
21 #include <linux/syscalls.h>
22 #include <linux/stackprotector.h>
23 #include <linux/string.h>
24 #include <linux/ctype.h>
25 #include <linux/delay.h>
26 #include <linux/ioport.h>
27 #include <linux/init.h>
28 #include <linux/initrd.h>
29 #include <linux/memblock.h>
30 #include <linux/acpi.h>
31 #include <linux/bootconfig.h>
32 #include <linux/console.h>
33 #include <linux/nmi.h>
34 #include <linux/percpu.h>
35 #include <linux/kmod.h>
36 #include <linux/kprobes.h>
37 #include <linux/kmsan.h>
38 #include <linux/vmalloc.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/start_kernel.h>
41 #include <linux/security.h>
42 #include <linux/smp.h>
43 #include <linux/profile.h>
44 #include <linux/kfence.h>
45 #include <linux/rcupdate.h>
46 #include <linux/srcu.h>
47 #include <linux/moduleparam.h>
48 #include <linux/kallsyms.h>
49 #include <linux/buildid.h>
50 #include <linux/writeback.h>
51 #include <linux/cpu.h>
52 #include <linux/cpuset.h>
53 #include <linux/memcontrol.h>
54 #include <linux/cgroup.h>
55 #include <linux/efi.h>
56 #include <linux/tick.h>
57 #include <linux/sched/isolation.h>
58 #include <linux/interrupt.h>
59 #include <linux/taskstats_kern.h>
60 #include <linux/delayacct.h>
61 #include <linux/unistd.h>
62 #include <linux/utsname.h>
63 #include <linux/rmap.h>
64 #include <linux/mempolicy.h>
65 #include <linux/key.h>
66 #include <linux/debug_locks.h>
67 #include <linux/debugobjects.h>
68 #include <linux/lockdep.h>
69 #include <linux/kmemleak.h>
70 #include <linux/padata.h>
71 #include <linux/pid_namespace.h>
72 #include <linux/device/driver.h>
73 #include <linux/kthread.h>
74 #include <linux/sched.h>
75 #include <linux/sched/init.h>
76 #include <linux/signal.h>
77 #include <linux/idr.h>
78 #include <linux/kgdb.h>
79 #include <linux/ftrace.h>
80 #include <linux/async.h>
81 #include <linux/shmem_fs.h>
82 #include <linux/slab.h>
83 #include <linux/perf_event.h>
84 #include <linux/ptrace.h>
85 #include <linux/pti.h>
86 #include <linux/blkdev.h>
87 #include <linux/sched/clock.h>
88 #include <linux/sched/task.h>
89 #include <linux/sched/task_stack.h>
90 #include <linux/context_tracking.h>
91 #include <linux/random.h>
92 #include <linux/moduleloader.h>
93 #include <linux/list.h>
94 #include <linux/integrity.h>
95 #include <linux/proc_ns.h>
96 #include <linux/io.h>
97 #include <linux/cache.h>
98 #include <linux/rodata_test.h>
99 #include <linux/jump_label.h>
100 #include <linux/kcsan.h>
101 #include <linux/init_syscalls.h>
102 #include <linux/stackdepot.h>
103 #include <linux/randomize_kstack.h>
104 #include <linux/pidfs.h>
105 #include <linux/ptdump.h>
106 #include <net/net_namespace.h>
107
108 #include <asm/io.h>
109 #include <asm/setup.h>
110 #include <asm/sections.h>
111 #include <asm/cacheflush.h>
112
113 #define CREATE_TRACE_POINTS
114 #include <trace/events/initcall.h>
115
116 #include <kunit/test.h>
117
118 static int kernel_init(void *);
119
120 /*
121 * Debug helper: via this flag we know that we are in 'early bootup code'
122 * where only the boot processor is running with IRQ disabled. This means
123 * two things - IRQ must not be enabled before the flag is cleared and some
124 * operations which are not allowed with IRQ disabled are allowed while the
125 * flag is set.
126 */
127 bool early_boot_irqs_disabled __read_mostly;
128
129 enum system_states system_state __read_mostly;
130 EXPORT_SYMBOL(system_state);
131
132 /*
133 * Boot command-line arguments
134 */
135 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
136 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
137
138 /* Default late time init is NULL. archs can override this later. */
139 void (*__initdata late_time_init)(void);
140
141 /* Untouched command line saved by arch-specific code. */
142 char __initdata boot_command_line[COMMAND_LINE_SIZE];
143 /* Untouched saved command line (eg. for /proc) */
144 char *saved_command_line __ro_after_init;
145 unsigned int saved_command_line_len __ro_after_init;
146 /* Command line for parameter parsing */
147 static char *static_command_line;
148 /* Untouched extra command line */
149 static char *extra_command_line;
150 /* Extra init arguments */
151 static char *extra_init_args;
152
153 #ifdef CONFIG_BOOT_CONFIG
154 /* Is bootconfig on command line? */
155 static bool bootconfig_found;
156 static size_t initargs_offs;
157 #else
158 # define bootconfig_found false
159 # define initargs_offs 0
160 #endif
161
162 static char *execute_command;
163 static char *ramdisk_execute_command = "/init";
164
165 /*
166 * Used to generate warnings if static_key manipulation functions are used
167 * before jump_label_init is called.
168 */
169 bool static_key_initialized __read_mostly;
170 EXPORT_SYMBOL_GPL(static_key_initialized);
171
172 /*
173 * If set, this is an indication to the drivers that reset the underlying
174 * device before going ahead with the initialization otherwise driver might
175 * rely on the BIOS and skip the reset operation.
176 *
177 * This is useful if kernel is booting in an unreliable environment.
178 * For ex. kdump situation where previous kernel has crashed, BIOS has been
179 * skipped and devices will be in unknown state.
180 */
181 unsigned int reset_devices;
182 EXPORT_SYMBOL(reset_devices);
183
set_reset_devices(char * str)184 static int __init set_reset_devices(char *str)
185 {
186 reset_devices = 1;
187 return 1;
188 }
189
190 __setup("reset_devices", set_reset_devices);
191
192 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
193 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
194 static const char *panic_later, *panic_param;
195
obsolete_checksetup(char * line)196 static bool __init obsolete_checksetup(char *line)
197 {
198 const struct obs_kernel_param *p;
199 bool had_early_param = false;
200
201 p = __setup_start;
202 do {
203 int n = strlen(p->str);
204 if (parameqn(line, p->str, n)) {
205 if (p->early) {
206 /* Already done in parse_early_param?
207 * (Needs exact match on param part).
208 * Keep iterating, as we can have early
209 * params and __setups of same names 8( */
210 if (line[n] == '\0' || line[n] == '=')
211 had_early_param = true;
212 } else if (!p->setup_func) {
213 pr_warn("Parameter %s is obsolete, ignored\n",
214 p->str);
215 return true;
216 } else if (p->setup_func(line + n))
217 return true;
218 }
219 p++;
220 } while (p < __setup_end);
221
222 return had_early_param;
223 }
224
225 /*
226 * This should be approx 2 Bo*oMips to start (note initial shift), and will
227 * still work even if initially too large, it will just take slightly longer
228 */
229 unsigned long loops_per_jiffy = (1<<12);
230 EXPORT_SYMBOL(loops_per_jiffy);
231
debug_kernel(char * str)232 static int __init debug_kernel(char *str)
233 {
234 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
235 return 0;
236 }
237
quiet_kernel(char * str)238 static int __init quiet_kernel(char *str)
239 {
240 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
241 return 0;
242 }
243
244 early_param("debug", debug_kernel);
245 early_param("quiet", quiet_kernel);
246
loglevel(char * str)247 static int __init loglevel(char *str)
248 {
249 int newlevel;
250
251 /*
252 * Only update loglevel value when a correct setting was passed,
253 * to prevent blind crashes (when loglevel being set to 0) that
254 * are quite hard to debug
255 */
256 if (get_option(&str, &newlevel)) {
257 console_loglevel = newlevel;
258 return 0;
259 }
260
261 return -EINVAL;
262 }
263
264 early_param("loglevel", loglevel);
265
266 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(size_t * _size)267 static void * __init get_boot_config_from_initrd(size_t *_size)
268 {
269 u32 size, csum;
270 char *data;
271 u32 *hdr;
272 int i;
273
274 if (!initrd_end)
275 return NULL;
276
277 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
278 /*
279 * Since Grub may align the size of initrd to 4, we must
280 * check the preceding 3 bytes as well.
281 */
282 for (i = 0; i < 4; i++) {
283 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
284 goto found;
285 data--;
286 }
287 return NULL;
288
289 found:
290 hdr = (u32 *)(data - 8);
291 size = le32_to_cpu(hdr[0]);
292 csum = le32_to_cpu(hdr[1]);
293
294 data = ((void *)hdr) - size;
295 if ((unsigned long)data < initrd_start) {
296 pr_err("bootconfig size %d is greater than initrd size %ld\n",
297 size, initrd_end - initrd_start);
298 return NULL;
299 }
300
301 if (xbc_calc_checksum(data, size) != csum) {
302 pr_err("bootconfig checksum failed\n");
303 return NULL;
304 }
305
306 /* Remove bootconfig from initramfs/initrd */
307 initrd_end = (unsigned long)data;
308 if (_size)
309 *_size = size;
310
311 return data;
312 }
313 #else
get_boot_config_from_initrd(size_t * _size)314 static void * __init get_boot_config_from_initrd(size_t *_size)
315 {
316 return NULL;
317 }
318 #endif
319
320 #ifdef CONFIG_BOOT_CONFIG
321
322 static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
323
324 #define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
325
xbc_snprint_cmdline(char * buf,size_t size,struct xbc_node * root)326 static int __init xbc_snprint_cmdline(char *buf, size_t size,
327 struct xbc_node *root)
328 {
329 struct xbc_node *knode, *vnode;
330 char *end = buf + size;
331 const char *val, *q;
332 int ret;
333
334 xbc_node_for_each_key_value(root, knode, val) {
335 ret = xbc_node_compose_key_after(root, knode,
336 xbc_namebuf, XBC_KEYLEN_MAX);
337 if (ret < 0)
338 return ret;
339
340 vnode = xbc_node_get_child(knode);
341 if (!vnode) {
342 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
343 if (ret < 0)
344 return ret;
345 buf += ret;
346 continue;
347 }
348 xbc_array_for_each_value(vnode, val) {
349 /*
350 * For prettier and more readable /proc/cmdline, only
351 * quote the value when necessary, i.e. when it contains
352 * whitespace.
353 */
354 q = strpbrk(val, " \t\r\n") ? "\"" : "";
355 ret = snprintf(buf, rest(buf, end), "%s=%s%s%s ",
356 xbc_namebuf, q, val, q);
357 if (ret < 0)
358 return ret;
359 buf += ret;
360 }
361 }
362
363 return buf - (end - size);
364 }
365 #undef rest
366
367 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)368 static char * __init xbc_make_cmdline(const char *key)
369 {
370 struct xbc_node *root;
371 char *new_cmdline;
372 int ret, len = 0;
373
374 root = xbc_find_node(key);
375 if (!root)
376 return NULL;
377
378 /* Count required buffer size */
379 len = xbc_snprint_cmdline(NULL, 0, root);
380 if (len <= 0)
381 return NULL;
382
383 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
384 if (!new_cmdline) {
385 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
386 return NULL;
387 }
388
389 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
390 if (ret < 0 || ret > len) {
391 pr_err("Failed to print extra kernel cmdline.\n");
392 memblock_free(new_cmdline, len + 1);
393 return NULL;
394 }
395
396 return new_cmdline;
397 }
398
bootconfig_params(char * param,char * val,const char * unused,void * arg)399 static int __init bootconfig_params(char *param, char *val,
400 const char *unused, void *arg)
401 {
402 if (strcmp(param, "bootconfig") == 0) {
403 bootconfig_found = true;
404 }
405 return 0;
406 }
407
warn_bootconfig(char * str)408 static int __init warn_bootconfig(char *str)
409 {
410 /* The 'bootconfig' has been handled by bootconfig_params(). */
411 return 0;
412 }
413
setup_boot_config(void)414 static void __init setup_boot_config(void)
415 {
416 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
417 const char *msg, *data;
418 int pos, ret;
419 size_t size;
420 char *err;
421
422 /* Cut out the bootconfig data even if we have no bootconfig option */
423 data = get_boot_config_from_initrd(&size);
424 /* If there is no bootconfig in initrd, try embedded one. */
425 if (!data)
426 data = xbc_get_embedded_bootconfig(&size);
427
428 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
429 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
430 bootconfig_params);
431
432 if (IS_ERR(err) || !(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
433 return;
434
435 /* parse_args() stops at the next param of '--' and returns an address */
436 if (err)
437 initargs_offs = err - tmp_cmdline;
438
439 if (!data) {
440 /* If user intended to use bootconfig, show an error level message */
441 if (bootconfig_found)
442 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
443 else
444 pr_info("No bootconfig data provided, so skipping bootconfig");
445 return;
446 }
447
448 if (size >= XBC_DATA_MAX) {
449 pr_err("bootconfig size %ld greater than max size %d\n",
450 (long)size, XBC_DATA_MAX);
451 return;
452 }
453
454 ret = xbc_init(data, size, &msg, &pos);
455 if (ret < 0) {
456 if (pos < 0)
457 pr_err("Failed to init bootconfig: %s.\n", msg);
458 else
459 pr_err("Failed to parse bootconfig: %s at %d.\n",
460 msg, pos);
461 } else {
462 xbc_get_info(&ret, NULL);
463 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
464 /* keys starting with "kernel." are passed via cmdline */
465 extra_command_line = xbc_make_cmdline("kernel");
466 /* Also, "init." keys are init arguments */
467 extra_init_args = xbc_make_cmdline("init");
468 }
469 return;
470 }
471
exit_boot_config(void)472 static void __init exit_boot_config(void)
473 {
474 xbc_exit();
475 }
476
477 #else /* !CONFIG_BOOT_CONFIG */
478
setup_boot_config(void)479 static void __init setup_boot_config(void)
480 {
481 /* Remove bootconfig data from initrd */
482 get_boot_config_from_initrd(NULL);
483 }
484
warn_bootconfig(char * str)485 static int __init warn_bootconfig(char *str)
486 {
487 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
488 return 0;
489 }
490
491 #define exit_boot_config() do {} while (0)
492
493 #endif /* CONFIG_BOOT_CONFIG */
494
495 early_param("bootconfig", warn_bootconfig);
496
cmdline_has_extra_options(void)497 bool __init cmdline_has_extra_options(void)
498 {
499 return extra_command_line || extra_init_args;
500 }
501
502 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)503 static void __init repair_env_string(char *param, char *val)
504 {
505 if (val) {
506 /* param=val or param="val"? */
507 if (val == param+strlen(param)+1)
508 val[-1] = '=';
509 else if (val == param+strlen(param)+2) {
510 val[-2] = '=';
511 memmove(val-1, val, strlen(val)+1);
512 } else
513 BUG();
514 }
515 }
516
517 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)518 static int __init set_init_arg(char *param, char *val,
519 const char *unused, void *arg)
520 {
521 unsigned int i;
522
523 if (panic_later)
524 return 0;
525
526 repair_env_string(param, val);
527
528 for (i = 0; argv_init[i]; i++) {
529 if (i == MAX_INIT_ARGS) {
530 panic_later = "init";
531 panic_param = param;
532 return 0;
533 }
534 }
535 argv_init[i] = param;
536 return 0;
537 }
538
539 /*
540 * Unknown boot options get handed to init, unless they look like
541 * unused parameters (modprobe will find them in /proc/cmdline).
542 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)543 static int __init unknown_bootoption(char *param, char *val,
544 const char *unused, void *arg)
545 {
546 size_t len = strlen(param);
547
548 /* Handle params aliased to sysctls */
549 if (sysctl_is_alias(param))
550 return 0;
551
552 repair_env_string(param, val);
553
554 /* Handle obsolete-style parameters */
555 if (obsolete_checksetup(param))
556 return 0;
557
558 /* Unused module parameter. */
559 if (strnchr(param, len, '.'))
560 return 0;
561
562 if (panic_later)
563 return 0;
564
565 if (val) {
566 /* Environment option */
567 unsigned int i;
568 for (i = 0; envp_init[i]; i++) {
569 if (i == MAX_INIT_ENVS) {
570 panic_later = "env";
571 panic_param = param;
572 }
573 if (!strncmp(param, envp_init[i], len+1))
574 break;
575 }
576 envp_init[i] = param;
577 } else {
578 /* Command line option */
579 unsigned int i;
580 for (i = 0; argv_init[i]; i++) {
581 if (i == MAX_INIT_ARGS) {
582 panic_later = "init";
583 panic_param = param;
584 }
585 }
586 argv_init[i] = param;
587 }
588 return 0;
589 }
590
init_setup(char * str)591 static int __init init_setup(char *str)
592 {
593 unsigned int i;
594
595 execute_command = str;
596 /*
597 * In case LILO is going to boot us with default command line,
598 * it prepends "auto" before the whole cmdline which makes
599 * the shell think it should execute a script with such name.
600 * So we ignore all arguments entered _before_ init=... [MJ]
601 */
602 for (i = 1; i < MAX_INIT_ARGS; i++)
603 argv_init[i] = NULL;
604 return 1;
605 }
606 __setup("init=", init_setup);
607
rdinit_setup(char * str)608 static int __init rdinit_setup(char *str)
609 {
610 unsigned int i;
611
612 ramdisk_execute_command = str;
613 /* See "auto" comment in init_setup */
614 for (i = 1; i < MAX_INIT_ARGS; i++)
615 argv_init[i] = NULL;
616 return 1;
617 }
618 __setup("rdinit=", rdinit_setup);
619
620 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)621 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)622 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
623 #endif
624
625 /*
626 * We need to store the untouched command line for future reference.
627 * We also need to store the touched command line since the parameter
628 * parsing is performed in place, and we should allow a component to
629 * store reference of name/value for future reference.
630 */
setup_command_line(char * command_line)631 static void __init setup_command_line(char *command_line)
632 {
633 size_t len, xlen = 0, ilen = 0;
634
635 if (extra_command_line)
636 xlen = strlen(extra_command_line);
637 if (extra_init_args) {
638 extra_init_args = strim(extra_init_args); /* remove trailing space */
639 ilen = strlen(extra_init_args) + 4; /* for " -- " */
640 }
641
642 len = xlen + strlen(boot_command_line) + ilen + 1;
643
644 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
645 if (!saved_command_line)
646 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
647
648 len = xlen + strlen(command_line) + 1;
649
650 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
651 if (!static_command_line)
652 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
653
654 if (xlen) {
655 /*
656 * We have to put extra_command_line before boot command
657 * lines because there could be dashes (separator of init
658 * command line) in the command lines.
659 */
660 strcpy(saved_command_line, extra_command_line);
661 strcpy(static_command_line, extra_command_line);
662 }
663 strcpy(saved_command_line + xlen, boot_command_line);
664 strcpy(static_command_line + xlen, command_line);
665
666 if (ilen) {
667 /*
668 * Append supplemental init boot args to saved_command_line
669 * so that user can check what command line options passed
670 * to init.
671 * The order should always be
672 * " -- "[bootconfig init-param][cmdline init-param]
673 */
674 if (initargs_offs) {
675 len = xlen + initargs_offs;
676 strcpy(saved_command_line + len, extra_init_args);
677 len += ilen - 4; /* strlen(extra_init_args) */
678 strcpy(saved_command_line + len,
679 boot_command_line + initargs_offs - 1);
680 } else {
681 len = strlen(saved_command_line);
682 strcpy(saved_command_line + len, " -- ");
683 len += 4;
684 strcpy(saved_command_line + len, extra_init_args);
685 }
686 }
687
688 saved_command_line_len = strlen(saved_command_line);
689 }
690
691 /*
692 * We need to finalize in a non-__init function or else race conditions
693 * between the root thread and the init thread may cause start_kernel to
694 * be reaped by free_initmem before the root thread has proceeded to
695 * cpu_idle.
696 *
697 * gcc-3.4 accidentally inlines this function, so use noinline.
698 */
699
700 static __initdata DECLARE_COMPLETION(kthreadd_done);
701
rest_init(void)702 static noinline void __ref __noreturn rest_init(void)
703 {
704 struct task_struct *tsk;
705 int pid;
706
707 rcu_scheduler_starting();
708 /*
709 * We need to spawn init first so that it obtains pid 1, however
710 * the init task will end up wanting to create kthreads, which, if
711 * we schedule it before we create kthreadd, will OOPS.
712 */
713 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
714 /*
715 * Pin init on the boot CPU. Task migration is not properly working
716 * until sched_init_smp() has been run. It will set the allowed
717 * CPUs for init to the non isolated CPUs.
718 */
719 rcu_read_lock();
720 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
721 tsk->flags |= PF_NO_SETAFFINITY;
722 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
723 rcu_read_unlock();
724
725 numa_default_policy();
726 pid = kernel_thread(kthreadd, NULL, NULL, CLONE_FS | CLONE_FILES);
727 rcu_read_lock();
728 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
729 rcu_read_unlock();
730
731 /*
732 * Enable might_sleep() and smp_processor_id() checks.
733 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
734 * kernel_thread() would trigger might_sleep() splats. With
735 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
736 * already, but it's stuck on the kthreadd_done completion.
737 */
738 system_state = SYSTEM_SCHEDULING;
739
740 complete(&kthreadd_done);
741
742 /*
743 * The boot idle thread must execute schedule()
744 * at least once to get things moving:
745 */
746 schedule_preempt_disabled();
747 /* Call into cpu_idle with preempt disabled */
748 cpu_startup_entry(CPUHP_ONLINE);
749 }
750
751 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)752 static int __init do_early_param(char *param, char *val,
753 const char *unused, void *arg)
754 {
755 const struct obs_kernel_param *p;
756
757 for (p = __setup_start; p < __setup_end; p++) {
758 if ((p->early && parameq(param, p->str)) ||
759 (strcmp(param, "console") == 0 &&
760 strcmp(p->str, "earlycon") == 0)
761 ) {
762 if (p->setup_func(val) != 0)
763 pr_warn("Malformed early option '%s'\n", param);
764 }
765 }
766 /* We accept everything at this stage. */
767 return 0;
768 }
769
parse_early_options(char * cmdline)770 void __init parse_early_options(char *cmdline)
771 {
772 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
773 do_early_param);
774 }
775
776 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)777 void __init parse_early_param(void)
778 {
779 static int done __initdata;
780 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
781
782 if (done)
783 return;
784
785 /* All fall through to do_early_param. */
786 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
787 parse_early_options(tmp_cmdline);
788 done = 1;
789 }
790
arch_post_acpi_subsys_init(void)791 void __init __weak arch_post_acpi_subsys_init(void) { }
792
smp_setup_processor_id(void)793 void __init __weak smp_setup_processor_id(void)
794 {
795 }
796
smp_prepare_boot_cpu(void)797 void __init __weak smp_prepare_boot_cpu(void)
798 {
799 }
800
801 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)802 void __init __weak thread_stack_cache_init(void)
803 {
804 }
805 #endif
806
poking_init(void)807 void __init __weak poking_init(void) { }
808
pgtable_cache_init(void)809 void __init __weak pgtable_cache_init(void) { }
810
trap_init(void)811 void __init __weak trap_init(void) { }
812
813 bool initcall_debug;
814 core_param(initcall_debug, initcall_debug, bool, 0644);
815
816 #ifdef TRACEPOINTS_ENABLED
817 static void __init initcall_debug_enable(void);
818 #else
initcall_debug_enable(void)819 static inline void initcall_debug_enable(void)
820 {
821 }
822 #endif
823
824 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
825 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
826 randomize_kstack_offset);
827 DEFINE_PER_CPU(u32, kstack_offset);
828
early_randomize_kstack_offset(char * buf)829 static int __init early_randomize_kstack_offset(char *buf)
830 {
831 int ret;
832 bool bool_result;
833
834 ret = kstrtobool(buf, &bool_result);
835 if (ret)
836 return ret;
837
838 if (bool_result)
839 static_branch_enable(&randomize_kstack_offset);
840 else
841 static_branch_disable(&randomize_kstack_offset);
842 return 0;
843 }
844 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
845 #endif
846
print_unknown_bootoptions(void)847 static void __init print_unknown_bootoptions(void)
848 {
849 char *unknown_options;
850 char *end;
851 const char *const *p;
852 size_t len;
853
854 if (panic_later || (!argv_init[1] && !envp_init[2]))
855 return;
856
857 /*
858 * Determine how many options we have to print out, plus a space
859 * before each
860 */
861 len = 1; /* null terminator */
862 for (p = &argv_init[1]; *p; p++) {
863 len++;
864 len += strlen(*p);
865 }
866 for (p = &envp_init[2]; *p; p++) {
867 len++;
868 len += strlen(*p);
869 }
870
871 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
872 if (!unknown_options) {
873 pr_err("%s: Failed to allocate %zu bytes\n",
874 __func__, len);
875 return;
876 }
877 end = unknown_options;
878
879 for (p = &argv_init[1]; *p; p++)
880 end += sprintf(end, " %s", *p);
881 for (p = &envp_init[2]; *p; p++)
882 end += sprintf(end, " %s", *p);
883
884 /* Start at unknown_options[1] to skip the initial space */
885 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
886 &unknown_options[1]);
887 memblock_free(unknown_options, len);
888 }
889
early_numa_node_init(void)890 static void __init early_numa_node_init(void)
891 {
892 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
893 #ifndef cpu_to_node
894 int cpu;
895
896 /* The early_cpu_to_node() should be ready here. */
897 for_each_possible_cpu(cpu)
898 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
899 #endif
900 #endif
901 }
902
903 #ifdef CONFIG_GKI_DYNAMIC_TASK_STRUCT_SIZE
setup_arch_task_struct_size(void)904 static void __init setup_arch_task_struct_size(void)
905 {
906 arch_task_struct_size = sizeof(struct task_struct);
907 }
908
task_struct_vendor_size_setup(char * str)909 static int __init task_struct_vendor_size_setup(char *str)
910 {
911 u64 size;
912
913 if (!str)
914 return -EINVAL;
915
916 size = memparse(str, &str);
917
918 if (size < 0 || size > CONFIG_GKI_TASK_STRUCT_VENDOR_SIZE_MAX)
919 return -EINVAL;
920
921 arch_task_struct_size = sizeof(struct task_struct) + size;
922
923 return 0;
924 }
925 early_param("android_arch_task_struct_size", task_struct_vendor_size_setup);
926 #endif
927
928 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)929 void start_kernel(void)
930 {
931 char *command_line;
932 char *after_dashes;
933
934 set_task_stack_end_magic(&init_task);
935 smp_setup_processor_id();
936 debug_objects_early_init();
937 init_vmlinux_build_id();
938
939 cgroup_init_early();
940
941 local_irq_disable();
942 early_boot_irqs_disabled = true;
943
944 /*
945 * Interrupts are still disabled. Do necessary setups, then
946 * enable them.
947 */
948 boot_cpu_init();
949 page_address_init();
950 pr_notice("%s", linux_banner);
951 #ifdef CONFIG_GKI_DYNAMIC_TASK_STRUCT_SIZE
952 setup_arch_task_struct_size();
953 #endif
954 setup_arch(&command_line);
955 /* Static keys and static calls are needed by LSMs */
956 jump_label_init();
957 static_call_init();
958 early_security_init();
959 setup_boot_config();
960 setup_command_line(command_line);
961 setup_nr_cpu_ids();
962 setup_per_cpu_areas();
963 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
964 early_numa_node_init();
965 boot_cpu_hotplug_init();
966
967 pr_notice("Kernel command line: %s\n", saved_command_line);
968 /* parameters may set static keys */
969 parse_early_param();
970 after_dashes = parse_args("Booting kernel",
971 static_command_line, __start___param,
972 __stop___param - __start___param,
973 -1, -1, NULL, &unknown_bootoption);
974 print_unknown_bootoptions();
975 if (!IS_ERR_OR_NULL(after_dashes))
976 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
977 NULL, set_init_arg);
978 if (extra_init_args)
979 parse_args("Setting extra init args", extra_init_args,
980 NULL, 0, -1, -1, NULL, set_init_arg);
981
982 /* Architectural and non-timekeeping rng init, before allocator init */
983 random_init_early(command_line);
984
985 /*
986 * These use large bootmem allocations and must precede
987 * initalization of page allocator
988 */
989 setup_log_buf(0);
990 vfs_caches_init_early();
991 sort_main_extable();
992 trap_init();
993 mm_core_init();
994 poking_init();
995 ftrace_init();
996
997 /* trace_printk can be enabled here */
998 early_trace_init();
999
1000 /*
1001 * Set up the scheduler prior starting any interrupts (such as the
1002 * timer interrupt). Full topology setup happens at smp_init()
1003 * time - but meanwhile we still have a functioning scheduler.
1004 */
1005 sched_init();
1006
1007 if (WARN(!irqs_disabled(),
1008 "Interrupts were enabled *very* early, fixing it\n"))
1009 local_irq_disable();
1010 radix_tree_init();
1011 maple_tree_init();
1012
1013 /*
1014 * Set up housekeeping before setting up workqueues to allow the unbound
1015 * workqueue to take non-housekeeping into account.
1016 */
1017 housekeeping_init();
1018
1019 /*
1020 * Allow workqueue creation and work item queueing/cancelling
1021 * early. Work item execution depends on kthreads and starts after
1022 * workqueue_init().
1023 */
1024 workqueue_init_early();
1025
1026 rcu_init();
1027
1028 /* Trace events are available after this */
1029 trace_init();
1030
1031 if (initcall_debug)
1032 initcall_debug_enable();
1033
1034 context_tracking_init();
1035 /* init some links before init_ISA_irqs() */
1036 early_irq_init();
1037 init_IRQ();
1038 tick_init();
1039 rcu_init_nohz();
1040 init_timers();
1041 srcu_init();
1042 hrtimers_init();
1043 softirq_init();
1044 timekeeping_init();
1045 time_init();
1046
1047 /* This must be after timekeeping is initialized */
1048 random_init();
1049
1050 /* These make use of the fully initialized rng */
1051 kfence_init();
1052 boot_init_stack_canary();
1053
1054 perf_event_init();
1055 profile_init();
1056 call_function_init();
1057 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1058
1059 early_boot_irqs_disabled = false;
1060 local_irq_enable();
1061
1062 kmem_cache_init_late();
1063
1064 /*
1065 * HACK ALERT! This is early. We're enabling the console before
1066 * we've done PCI setups etc, and console_init() must be aware of
1067 * this. But we do want output early, in case something goes wrong.
1068 */
1069 console_init();
1070 if (panic_later)
1071 panic("Too many boot %s vars at `%s'", panic_later,
1072 panic_param);
1073
1074 lockdep_init();
1075
1076 /*
1077 * Need to run this when irqs are enabled, because it wants
1078 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1079 * too:
1080 */
1081 locking_selftest();
1082
1083 #ifdef CONFIG_BLK_DEV_INITRD
1084 if (initrd_start && !initrd_below_start_ok &&
1085 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1086 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1087 page_to_pfn(virt_to_page((void *)initrd_start)),
1088 min_low_pfn);
1089 initrd_start = 0;
1090 }
1091 #endif
1092 setup_per_cpu_pageset();
1093 numa_policy_init();
1094 acpi_early_init();
1095 if (late_time_init)
1096 late_time_init();
1097 sched_clock_init();
1098 calibrate_delay();
1099
1100 arch_cpu_finalize_init();
1101
1102 pid_idr_init();
1103 anon_vma_init();
1104 #ifdef CONFIG_X86
1105 if (efi_enabled(EFI_RUNTIME_SERVICES))
1106 efi_enter_virtual_mode();
1107 #endif
1108 thread_stack_cache_init();
1109 cred_init();
1110 fork_init();
1111 proc_caches_init();
1112 uts_ns_init();
1113 key_init();
1114 security_init();
1115 dbg_late_init();
1116 net_ns_init();
1117 vfs_caches_init();
1118 pagecache_init();
1119 signals_init();
1120 seq_file_init();
1121 proc_root_init();
1122 nsfs_init();
1123 pidfs_init();
1124 cpuset_init();
1125 mem_cgroup_init();
1126 cgroup_init();
1127 taskstats_init_early();
1128 delayacct_init();
1129
1130 acpi_subsystem_init();
1131 arch_post_acpi_subsys_init();
1132 kcsan_init();
1133
1134 /* Do the rest non-__init'ed, we're now alive */
1135 rest_init();
1136
1137 /*
1138 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1139 * and older.
1140 */
1141 #if !__has_attribute(__no_stack_protector__)
1142 prevent_tail_call_optimization();
1143 #endif
1144 }
1145
1146 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1147 static void __init do_ctors(void)
1148 {
1149 /*
1150 * For UML, the constructors have already been called by the
1151 * normal setup code as it's just a normal ELF binary, so we
1152 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1153 * even on UML for modules.
1154 */
1155 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1156 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1157
1158 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1159 (*fn)();
1160 #endif
1161 }
1162
1163 #ifdef CONFIG_KALLSYMS
1164 struct blacklist_entry {
1165 struct list_head next;
1166 char *buf;
1167 };
1168
1169 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1170
initcall_blacklist(char * str)1171 static int __init initcall_blacklist(char *str)
1172 {
1173 char *str_entry;
1174 struct blacklist_entry *entry;
1175
1176 /* str argument is a comma-separated list of functions */
1177 do {
1178 str_entry = strsep(&str, ",");
1179 if (str_entry) {
1180 pr_debug("blacklisting initcall %s\n", str_entry);
1181 entry = memblock_alloc(sizeof(*entry),
1182 SMP_CACHE_BYTES);
1183 if (!entry)
1184 panic("%s: Failed to allocate %zu bytes\n",
1185 __func__, sizeof(*entry));
1186 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1187 SMP_CACHE_BYTES);
1188 if (!entry->buf)
1189 panic("%s: Failed to allocate %zu bytes\n",
1190 __func__, strlen(str_entry) + 1);
1191 strcpy(entry->buf, str_entry);
1192 list_add(&entry->next, &blacklisted_initcalls);
1193 }
1194 } while (str_entry);
1195
1196 return 1;
1197 }
1198
initcall_blacklisted(initcall_t fn)1199 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1200 {
1201 struct blacklist_entry *entry;
1202 char fn_name[KSYM_SYMBOL_LEN];
1203 unsigned long addr;
1204
1205 if (list_empty(&blacklisted_initcalls))
1206 return false;
1207
1208 addr = (unsigned long) dereference_function_descriptor(fn);
1209 sprint_symbol_no_offset(fn_name, addr);
1210
1211 /*
1212 * fn will be "function_name [module_name]" where [module_name] is not
1213 * displayed for built-in init functions. Strip off the [module_name].
1214 */
1215 strreplace(fn_name, ' ', '\0');
1216
1217 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1218 if (!strcmp(fn_name, entry->buf)) {
1219 pr_debug("initcall %s blacklisted\n", fn_name);
1220 return true;
1221 }
1222 }
1223
1224 return false;
1225 }
1226 #else
initcall_blacklist(char * str)1227 static int __init initcall_blacklist(char *str)
1228 {
1229 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1230 return 0;
1231 }
1232
initcall_blacklisted(initcall_t fn)1233 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1234 {
1235 return false;
1236 }
1237 #endif
1238 __setup("initcall_blacklist=", initcall_blacklist);
1239
1240 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1241 trace_initcall_start_cb(void *data, initcall_t fn)
1242 {
1243 ktime_t *calltime = data;
1244
1245 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1246 *calltime = ktime_get();
1247 }
1248
1249 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1250 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1251 {
1252 ktime_t rettime, *calltime = data;
1253
1254 rettime = ktime_get();
1255 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1256 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1257 }
1258
1259 static ktime_t initcall_calltime;
1260
1261 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1262 static void __init initcall_debug_enable(void)
1263 {
1264 int ret;
1265
1266 ret = register_trace_initcall_start(trace_initcall_start_cb,
1267 &initcall_calltime);
1268 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1269 &initcall_calltime);
1270 WARN(ret, "Failed to register initcall tracepoints\n");
1271 }
1272 # define do_trace_initcall_start trace_initcall_start
1273 # define do_trace_initcall_finish trace_initcall_finish
1274 #else
do_trace_initcall_start(initcall_t fn)1275 static inline void do_trace_initcall_start(initcall_t fn)
1276 {
1277 if (!initcall_debug)
1278 return;
1279 trace_initcall_start_cb(&initcall_calltime, fn);
1280 }
do_trace_initcall_finish(initcall_t fn,int ret)1281 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1282 {
1283 if (!initcall_debug)
1284 return;
1285 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1286 }
1287 #endif /* !TRACEPOINTS_ENABLED */
1288
do_one_initcall(initcall_t fn)1289 int __init_or_module do_one_initcall(initcall_t fn)
1290 {
1291 int count = preempt_count();
1292 char msgbuf[64];
1293 int ret;
1294
1295 if (initcall_blacklisted(fn))
1296 return -EPERM;
1297
1298 do_trace_initcall_start(fn);
1299 ret = fn();
1300 do_trace_initcall_finish(fn, ret);
1301
1302 msgbuf[0] = 0;
1303
1304 if (preempt_count() != count) {
1305 sprintf(msgbuf, "preemption imbalance ");
1306 preempt_count_set(count);
1307 }
1308 if (irqs_disabled()) {
1309 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1310 local_irq_enable();
1311 }
1312 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1313
1314 add_latent_entropy();
1315 return ret;
1316 }
1317
1318
1319 static initcall_entry_t *initcall_levels[] __initdata = {
1320 __initcall0_start,
1321 __initcall1_start,
1322 __initcall2_start,
1323 __initcall3_start,
1324 __initcall4_start,
1325 __initcall5_start,
1326 __initcall6_start,
1327 __initcall7_start,
1328 __initcall_end,
1329 };
1330
1331 /* Keep these in sync with initcalls in include/linux/init.h */
1332 static const char *initcall_level_names[] __initdata = {
1333 "pure",
1334 "core",
1335 "postcore",
1336 "arch",
1337 "subsys",
1338 "fs",
1339 "device",
1340 "late",
1341 };
1342
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1343 static int __init ignore_unknown_bootoption(char *param, char *val,
1344 const char *unused, void *arg)
1345 {
1346 return 0;
1347 }
1348
do_initcall_level(int level,char * command_line)1349 static void __init do_initcall_level(int level, char *command_line)
1350 {
1351 initcall_entry_t *fn;
1352
1353 parse_args(initcall_level_names[level],
1354 command_line, __start___param,
1355 __stop___param - __start___param,
1356 level, level,
1357 NULL, ignore_unknown_bootoption);
1358
1359 trace_initcall_level(initcall_level_names[level]);
1360 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1361 do_one_initcall(initcall_from_entry(fn));
1362 }
1363
do_initcalls(void)1364 static void __init do_initcalls(void)
1365 {
1366 int level;
1367 size_t len = saved_command_line_len + 1;
1368 char *command_line;
1369
1370 command_line = kzalloc(len, GFP_KERNEL);
1371 if (!command_line)
1372 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1373
1374 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1375 /* Parser modifies command_line, restore it each time */
1376 strcpy(command_line, saved_command_line);
1377 do_initcall_level(level, command_line);
1378 }
1379
1380 kfree(command_line);
1381 }
1382
1383 /*
1384 * Ok, the machine is now initialized. None of the devices
1385 * have been touched yet, but the CPU subsystem is up and
1386 * running, and memory and process management works.
1387 *
1388 * Now we can finally start doing some real work..
1389 */
do_basic_setup(void)1390 static void __init do_basic_setup(void)
1391 {
1392 cpuset_init_smp();
1393 driver_init();
1394 init_irq_proc();
1395 do_ctors();
1396 do_initcalls();
1397 }
1398
do_pre_smp_initcalls(void)1399 static void __init do_pre_smp_initcalls(void)
1400 {
1401 initcall_entry_t *fn;
1402
1403 trace_initcall_level("early");
1404 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1405 do_one_initcall(initcall_from_entry(fn));
1406 }
1407
run_init_process(const char * init_filename)1408 static int run_init_process(const char *init_filename)
1409 {
1410 const char *const *p;
1411
1412 argv_init[0] = init_filename;
1413 pr_info("Run %s as init process\n", init_filename);
1414 pr_debug(" with arguments:\n");
1415 for (p = argv_init; *p; p++)
1416 pr_debug(" %s\n", *p);
1417 pr_debug(" with environment:\n");
1418 for (p = envp_init; *p; p++)
1419 pr_debug(" %s\n", *p);
1420 return kernel_execve(init_filename, argv_init, envp_init);
1421 }
1422
try_to_run_init_process(const char * init_filename)1423 static int try_to_run_init_process(const char *init_filename)
1424 {
1425 int ret;
1426
1427 ret = run_init_process(init_filename);
1428
1429 if (ret && ret != -ENOENT) {
1430 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1431 init_filename, ret);
1432 }
1433
1434 return ret;
1435 }
1436
1437 static noinline void __init kernel_init_freeable(void);
1438
1439 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1440 bool rodata_enabled __ro_after_init = true;
1441
1442 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1443 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1444 #endif
1445
set_debug_rodata(char * str)1446 static int __init set_debug_rodata(char *str)
1447 {
1448 if (arch_parse_debug_rodata(str))
1449 return 0;
1450
1451 if (str && !strcmp(str, "on"))
1452 rodata_enabled = true;
1453 else if (str && !strcmp(str, "off"))
1454 rodata_enabled = false;
1455 else
1456 pr_warn("Invalid option string for rodata: '%s'\n", str);
1457 return 0;
1458 }
1459 early_param("rodata", set_debug_rodata);
1460 #endif
1461
mark_readonly(void)1462 static void mark_readonly(void)
1463 {
1464 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1465 /*
1466 * load_module() results in W+X mappings, which are cleaned
1467 * up with init_free_wq. Let's make sure that queued work is
1468 * flushed so that we don't hit false positives looking for
1469 * insecure pages which are W+X.
1470 */
1471 flush_module_init_free_work();
1472 jump_label_init_ro();
1473 mark_rodata_ro();
1474 debug_checkwx();
1475 rodata_test();
1476 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1477 pr_info("Kernel memory protection disabled.\n");
1478 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1479 pr_warn("Kernel memory protection not selected by kernel config.\n");
1480 } else {
1481 pr_warn("This architecture does not have kernel memory protection.\n");
1482 }
1483 }
1484
free_initmem(void)1485 void __weak free_initmem(void)
1486 {
1487 free_initmem_default(POISON_FREE_INITMEM);
1488 }
1489
kernel_init(void * unused)1490 static int __ref kernel_init(void *unused)
1491 {
1492 int ret;
1493
1494 /*
1495 * Wait until kthreadd is all set-up.
1496 */
1497 wait_for_completion(&kthreadd_done);
1498
1499 kernel_init_freeable();
1500 /* need to finish all async __init code before freeing the memory */
1501 async_synchronize_full();
1502
1503 system_state = SYSTEM_FREEING_INITMEM;
1504 kprobe_free_init_mem();
1505 ftrace_free_init_mem();
1506 kgdb_free_init_mem();
1507 exit_boot_config();
1508 free_initmem();
1509 mark_readonly();
1510
1511 /*
1512 * Kernel mappings are now finalized - update the userspace page-table
1513 * to finalize PTI.
1514 */
1515 pti_finalize();
1516
1517 system_state = SYSTEM_RUNNING;
1518 numa_default_policy();
1519
1520 rcu_end_inkernel_boot();
1521
1522 do_sysctl_args();
1523
1524 if (ramdisk_execute_command) {
1525 ret = run_init_process(ramdisk_execute_command);
1526 if (!ret)
1527 return 0;
1528 pr_err("Failed to execute %s (error %d)\n",
1529 ramdisk_execute_command, ret);
1530 }
1531
1532 /*
1533 * We try each of these until one succeeds.
1534 *
1535 * The Bourne shell can be used instead of init if we are
1536 * trying to recover a really broken machine.
1537 */
1538 if (execute_command) {
1539 ret = run_init_process(execute_command);
1540 if (!ret)
1541 return 0;
1542 panic("Requested init %s failed (error %d).",
1543 execute_command, ret);
1544 }
1545
1546 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1547 ret = run_init_process(CONFIG_DEFAULT_INIT);
1548 if (ret)
1549 pr_err("Default init %s failed (error %d)\n",
1550 CONFIG_DEFAULT_INIT, ret);
1551 else
1552 return 0;
1553 }
1554
1555 if (!try_to_run_init_process("/sbin/init") ||
1556 !try_to_run_init_process("/etc/init") ||
1557 !try_to_run_init_process("/bin/init") ||
1558 !try_to_run_init_process("/bin/sh"))
1559 return 0;
1560
1561 panic("No working init found. Try passing init= option to kernel. "
1562 "See Linux Documentation/admin-guide/init.rst for guidance.");
1563 }
1564
1565 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1566 void __init console_on_rootfs(void)
1567 {
1568 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1569
1570 if (IS_ERR(file)) {
1571 pr_err("Warning: unable to open an initial console.\n");
1572 return;
1573 }
1574 init_dup(file);
1575 init_dup(file);
1576 init_dup(file);
1577 fput(file);
1578 }
1579
kernel_init_freeable(void)1580 static noinline void __init kernel_init_freeable(void)
1581 {
1582 /* Now the scheduler is fully set up and can do blocking allocations */
1583 gfp_allowed_mask = __GFP_BITS_MASK;
1584
1585 /*
1586 * init can allocate pages on any node
1587 */
1588 set_mems_allowed(node_states[N_MEMORY]);
1589
1590 cad_pid = get_pid(task_pid(current));
1591
1592 smp_prepare_cpus(setup_max_cpus);
1593
1594 workqueue_init();
1595
1596 init_mm_internals();
1597
1598 rcu_init_tasks_generic();
1599 do_pre_smp_initcalls();
1600 lockup_detector_init();
1601
1602 smp_init();
1603 sched_init_smp();
1604
1605 workqueue_init_topology();
1606 async_init();
1607 padata_init();
1608 page_alloc_init_late();
1609
1610 do_basic_setup();
1611
1612 kunit_run_all_tests();
1613
1614 wait_for_initramfs();
1615 console_on_rootfs();
1616
1617 /*
1618 * check if there is an early userspace init. If yes, let it do all
1619 * the work
1620 */
1621 if (init_eaccess(ramdisk_execute_command) != 0) {
1622 ramdisk_execute_command = NULL;
1623 prepare_namespace();
1624 }
1625
1626 /*
1627 * Ok, we have completed the initial bootup, and
1628 * we're essentially up and running. Get rid of the
1629 * initmem segments and start the user-mode stuff..
1630 *
1631 * rootfs is available now, try loading the public keys
1632 * and default modules
1633 */
1634
1635 integrity_load_keys();
1636 }
1637