projects/linux·blob v6.3 · refs

1664 lines · 40.3 KiB

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 <[email protected]>
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/page_ext.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/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/mem_encrypt.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 <net/net_namespace.h>
105
106#include <asm/io.h>
107#include <asm/bugs.h>
108#include <asm/setup.h>
109#include <asm/sections.h>
110#include <asm/cacheflush.h>
111
112#define CREATE_TRACE_POINTS
113#include <trace/events/initcall.h>
114
115#include <kunit/test.h>
116
117static int kernel_init(void *);
118
119extern void init_IRQ(void);
120extern void radix_tree_init(void);
121extern void maple_tree_init(void);
122
123/*
124 * Debug helper: via this flag we know that we are in 'early bootup code'
125 * where only the boot processor is running with IRQ disabled. This means
126 * two things - IRQ must not be enabled before the flag is cleared and some
127 * operations which are not allowed with IRQ disabled are allowed while the
128 * flag is set.
129 */
130bool early_boot_irqs_disabled __read_mostly;
131
132enum system_states system_state __read_mostly;
133EXPORT_SYMBOL(system_state);
134
135/*
136 * Boot command-line arguments
137 */
138#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
139#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
140
141extern void time_init(void);
142/* Default late time init is NULL. archs can override this later. */
143void (*__initdata late_time_init)(void);
144
145/* Untouched command line saved by arch-specific code. */
146char __initdata boot_command_line[COMMAND_LINE_SIZE];
147/* Untouched saved command line (eg. for /proc) */
148char *saved_command_line __ro_after_init;
149unsigned int saved_command_line_len __ro_after_init;
150/* Command line for parameter parsing */
151static char *static_command_line;
152/* Untouched extra command line */
153static char *extra_command_line;
154/* Extra init arguments */
155static char *extra_init_args;
156
157#ifdef CONFIG_BOOT_CONFIG
158/* Is bootconfig on command line? */
159static bool bootconfig_found;
160static size_t initargs_offs;
161#else
162# define bootconfig_found false
163# define initargs_offs 0
164#endif
165
166static char *execute_command;
167static char *ramdisk_execute_command = "/init";
168
169/*
170 * Used to generate warnings if static_key manipulation functions are used
171 * before jump_label_init is called.
172 */
173bool static_key_initialized __read_mostly;
174EXPORT_SYMBOL_GPL(static_key_initialized);
175
176/*
177 * If set, this is an indication to the drivers that reset the underlying
178 * device before going ahead with the initialization otherwise driver might
179 * rely on the BIOS and skip the reset operation.
180 *
181 * This is useful if kernel is booting in an unreliable environment.
182 * For ex. kdump situation where previous kernel has crashed, BIOS has been
183 * skipped and devices will be in unknown state.
184 */
185unsigned int reset_devices;
186EXPORT_SYMBOL(reset_devices);
187
188static int __init set_reset_devices(char *str)
189{
190 reset_devices = 1;
191 return 1;
192}
193
194__setup("reset_devices", set_reset_devices);
195
196static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
197const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
198static const char *panic_later, *panic_param;
199
200extern const struct obs_kernel_param __setup_start[], __setup_end[];
201
202static bool __init obsolete_checksetup(char *line)
203{
204 const struct obs_kernel_param *p;
205 bool had_early_param = false;
206
207 p = __setup_start;
208 do {
209 int n = strlen(p->str);
210 if (parameqn(line, p->str, n)) {
211 if (p->early) {
212 /* Already done in parse_early_param?
213 * (Needs exact match on param part).
214 * Keep iterating, as we can have early
215 * params and __setups of same names 8( */
216 if (line[n] == '\0' || line[n] == '=')
217 had_early_param = true;
218 } else if (!p->setup_func) {
219 pr_warn("Parameter %s is obsolete, ignored\n",
220 p->str);
221 return true;
222 } else if (p->setup_func(line + n))
223 return true;
224 }
225 p++;
226 } while (p < __setup_end);
227
228 return had_early_param;
229}
230
231/*
232 * This should be approx 2 Bo*oMips to start (note initial shift), and will
233 * still work even if initially too large, it will just take slightly longer
234 */
235unsigned long loops_per_jiffy = (1<<12);
236EXPORT_SYMBOL(loops_per_jiffy);
237
238static int __init debug_kernel(char *str)
239{
240 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
241 return 0;
242}
243
244static int __init quiet_kernel(char *str)
245{
246 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
247 return 0;
248}
249
250early_param("debug", debug_kernel);
251early_param("quiet", quiet_kernel);
252
253static int __init loglevel(char *str)
254{
255 int newlevel;
256
257 /*
258 * Only update loglevel value when a correct setting was passed,
259 * to prevent blind crashes (when loglevel being set to 0) that
260 * are quite hard to debug
261 */
262 if (get_option(&str, &newlevel)) {
263 console_loglevel = newlevel;
264 return 0;
265 }
266
267 return -EINVAL;
268}
269
270early_param("loglevel", loglevel);
271
272#ifdef CONFIG_BLK_DEV_INITRD
273static void * __init get_boot_config_from_initrd(size_t *_size)
274{
275 u32 size, csum;
276 char *data;
277 u32 *hdr;
278 int i;
279
280 if (!initrd_end)
281 return NULL;
282
283 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
284 /*
285 * Since Grub may align the size of initrd to 4, we must
286 * check the preceding 3 bytes as well.
287 */
288 for (i = 0; i < 4; i++) {
289 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
290 goto found;
291 data--;
292 }
293 return NULL;
294
295found:
296 hdr = (u32 *)(data - 8);
297 size = le32_to_cpu(hdr[0]);
298 csum = le32_to_cpu(hdr[1]);
299
300 data = ((void *)hdr) - size;
301 if ((unsigned long)data < initrd_start) {
302 pr_err("bootconfig size %d is greater than initrd size %ld\n",
303 size, initrd_end - initrd_start);
304 return NULL;
305 }
306
307 if (xbc_calc_checksum(data, size) != csum) {
308 pr_err("bootconfig checksum failed\n");
309 return NULL;
310 }
311
312 /* Remove bootconfig from initramfs/initrd */
313 initrd_end = (unsigned long)data;
314 if (_size)
315 *_size = size;
316
317 return data;
318}
319#else
320static void * __init get_boot_config_from_initrd(size_t *_size)
321{
322 return NULL;
323}
324#endif
325
326#ifdef CONFIG_BOOT_CONFIG
327
328static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
329
330#define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
331
332static int __init xbc_snprint_cmdline(char *buf, size_t size,
333 struct xbc_node *root)
334{
335 struct xbc_node *knode, *vnode;
336 char *end = buf + size;
337 const char *val;
338 int ret;
339
340 xbc_node_for_each_key_value(root, knode, val) {
341 ret = xbc_node_compose_key_after(root, knode,
342 xbc_namebuf, XBC_KEYLEN_MAX);
343 if (ret < 0)
344 return ret;
345
346 vnode = xbc_node_get_child(knode);
347 if (!vnode) {
348 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
349 if (ret < 0)
350 return ret;
351 buf += ret;
352 continue;
353 }
354 xbc_array_for_each_value(vnode, val) {
355 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
356 xbc_namebuf, val);
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 */
368static 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
399static 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
408static int __init warn_bootconfig(char *str)
409{
410 /* The 'bootconfig' has been handled by bootconfig_params(). */
411 return 0;
412}
413
414static 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
472static void __init exit_boot_config(void)
473{
474 xbc_exit();
475}
476
477#else /* !CONFIG_BOOT_CONFIG */
478
479static void __init setup_boot_config(void)
480{
481 /* Remove bootconfig data from initrd */
482 get_boot_config_from_initrd(NULL);
483}
484
485static 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
495early_param("bootconfig", warn_bootconfig);
496
497/* Change NUL term back to "=", to make "param" the whole string. */
498static void __init repair_env_string(char *param, char *val)
499{
500 if (val) {
501 /* param=val or param="val"? */
502 if (val == param+strlen(param)+1)
503 val[-1] = '=';
504 else if (val == param+strlen(param)+2) {
505 val[-2] = '=';
506 memmove(val-1, val, strlen(val)+1);
507 } else
508 BUG();
509 }
510}
511
512/* Anything after -- gets handed straight to init. */
513static int __init set_init_arg(char *param, char *val,
514 const char *unused, void *arg)
515{
516 unsigned int i;
517
518 if (panic_later)
519 return 0;
520
521 repair_env_string(param, val);
522
523 for (i = 0; argv_init[i]; i++) {
524 if (i == MAX_INIT_ARGS) {
525 panic_later = "init";
526 panic_param = param;
527 return 0;
528 }
529 }
530 argv_init[i] = param;
531 return 0;
532}
533
534/*
535 * Unknown boot options get handed to init, unless they look like
536 * unused parameters (modprobe will find them in /proc/cmdline).
537 */
538static int __init unknown_bootoption(char *param, char *val,
539 const char *unused, void *arg)
540{
541 size_t len = strlen(param);
542
543 repair_env_string(param, val);
544
545 /* Handle obsolete-style parameters */
546 if (obsolete_checksetup(param))
547 return 0;
548
549 /* Unused module parameter. */
550 if (strnchr(param, len, '.'))
551 return 0;
552
553 if (panic_later)
554 return 0;
555
556 if (val) {
557 /* Environment option */
558 unsigned int i;
559 for (i = 0; envp_init[i]; i++) {
560 if (i == MAX_INIT_ENVS) {
561 panic_later = "env";
562 panic_param = param;
563 }
564 if (!strncmp(param, envp_init[i], len+1))
565 break;
566 }
567 envp_init[i] = param;
568 } else {
569 /* Command line option */
570 unsigned int i;
571 for (i = 0; argv_init[i]; i++) {
572 if (i == MAX_INIT_ARGS) {
573 panic_later = "init";
574 panic_param = param;
575 }
576 }
577 argv_init[i] = param;
578 }
579 return 0;
580}
581
582static int __init init_setup(char *str)
583{
584 unsigned int i;
585
586 execute_command = str;
587 /*
588 * In case LILO is going to boot us with default command line,
589 * it prepends "auto" before the whole cmdline which makes
590 * the shell think it should execute a script with such name.
591 * So we ignore all arguments entered _before_ init=... [MJ]
592 */
593 for (i = 1; i < MAX_INIT_ARGS; i++)
594 argv_init[i] = NULL;
595 return 1;
596}
597__setup("init=", init_setup);
598
599static int __init rdinit_setup(char *str)
600{
601 unsigned int i;
602
603 ramdisk_execute_command = str;
604 /* See "auto" comment in init_setup */
605 for (i = 1; i < MAX_INIT_ARGS; i++)
606 argv_init[i] = NULL;
607 return 1;
608}
609__setup("rdinit=", rdinit_setup);
610
611#ifndef CONFIG_SMP
612static const unsigned int setup_max_cpus = NR_CPUS;
613static inline void setup_nr_cpu_ids(void) { }
614static inline void smp_prepare_cpus(unsigned int maxcpus) { }
615#endif
616
617/*
618 * We need to store the untouched command line for future reference.
619 * We also need to store the touched command line since the parameter
620 * parsing is performed in place, and we should allow a component to
621 * store reference of name/value for future reference.
622 */
623static void __init setup_command_line(char *command_line)
624{
625 size_t len, xlen = 0, ilen = 0;
626
627 if (extra_command_line)
628 xlen = strlen(extra_command_line);
629 if (extra_init_args)
630 ilen = strlen(extra_init_args) + 4; /* for " -- " */
631
632 len = xlen + strlen(boot_command_line) + 1;
633
634 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
635 if (!saved_command_line)
636 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
637
638 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
639 if (!static_command_line)
640 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
641
642 if (xlen) {
643 /*
644 * We have to put extra_command_line before boot command
645 * lines because there could be dashes (separator of init
646 * command line) in the command lines.
647 */
648 strcpy(saved_command_line, extra_command_line);
649 strcpy(static_command_line, extra_command_line);
650 }
651 strcpy(saved_command_line + xlen, boot_command_line);
652 strcpy(static_command_line + xlen, command_line);
653
654 if (ilen) {
655 /*
656 * Append supplemental init boot args to saved_command_line
657 * so that user can check what command line options passed
658 * to init.
659 * The order should always be
660 * " -- "[bootconfig init-param][cmdline init-param]
661 */
662 if (initargs_offs) {
663 len = xlen + initargs_offs;
664 strcpy(saved_command_line + len, extra_init_args);
665 len += ilen - 4; /* strlen(extra_init_args) */
666 strcpy(saved_command_line + len,
667 boot_command_line + initargs_offs - 1);
668 } else {
669 len = strlen(saved_command_line);
670 strcpy(saved_command_line + len, " -- ");
671 len += 4;
672 strcpy(saved_command_line + len, extra_init_args);
673 }
674 }
675
676 saved_command_line_len = strlen(saved_command_line);
677}
678
679/*
680 * We need to finalize in a non-__init function or else race conditions
681 * between the root thread and the init thread may cause start_kernel to
682 * be reaped by free_initmem before the root thread has proceeded to
683 * cpu_idle.
684 *
685 * gcc-3.4 accidentally inlines this function, so use noinline.
686 */
687
688static __initdata DECLARE_COMPLETION(kthreadd_done);
689
690noinline void __ref rest_init(void)
691{
692 struct task_struct *tsk;
693 int pid;
694
695 rcu_scheduler_starting();
696 /*
697 * We need to spawn init first so that it obtains pid 1, however
698 * the init task will end up wanting to create kthreads, which, if
699 * we schedule it before we create kthreadd, will OOPS.
700 */
701 pid = user_mode_thread(kernel_init, NULL, CLONE_FS);
702 /*
703 * Pin init on the boot CPU. Task migration is not properly working
704 * until sched_init_smp() has been run. It will set the allowed
705 * CPUs for init to the non isolated CPUs.
706 */
707 rcu_read_lock();
708 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
709 tsk->flags |= PF_NO_SETAFFINITY;
710 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
711 rcu_read_unlock();
712
713 numa_default_policy();
714 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
715 rcu_read_lock();
716 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
717 rcu_read_unlock();
718
719 /*
720 * Enable might_sleep() and smp_processor_id() checks.
721 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
722 * kernel_thread() would trigger might_sleep() splats. With
723 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
724 * already, but it's stuck on the kthreadd_done completion.
725 */
726 system_state = SYSTEM_SCHEDULING;
727
728 complete(&kthreadd_done);
729
730 /*
731 * The boot idle thread must execute schedule()
732 * at least once to get things moving:
733 */
734 schedule_preempt_disabled();
735 /* Call into cpu_idle with preempt disabled */
736 cpu_startup_entry(CPUHP_ONLINE);
737}
738
739/* Check for early params. */
740static int __init do_early_param(char *param, char *val,
741 const char *unused, void *arg)
742{
743 const struct obs_kernel_param *p;
744
745 for (p = __setup_start; p < __setup_end; p++) {
746 if ((p->early && parameq(param, p->str)) ||
747 (strcmp(param, "console") == 0 &&
748 strcmp(p->str, "earlycon") == 0)
749 ) {
750 if (p->setup_func(val) != 0)
751 pr_warn("Malformed early option '%s'\n", param);
752 }
753 }
754 /* We accept everything at this stage. */
755 return 0;
756}
757
758void __init parse_early_options(char *cmdline)
759{
760 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
761 do_early_param);
762}
763
764/* Arch code calls this early on, or if not, just before other parsing. */
765void __init parse_early_param(void)
766{
767 static int done __initdata;
768 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
769
770 if (done)
771 return;
772
773 /* All fall through to do_early_param. */
774 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
775 parse_early_options(tmp_cmdline);
776 done = 1;
777}
778
779void __init __weak arch_post_acpi_subsys_init(void) { }
780
781void __init __weak smp_setup_processor_id(void)
782{
783}
784
785# if THREAD_SIZE >= PAGE_SIZE
786void __init __weak thread_stack_cache_init(void)
787{
788}
789#endif
790
791void __init __weak mem_encrypt_init(void) { }
792
793void __init __weak poking_init(void) { }
794
795void __init __weak pgtable_cache_init(void) { }
796
797void __init __weak trap_init(void) { }
798
799bool initcall_debug;
800core_param(initcall_debug, initcall_debug, bool, 0644);
801
802#ifdef TRACEPOINTS_ENABLED
803static void __init initcall_debug_enable(void);
804#else
805static inline void initcall_debug_enable(void)
806{
807}
808#endif
809
810/* Report memory auto-initialization states for this boot. */
811static void __init report_meminit(void)
812{
813 const char *stack;
814
815 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
816 stack = "all(pattern)";
817 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
818 stack = "all(zero)";
819 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
820 stack = "byref_all(zero)";
821 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
822 stack = "byref(zero)";
823 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
824 stack = "__user(zero)";
825 else
826 stack = "off";
827
828 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
829 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
830 want_init_on_free() ? "on" : "off");
831 if (want_init_on_free())
832 pr_info("mem auto-init: clearing system memory may take some time...\n");
833}
834
835/*
836 * Set up kernel memory allocators
837 */
838static void __init mm_init(void)
839{
840 /*
841 * page_ext requires contiguous pages,
842 * bigger than MAX_ORDER unless SPARSEMEM.
843 */
844 page_ext_init_flatmem();
845 init_mem_debugging_and_hardening();
846 kfence_alloc_pool();
847 report_meminit();
848 kmsan_init_shadow();
849 stack_depot_early_init();
850 mem_init();
851 mem_init_print_info();
852 kmem_cache_init();
853 /*
854 * page_owner must be initialized after buddy is ready, and also after
855 * slab is ready so that stack_depot_init() works properly
856 */
857 page_ext_init_flatmem_late();
858 kmemleak_init();
859 pgtable_init();
860 debug_objects_mem_init();
861 vmalloc_init();
862 /* If no deferred init page_ext now, as vmap is fully initialized */
863 if (!deferred_struct_pages)
864 page_ext_init();
865 /* Should be run before the first non-init thread is created */
866 init_espfix_bsp();
867 /* Should be run after espfix64 is set up. */
868 pti_init();
869 kmsan_init_runtime();
870 mm_cache_init();
871}
872
873#ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
874DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
875 randomize_kstack_offset);
876DEFINE_PER_CPU(u32, kstack_offset);
877
878static int __init early_randomize_kstack_offset(char *buf)
879{
880 int ret;
881 bool bool_result;
882
883 ret = kstrtobool(buf, &bool_result);
884 if (ret)
885 return ret;
886
887 if (bool_result)
888 static_branch_enable(&randomize_kstack_offset);
889 else
890 static_branch_disable(&randomize_kstack_offset);
891 return 0;
892}
893early_param("randomize_kstack_offset", early_randomize_kstack_offset);
894#endif
895
896void __init __weak arch_call_rest_init(void)
897{
898 rest_init();
899}
900
901static void __init print_unknown_bootoptions(void)
902{
903 char *unknown_options;
904 char *end;
905 const char *const *p;
906 size_t len;
907
908 if (panic_later || (!argv_init[1] && !envp_init[2]))
909 return;
910
911 /*
912 * Determine how many options we have to print out, plus a space
913 * before each
914 */
915 len = 1; /* null terminator */
916 for (p = &argv_init[1]; *p; p++) {
917 len++;
918 len += strlen(*p);
919 }
920 for (p = &envp_init[2]; *p; p++) {
921 len++;
922 len += strlen(*p);
923 }
924
925 unknown_options = memblock_alloc(len, SMP_CACHE_BYTES);
926 if (!unknown_options) {
927 pr_err("%s: Failed to allocate %zu bytes\n",
928 __func__, len);
929 return;
930 }
931 end = unknown_options;
932
933 for (p = &argv_init[1]; *p; p++)
934 end += sprintf(end, " %s", *p);
935 for (p = &envp_init[2]; *p; p++)
936 end += sprintf(end, " %s", *p);
937
938 /* Start at unknown_options[1] to skip the initial space */
939 pr_notice("Unknown kernel command line parameters \"%s\", will be passed to user space.\n",
940 &unknown_options[1]);
941 memblock_free(unknown_options, len);
942}
943
944asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
945{
946 char *command_line;
947 char *after_dashes;
948
949 set_task_stack_end_magic(&init_task);
950 smp_setup_processor_id();
951 debug_objects_early_init();
952 init_vmlinux_build_id();
953
954 cgroup_init_early();
955
956 local_irq_disable();
957 early_boot_irqs_disabled = true;
958
959 /*
960 * Interrupts are still disabled. Do necessary setups, then
961 * enable them.
962 */
963 boot_cpu_init();
964 page_address_init();
965 pr_notice("%s", linux_banner);
966 early_security_init();
967 setup_arch(&command_line);
968 setup_boot_config();
969 setup_command_line(command_line);
970 setup_nr_cpu_ids();
971 setup_per_cpu_areas();
972 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
973 boot_cpu_hotplug_init();
974
975 build_all_zonelists(NULL);
976 page_alloc_init();
977
978 pr_notice("Kernel command line: %s\n", saved_command_line);
979 /* parameters may set static keys */
980 jump_label_init();
981 parse_early_param();
982 after_dashes = parse_args("Booting kernel",
983 static_command_line, __start___param,
984 __stop___param - __start___param,
985 -1, -1, NULL, &unknown_bootoption);
986 print_unknown_bootoptions();
987 if (!IS_ERR_OR_NULL(after_dashes))
988 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
989 NULL, set_init_arg);
990 if (extra_init_args)
991 parse_args("Setting extra init args", extra_init_args,
992 NULL, 0, -1, -1, NULL, set_init_arg);
993
994 /* Architectural and non-timekeeping rng init, before allocator init */
995 random_init_early(command_line);
996
997 /*
998 * These use large bootmem allocations and must precede
999 * kmem_cache_init()
1000 */
1001 setup_log_buf(0);
1002 vfs_caches_init_early();
1003 sort_main_extable();
1004 trap_init();
1005 mm_init();
1006 poking_init();
1007 ftrace_init();
1008
1009 /* trace_printk can be enabled here */
1010 early_trace_init();
1011
1012 /*
1013 * Set up the scheduler prior starting any interrupts (such as the
1014 * timer interrupt). Full topology setup happens at smp_init()
1015 * time - but meanwhile we still have a functioning scheduler.
1016 */
1017 sched_init();
1018
1019 if (WARN(!irqs_disabled(),
1020 "Interrupts were enabled *very* early, fixing it\n"))
1021 local_irq_disable();
1022 radix_tree_init();
1023 maple_tree_init();
1024
1025 /*
1026 * Set up housekeeping before setting up workqueues to allow the unbound
1027 * workqueue to take non-housekeeping into account.
1028 */
1029 housekeeping_init();
1030
1031 /*
1032 * Allow workqueue creation and work item queueing/cancelling
1033 * early. Work item execution depends on kthreads and starts after
1034 * workqueue_init().
1035 */
1036 workqueue_init_early();
1037
1038 rcu_init();
1039
1040 /* Trace events are available after this */
1041 trace_init();
1042
1043 if (initcall_debug)
1044 initcall_debug_enable();
1045
1046 context_tracking_init();
1047 /* init some links before init_ISA_irqs() */
1048 early_irq_init();
1049 init_IRQ();
1050 tick_init();
1051 rcu_init_nohz();
1052 init_timers();
1053 srcu_init();
1054 hrtimers_init();
1055 softirq_init();
1056 timekeeping_init();
1057 time_init();
1058
1059 /* This must be after timekeeping is initialized */
1060 random_init();
1061
1062 /* These make use of the fully initialized rng */
1063 kfence_init();
1064 boot_init_stack_canary();
1065
1066 perf_event_init();
1067 profile_init();
1068 call_function_init();
1069 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1070
1071 early_boot_irqs_disabled = false;
1072 local_irq_enable();
1073
1074 kmem_cache_init_late();
1075
1076 /*
1077 * HACK ALERT! This is early. We're enabling the console before
1078 * we've done PCI setups etc, and console_init() must be aware of
1079 * this. But we do want output early, in case something goes wrong.
1080 */
1081 console_init();
1082 if (panic_later)
1083 panic("Too many boot %s vars at `%s'", panic_later,
1084 panic_param);
1085
1086 lockdep_init();
1087
1088 /*
1089 * Need to run this when irqs are enabled, because it wants
1090 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1091 * too:
1092 */
1093 locking_selftest();
1094
1095 /*
1096 * This needs to be called before any devices perform DMA
1097 * operations that might use the SWIOTLB bounce buffers. It will
1098 * mark the bounce buffers as decrypted so that their usage will
1099 * not cause "plain-text" data to be decrypted when accessed.
1100 */
1101 mem_encrypt_init();
1102
1103#ifdef CONFIG_BLK_DEV_INITRD
1104 if (initrd_start && !initrd_below_start_ok &&
1105 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1106 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1107 page_to_pfn(virt_to_page((void *)initrd_start)),
1108 min_low_pfn);
1109 initrd_start = 0;
1110 }
1111#endif
1112 setup_per_cpu_pageset();
1113 numa_policy_init();
1114 acpi_early_init();
1115 if (late_time_init)
1116 late_time_init();
1117 sched_clock_init();
1118 calibrate_delay();
1119 pid_idr_init();
1120 anon_vma_init();
1121#ifdef CONFIG_X86
1122 if (efi_enabled(EFI_RUNTIME_SERVICES))
1123 efi_enter_virtual_mode();
1124#endif
1125 thread_stack_cache_init();
1126 cred_init();
1127 fork_init();
1128 proc_caches_init();
1129 uts_ns_init();
1130 key_init();
1131 security_init();
1132 dbg_late_init();
1133 net_ns_init();
1134 vfs_caches_init();
1135 pagecache_init();
1136 signals_init();
1137 seq_file_init();
1138 proc_root_init();
1139 nsfs_init();
1140 cpuset_init();
1141 cgroup_init();
1142 taskstats_init_early();
1143 delayacct_init();
1144
1145 check_bugs();
1146
1147 acpi_subsystem_init();
1148 arch_post_acpi_subsys_init();
1149 kcsan_init();
1150
1151 /* Do the rest non-__init'ed, we're now alive */
1152 arch_call_rest_init();
1153
1154 prevent_tail_call_optimization();
1155}
1156
1157/* Call all constructor functions linked into the kernel. */
1158static void __init do_ctors(void)
1159{
1160/*
1161 * For UML, the constructors have already been called by the
1162 * normal setup code as it's just a normal ELF binary, so we
1163 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1164 * even on UML for modules.
1165 */
1166#if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1167 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1168
1169 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1170 (*fn)();
1171#endif
1172}
1173
1174#ifdef CONFIG_KALLSYMS
1175struct blacklist_entry {
1176 struct list_head next;
1177 char *buf;
1178};
1179
1180static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1181
1182static int __init initcall_blacklist(char *str)
1183{
1184 char *str_entry;
1185 struct blacklist_entry *entry;
1186
1187 /* str argument is a comma-separated list of functions */
1188 do {
1189 str_entry = strsep(&str, ",");
1190 if (str_entry) {
1191 pr_debug("blacklisting initcall %s\n", str_entry);
1192 entry = memblock_alloc(sizeof(*entry),
1193 SMP_CACHE_BYTES);
1194 if (!entry)
1195 panic("%s: Failed to allocate %zu bytes\n",
1196 __func__, sizeof(*entry));
1197 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1198 SMP_CACHE_BYTES);
1199 if (!entry->buf)
1200 panic("%s: Failed to allocate %zu bytes\n",
1201 __func__, strlen(str_entry) + 1);
1202 strcpy(entry->buf, str_entry);
1203 list_add(&entry->next, &blacklisted_initcalls);
1204 }
1205 } while (str_entry);
1206
1207 return 1;
1208}
1209
1210static bool __init_or_module initcall_blacklisted(initcall_t fn)
1211{
1212 struct blacklist_entry *entry;
1213 char fn_name[KSYM_SYMBOL_LEN];
1214 unsigned long addr;
1215
1216 if (list_empty(&blacklisted_initcalls))
1217 return false;
1218
1219 addr = (unsigned long) dereference_function_descriptor(fn);
1220 sprint_symbol_no_offset(fn_name, addr);
1221
1222 /*
1223 * fn will be "function_name [module_name]" where [module_name] is not
1224 * displayed for built-in init functions. Strip off the [module_name].
1225 */
1226 strreplace(fn_name, ' ', '\0');
1227
1228 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1229 if (!strcmp(fn_name, entry->buf)) {
1230 pr_debug("initcall %s blacklisted\n", fn_name);
1231 return true;
1232 }
1233 }
1234
1235 return false;
1236}
1237#else
1238static int __init initcall_blacklist(char *str)
1239{
1240 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1241 return 0;
1242}
1243
1244static bool __init_or_module initcall_blacklisted(initcall_t fn)
1245{
1246 return false;
1247}
1248#endif
1249__setup("initcall_blacklist=", initcall_blacklist);
1250
1251static __init_or_module void
1252trace_initcall_start_cb(void *data, initcall_t fn)
1253{
1254 ktime_t *calltime = data;
1255
1256 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1257 *calltime = ktime_get();
1258}
1259
1260static __init_or_module void
1261trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1262{
1263 ktime_t rettime, *calltime = data;
1264
1265 rettime = ktime_get();
1266 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1267 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1268}
1269
1270static ktime_t initcall_calltime;
1271
1272#ifdef TRACEPOINTS_ENABLED
1273static void __init initcall_debug_enable(void)
1274{
1275 int ret;
1276
1277 ret = register_trace_initcall_start(trace_initcall_start_cb,
1278 &initcall_calltime);
1279 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1280 &initcall_calltime);
1281 WARN(ret, "Failed to register initcall tracepoints\n");
1282}
1283# define do_trace_initcall_start trace_initcall_start
1284# define do_trace_initcall_finish trace_initcall_finish
1285#else
1286static inline void do_trace_initcall_start(initcall_t fn)
1287{
1288 if (!initcall_debug)
1289 return;
1290 trace_initcall_start_cb(&initcall_calltime, fn);
1291}
1292static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1293{
1294 if (!initcall_debug)
1295 return;
1296 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1297}
1298#endif /* !TRACEPOINTS_ENABLED */
1299
1300int __init_or_module do_one_initcall(initcall_t fn)
1301{
1302 int count = preempt_count();
1303 char msgbuf[64];
1304 int ret;
1305
1306 if (initcall_blacklisted(fn))
1307 return -EPERM;
1308
1309 do_trace_initcall_start(fn);
1310 ret = fn();
1311 do_trace_initcall_finish(fn, ret);
1312
1313 msgbuf[0] = 0;
1314
1315 if (preempt_count() != count) {
1316 sprintf(msgbuf, "preemption imbalance ");
1317 preempt_count_set(count);
1318 }
1319 if (irqs_disabled()) {
1320 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1321 local_irq_enable();
1322 }
1323 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1324
1325 add_latent_entropy();
1326 return ret;
1327}
1328
1329
1330extern initcall_entry_t __initcall_start[];
1331extern initcall_entry_t __initcall0_start[];
1332extern initcall_entry_t __initcall1_start[];
1333extern initcall_entry_t __initcall2_start[];
1334extern initcall_entry_t __initcall3_start[];
1335extern initcall_entry_t __initcall4_start[];
1336extern initcall_entry_t __initcall5_start[];
1337extern initcall_entry_t __initcall6_start[];
1338extern initcall_entry_t __initcall7_start[];
1339extern initcall_entry_t __initcall_end[];
1340
1341static initcall_entry_t *initcall_levels[] __initdata = {
1342 __initcall0_start,
1343 __initcall1_start,
1344 __initcall2_start,
1345 __initcall3_start,
1346 __initcall4_start,
1347 __initcall5_start,
1348 __initcall6_start,
1349 __initcall7_start,
1350 __initcall_end,
1351};
1352
1353/* Keep these in sync with initcalls in include/linux/init.h */
1354static const char *initcall_level_names[] __initdata = {
1355 "pure",
1356 "core",
1357 "postcore",
1358 "arch",
1359 "subsys",
1360 "fs",
1361 "device",
1362 "late",
1363};
1364
1365static int __init ignore_unknown_bootoption(char *param, char *val,
1366 const char *unused, void *arg)
1367{
1368 return 0;
1369}
1370
1371static void __init do_initcall_level(int level, char *command_line)
1372{
1373 initcall_entry_t *fn;
1374
1375 parse_args(initcall_level_names[level],
1376 command_line, __start___param,
1377 __stop___param - __start___param,
1378 level, level,
1379 NULL, ignore_unknown_bootoption);
1380
1381 trace_initcall_level(initcall_level_names[level]);
1382 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1383 do_one_initcall(initcall_from_entry(fn));
1384}
1385
1386static void __init do_initcalls(void)
1387{
1388 int level;
1389 size_t len = saved_command_line_len + 1;
1390 char *command_line;
1391
1392 command_line = kzalloc(len, GFP_KERNEL);
1393 if (!command_line)
1394 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1395
1396 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1397 /* Parser modifies command_line, restore it each time */
1398 strcpy(command_line, saved_command_line);
1399 do_initcall_level(level, command_line);
1400 }
1401
1402 kfree(command_line);
1403}
1404
1405/*
1406 * Ok, the machine is now initialized. None of the devices
1407 * have been touched yet, but the CPU subsystem is up and
1408 * running, and memory and process management works.
1409 *
1410 * Now we can finally start doing some real work..
1411 */
1412static void __init do_basic_setup(void)
1413{
1414 cpuset_init_smp();
1415 driver_init();
1416 init_irq_proc();
1417 do_ctors();
1418 do_initcalls();
1419}
1420
1421static void __init do_pre_smp_initcalls(void)
1422{
1423 initcall_entry_t *fn;
1424
1425 trace_initcall_level("early");
1426 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1427 do_one_initcall(initcall_from_entry(fn));
1428}
1429
1430static int run_init_process(const char *init_filename)
1431{
1432 const char *const *p;
1433
1434 argv_init[0] = init_filename;
1435 pr_info("Run %s as init process\n", init_filename);
1436 pr_debug(" with arguments:\n");
1437 for (p = argv_init; *p; p++)
1438 pr_debug(" %s\n", *p);
1439 pr_debug(" with environment:\n");
1440 for (p = envp_init; *p; p++)
1441 pr_debug(" %s\n", *p);
1442 return kernel_execve(init_filename, argv_init, envp_init);
1443}
1444
1445static int try_to_run_init_process(const char *init_filename)
1446{
1447 int ret;
1448
1449 ret = run_init_process(init_filename);
1450
1451 if (ret && ret != -ENOENT) {
1452 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1453 init_filename, ret);
1454 }
1455
1456 return ret;
1457}
1458
1459static noinline void __init kernel_init_freeable(void);
1460
1461#if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1462bool rodata_enabled __ro_after_init = true;
1463
1464#ifndef arch_parse_debug_rodata
1465static inline bool arch_parse_debug_rodata(char *str) { return false; }
1466#endif
1467
1468static int __init set_debug_rodata(char *str)
1469{
1470 if (arch_parse_debug_rodata(str))
1471 return 0;
1472
1473 if (str && !strcmp(str, "on"))
1474 rodata_enabled = true;
1475 else if (str && !strcmp(str, "off"))
1476 rodata_enabled = false;
1477 else
1478 pr_warn("Invalid option string for rodata: '%s'\n", str);
1479 return 0;
1480}
1481early_param("rodata", set_debug_rodata);
1482#endif
1483
1484#ifdef CONFIG_STRICT_KERNEL_RWX
1485static void mark_readonly(void)
1486{
1487 if (rodata_enabled) {
1488 /*
1489 * load_module() results in W+X mappings, which are cleaned
1490 * up with call_rcu(). Let's make sure that queued work is
1491 * flushed so that we don't hit false positives looking for
1492 * insecure pages which are W+X.
1493 */
1494 rcu_barrier();
1495 mark_rodata_ro();
1496 rodata_test();
1497 } else
1498 pr_info("Kernel memory protection disabled.\n");
1499}
1500#elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1501static inline void mark_readonly(void)
1502{
1503 pr_warn("Kernel memory protection not selected by kernel config.\n");
1504}
1505#else
1506static inline void mark_readonly(void)
1507{
1508 pr_warn("This architecture does not have kernel memory protection.\n");
1509}
1510#endif
1511
1512void __weak free_initmem(void)
1513{
1514 free_initmem_default(POISON_FREE_INITMEM);
1515}
1516
1517static int __ref kernel_init(void *unused)
1518{
1519 int ret;
1520
1521 /*
1522 * Wait until kthreadd is all set-up.
1523 */
1524 wait_for_completion(&kthreadd_done);
1525
1526 kernel_init_freeable();
1527 /* need to finish all async __init code before freeing the memory */
1528 async_synchronize_full();
1529
1530 system_state = SYSTEM_FREEING_INITMEM;
1531 kprobe_free_init_mem();
1532 ftrace_free_init_mem();
1533 kgdb_free_init_mem();
1534 exit_boot_config();
1535 free_initmem();
1536 mark_readonly();
1537
1538 /*
1539 * Kernel mappings are now finalized - update the userspace page-table
1540 * to finalize PTI.
1541 */
1542 pti_finalize();
1543
1544 system_state = SYSTEM_RUNNING;
1545 numa_default_policy();
1546
1547 rcu_end_inkernel_boot();
1548
1549 do_sysctl_args();
1550
1551 if (ramdisk_execute_command) {
1552 ret = run_init_process(ramdisk_execute_command);
1553 if (!ret)
1554 return 0;
1555 pr_err("Failed to execute %s (error %d)\n",
1556 ramdisk_execute_command, ret);
1557 }
1558
1559 /*
1560 * We try each of these until one succeeds.
1561 *
1562 * The Bourne shell can be used instead of init if we are
1563 * trying to recover a really broken machine.
1564 */
1565 if (execute_command) {
1566 ret = run_init_process(execute_command);
1567 if (!ret)
1568 return 0;
1569 panic("Requested init %s failed (error %d).",
1570 execute_command, ret);
1571 }
1572
1573 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1574 ret = run_init_process(CONFIG_DEFAULT_INIT);
1575 if (ret)
1576 pr_err("Default init %s failed (error %d)\n",
1577 CONFIG_DEFAULT_INIT, ret);
1578 else
1579 return 0;
1580 }
1581
1582 if (!try_to_run_init_process("/sbin/init") ||
1583 !try_to_run_init_process("/etc/init") ||
1584 !try_to_run_init_process("/bin/init") ||
1585 !try_to_run_init_process("/bin/sh"))
1586 return 0;
1587
1588 panic("No working init found. Try passing init= option to kernel. "
1589 "See Linux Documentation/admin-guide/init.rst for guidance.");
1590}
1591
1592/* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1593void __init console_on_rootfs(void)
1594{
1595 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1596
1597 if (IS_ERR(file)) {
1598 pr_err("Warning: unable to open an initial console.\n");
1599 return;
1600 }
1601 init_dup(file);
1602 init_dup(file);
1603 init_dup(file);
1604 fput(file);
1605}
1606
1607static noinline void __init kernel_init_freeable(void)
1608{
1609 /* Now the scheduler is fully set up and can do blocking allocations */
1610 gfp_allowed_mask = __GFP_BITS_MASK;
1611
1612 /*
1613 * init can allocate pages on any node
1614 */
1615 set_mems_allowed(node_states[N_MEMORY]);
1616
1617 cad_pid = get_pid(task_pid(current));
1618
1619 smp_prepare_cpus(setup_max_cpus);
1620
1621 workqueue_init();
1622
1623 init_mm_internals();
1624
1625 rcu_init_tasks_generic();
1626 do_pre_smp_initcalls();
1627 lockup_detector_init();
1628
1629 smp_init();
1630 sched_init_smp();
1631
1632 padata_init();
1633 page_alloc_init_late();
1634 /* Initialize page ext after all struct pages are initialized. */
1635 if (deferred_struct_pages)
1636 page_ext_init();
1637
1638 do_basic_setup();
1639
1640 kunit_run_all_tests();
1641
1642 wait_for_initramfs();
1643 console_on_rootfs();
1644
1645 /*
1646 * check if there is an early userspace init. If yes, let it do all
1647 * the work
1648 */
1649 if (init_eaccess(ramdisk_execute_command) != 0) {
1650 ramdisk_execute_command = NULL;
1651 prepare_namespace();
1652 }
1653
1654 /*
1655 * Ok, we have completed the initial bootup, and
1656 * we're essentially up and running. Get rid of the
1657 * initmem segments and start the user-mode stuff..
1658 *
1659 * rootfs is available now, try loading the public keys
1660 * and default modules
1661 */
1662
1663 integrity_load_keys();
1664}