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