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