1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/init/main.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
11 */
12
13 #define DEBUG /* Enable initcall_debug */
14
15 #include <linux/types.h>
16 #include <linux/export.h>
17 #include <linux/extable.h>
18 #include <linux/module.h>
19 #include <linux/proc_fs.h>
20 #include <linux/binfmts.h>
21 #include <linux/kernel.h>
22 #include <linux/syscalls.h>
23 #include <linux/stackprotector.h>
24 #include <linux/string.h>
25 #include <linux/ctype.h>
26 #include <linux/delay.h>
27 #include <linux/ioport.h>
28 #include <linux/init.h>
29 #include <linux/initrd.h>
30 #include <linux/memblock.h>
31 #include <linux/acpi.h>
32 #include <linux/bootconfig.h>
33 #include <linux/console.h>
34 #include <linux/nmi.h>
35 #include <linux/percpu.h>
36 #include <linux/kmod.h>
37 #include <linux/kprobes.h>
38 #include <linux/kmsan.h>
39 #include <linux/ksysfs.h>
40 #include <linux/vmalloc.h>
41 #include <linux/kernel_stat.h>
42 #include <linux/start_kernel.h>
43 #include <linux/security.h>
44 #include <linux/smp.h>
45 #include <linux/profile.h>
46 #include <linux/kfence.h>
47 #include <linux/rcupdate.h>
48 #include <linux/srcu.h>
49 #include <linux/moduleparam.h>
50 #include <linux/kallsyms.h>
51 #include <linux/buildid.h>
52 #include <linux/writeback.h>
53 #include <linux/cpu.h>
54 #include <linux/cpuset.h>
55 #include <linux/memcontrol.h>
56 #include <linux/cgroup.h>
57 #include <linux/tick.h>
58 #include <linux/sched/isolation.h>
59 #include <linux/interrupt.h>
60 #include <linux/taskstats_kern.h>
61 #include <linux/delayacct.h>
62 #include <linux/unistd.h>
63 #include <linux/utsname.h>
64 #include <linux/rmap.h>
65 #include <linux/mempolicy.h>
66 #include <linux/key.h>
67 #include <linux/debug_locks.h>
68 #include <linux/debugobjects.h>
69 #include <linux/lockdep.h>
70 #include <linux/kmemleak.h>
71 #include <linux/padata.h>
72 #include <linux/pid_namespace.h>
73 #include <linux/device/driver.h>
74 #include <linux/kthread.h>
75 #include <linux/sched.h>
76 #include <linux/sched/init.h>
77 #include <linux/signal.h>
78 #include <linux/idr.h>
79 #include <linux/kgdb.h>
80 #include <linux/ftrace.h>
81 #include <linux/async.h>
82 #include <linux/shmem_fs.h>
83 #include <linux/slab.h>
84 #include <linux/perf_event.h>
85 #include <linux/ptrace.h>
86 #include <linux/pti.h>
87 #include <linux/blkdev.h>
88 #include <linux/sched/clock.h>
89 #include <linux/sched/task.h>
90 #include <linux/sched/task_stack.h>
91 #include <linux/context_tracking.h>
92 #include <linux/random.h>
93 #include <linux/moduleloader.h>
94 #include <linux/list.h>
95 #include <linux/integrity.h>
96 #include <linux/proc_ns.h>
97 #include <linux/io.h>
98 #include <linux/cache.h>
99 #include <linux/rodata_test.h>
100 #include <linux/jump_label.h>
101 #include <linux/kcsan.h>
102 #include <linux/init_syscalls.h>
103 #include <linux/stackdepot.h>
104 #include <linux/randomize_kstack.h>
105 #include <linux/pidfs.h>
106 #include <linux/fs_struct.h>
107 #include <linux/ptdump.h>
108 #include <linux/time_namespace.h>
109 #include <linux/unaligned.h>
110 #include <linux/vdso_datastore.h>
111 #include <net/net_namespace.h>
112
113 #include <asm/io.h>
114 #include <asm/setup.h>
115 #include <asm/sections.h>
116 #include <asm/cacheflush.h>
117
118 #define CREATE_TRACE_POINTS
119 #include <trace/events/initcall.h>
120
121 #include <kunit/test.h>
122
123 static int kernel_init(void *);
124
125 /*
126 * Debug helper: via this flag we know that we are in 'early bootup code'
127 * where only the boot processor is running with IRQ disabled. This means
128 * two things - IRQ must not be enabled before the flag is cleared and some
129 * operations which are not allowed with IRQ disabled are allowed while the
130 * flag is set.
131 */
132 bool early_boot_irqs_disabled __read_mostly;
133
134 enum system_states system_state __read_mostly;
135 EXPORT_SYMBOL(system_state);
136
137 /*
138 * Boot command-line arguments
139 */
140 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
141 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
142
143 /* Default late time init is NULL. archs can override this later. */
144 void (*__initdata late_time_init)(void);
145
146 /* Untouched command line saved by arch-specific code. */
147 char __initdata boot_command_line[COMMAND_LINE_SIZE];
148 /* Untouched saved command line (eg. for /proc) */
149 char *saved_command_line __ro_after_init;
150 unsigned int saved_command_line_len __ro_after_init;
151 /* Command line for parameter parsing */
152 static char *static_command_line;
153 /* Untouched extra command line */
154 static char *extra_command_line;
155 /* Extra init arguments */
156 static char *extra_init_args;
157
158 #ifdef CONFIG_BOOT_CONFIG
159 /* Is bootconfig on command line? */
160 static bool bootconfig_found;
161 static size_t initargs_offs;
162 #else
163 # define bootconfig_found false
164 # define initargs_offs 0
165 #endif
166
167 static char *execute_command;
168 static char *ramdisk_execute_command = "/init";
169 static bool __initdata ramdisk_execute_command_set;
170
171 /*
172 * Used to generate warnings if static_key manipulation functions are used
173 * before jump_label_init is called.
174 */
175 bool static_key_initialized __read_mostly;
176 EXPORT_SYMBOL_GPL(static_key_initialized);
177
178 /*
179 * If set, this is an indication to the drivers that reset the underlying
180 * device before going ahead with the initialization otherwise driver might
181 * rely on the BIOS and skip the reset operation.
182 *
183 * This is useful if kernel is booting in an unreliable environment.
184 * For ex. kdump situation where previous kernel has crashed, BIOS has been
185 * skipped and devices will be in unknown state.
186 */
187 unsigned int reset_devices;
188 EXPORT_SYMBOL(reset_devices);
189
set_reset_devices(char * str)190 static int __init set_reset_devices(char *str)
191 {
192 reset_devices = 1;
193 return 1;
194 }
195
196 __setup("reset_devices", set_reset_devices);
197
198 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
199 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
200 static const char *panic_later, *panic_param;
201
obsolete_checksetup(char * line)202 static 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 */
235 unsigned long loops_per_jiffy = (1<<12);
236 EXPORT_SYMBOL(loops_per_jiffy);
237
debug_kernel(char * str)238 static int __init debug_kernel(char *str)
239 {
240 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
241 return 0;
242 }
243
quiet_kernel(char * str)244 static int __init quiet_kernel(char *str)
245 {
246 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
247 return 0;
248 }
249
250 early_param("debug", debug_kernel);
251 early_param("quiet", quiet_kernel);
252
loglevel(char * str)253 static 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
270 early_param("loglevel", loglevel);
271
272 #ifdef CONFIG_BLK_DEV_INITRD
get_boot_config_from_initrd(size_t * _size)273 static void * __init get_boot_config_from_initrd(size_t *_size)
274 {
275 u32 size, csum;
276 char *data;
277 u8 *hdr;
278 int i;
279
280 if (!initrd_end || initrd_end < initrd_start ||
281 initrd_end - initrd_start < BOOTCONFIG_MAGIC_LEN + 8)
282 return NULL;
283
284 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
285 /*
286 * Since Grub may align the size of initrd to 4, we must
287 * check the preceding 3 bytes as well.
288 */
289 for (i = 0; i < 4; i++) {
290 if (!memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
291 goto found;
292 data--;
293 }
294 return NULL;
295
296 found:
297 hdr = (u8 *)(data - 8);
298 if ((unsigned long)hdr < initrd_start)
299 return NULL;
300
301 size = get_unaligned_le32(hdr);
302 csum = get_unaligned_le32(hdr + 4);
303
304 if (size > XBC_DATA_MAX) {
305 pr_err("bootconfig size %u is greater than max size %d\n",
306 size, XBC_DATA_MAX);
307 return NULL;
308 }
309
310 if (size > ((unsigned long)hdr - initrd_start)) {
311 pr_err("bootconfig size %u is greater than initrd size %lu\n",
312 size, initrd_end - initrd_start);
313 return NULL;
314 }
315
316 data = ((void *)hdr) - size;
317
318 if (xbc_calc_checksum(data, size) != csum) {
319 pr_err("bootconfig checksum failed\n");
320 return NULL;
321 }
322
323 /* Remove bootconfig from initramfs/initrd */
324 initrd_end = (unsigned long)data;
325 if (_size)
326 *_size = size;
327
328 return data;
329 }
330 #else
get_boot_config_from_initrd(size_t * _size)331 static void * __init get_boot_config_from_initrd(size_t *_size)
332 {
333 return NULL;
334 }
335 #endif
336
337 #ifdef CONFIG_BOOT_CONFIG
338
339 /* Make an extra command line under given key word */
xbc_make_cmdline(const char * key)340 static char * __init xbc_make_cmdline(const char *key)
341 {
342 struct xbc_node *root;
343 char *new_cmdline;
344 int ret, len = 0;
345
346 root = xbc_find_node(key);
347 if (!root)
348 return NULL;
349
350 /* Count required buffer size */
351 len = xbc_snprint_cmdline(NULL, 0, root);
352 if (len <= 0)
353 return NULL;
354
355 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
356 if (!new_cmdline) {
357 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
358 return NULL;
359 }
360
361 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
362 if (ret < 0 || ret > len) {
363 pr_err("Failed to print extra kernel cmdline.\n");
364 memblock_free(new_cmdline, len + 1);
365 return NULL;
366 }
367
368 return new_cmdline;
369 }
370
warn_bootconfig(char * str)371 static int __init warn_bootconfig(char *str)
372 {
373 /* The 'bootconfig' option is handled by setup_boot_config(). */
374 return 0;
375 }
376
setup_boot_config(void)377 static void __init setup_boot_config(void)
378 {
379 const char *msg, *data;
380 int pos, ret, offs;
381 size_t size;
382 bool from_embedded = false;
383
384 /* Cut out the bootconfig data even if we have no bootconfig option */
385 data = get_boot_config_from_initrd(&size);
386 /* If there is no bootconfig in initrd, try embedded one. */
387 if (!data) {
388 data = xbc_get_embedded_bootconfig(&size);
389 from_embedded = true;
390 }
391
392 bootconfig_found = bootconfig_cmdline_requested(boot_command_line, &offs);
393 if (!(bootconfig_found || IS_ENABLED(CONFIG_BOOT_CONFIG_FORCE)))
394 return;
395
396 /* Offset of the init arguments after a "--", located by the helper. */
397 initargs_offs = offs;
398
399 if (!data) {
400 /* If user intended to use bootconfig, show an error level message */
401 if (bootconfig_found)
402 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
403 else
404 pr_info("No bootconfig data provided, so skipping bootconfig");
405 return;
406 }
407
408 ret = xbc_init(data, size, &msg, &pos);
409 if (ret < 0) {
410 if (pos < 0)
411 pr_err("Failed to init bootconfig: %s.\n", msg);
412 else
413 pr_err("Failed to parse bootconfig: %s at %d.\n",
414 msg, pos);
415 } else {
416 xbc_get_info(&ret, NULL);
417 pr_info("Load bootconfig: %ld bytes %d nodes\n", (long)size, ret);
418 /*
419 * keys starting with "kernel." are passed via cmdline. When
420 * this bootconfig came from the embedded source and
421 * setup_arch() already prepended the rendered "kernel" subtree
422 * to boot_command_line, rendering again here would duplicate
423 * the keys in saved_command_line and make accumulating handlers
424 * (console=, earlycon=, ...) re-register the same value. Skip
425 * only when the prepend really happened.
426 *
427 * On arches that do not select ARCH_SUPPORTS_CMDLINE_FROM_BOOTCONFIG,
428 * CONFIG_CMDLINE_FROM_BOOTCONFIG is unselectable and
429 * xbc_embedded_cmdline_applied() collapses to a stub returning
430 * false, so this path still runs and the embedded "kernel"
431 * keys reach the cmdline via the runtime parser exactly as
432 * before this series.
433 */
434 if (!from_embedded || !xbc_embedded_cmdline_applied())
435 extra_command_line = xbc_make_cmdline("kernel");
436 /* Also, "init." keys are init arguments */
437 extra_init_args = xbc_make_cmdline("init");
438 }
439 return;
440 }
441
exit_boot_config(void)442 static void __init exit_boot_config(void)
443 {
444 xbc_exit();
445 }
446
447 #else /* !CONFIG_BOOT_CONFIG */
448
setup_boot_config(void)449 static void __init setup_boot_config(void)
450 {
451 /* Remove bootconfig data from initrd */
452 get_boot_config_from_initrd(NULL);
453 }
454
warn_bootconfig(char * str)455 static int __init warn_bootconfig(char *str)
456 {
457 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOT_CONFIG is not set.\n");
458 return 0;
459 }
460
461 #define exit_boot_config() do {} while (0)
462
463 #endif /* CONFIG_BOOT_CONFIG */
464
465 early_param("bootconfig", warn_bootconfig);
466
cmdline_has_extra_options(void)467 bool __init cmdline_has_extra_options(void)
468 {
469 return extra_command_line || extra_init_args;
470 }
471
472 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val)473 static void __init repair_env_string(char *param, char *val)
474 {
475 if (val) {
476 /* param=val or param="val"? */
477 if (val == param+strlen(param)+1)
478 val[-1] = '=';
479 else if (val == param+strlen(param)+2) {
480 val[-2] = '=';
481 memmove(val-1, val, strlen(val)+1);
482 } else
483 BUG();
484 }
485 }
486
487 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)488 static int __init set_init_arg(char *param, char *val,
489 const char *unused, void *arg)
490 {
491 unsigned int i;
492
493 if (panic_later)
494 return 0;
495
496 repair_env_string(param, val);
497
498 for (i = 0; argv_init[i]; i++) {
499 if (i == MAX_INIT_ARGS) {
500 panic_later = "init";
501 panic_param = param;
502 return 0;
503 }
504 }
505 argv_init[i] = param;
506 return 0;
507 }
508
509 /*
510 * Unknown boot options get handed to init, unless they look like
511 * unused parameters (modprobe will find them in /proc/cmdline).
512 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)513 static int __init unknown_bootoption(char *param, char *val,
514 const char *unused, void *arg)
515 {
516 size_t len = strlen(param);
517 /*
518 * Well-known bootloader identifiers:
519 * 1. LILO/Grub pass "BOOT_IMAGE=...";
520 * 2. kexec/kdump (kexec-tools) pass "kexec".
521 */
522 const char *bootloader[] = { "BOOT_IMAGE=", "kexec", NULL };
523
524 /* Handle params aliased to sysctls */
525 if (sysctl_is_alias(param))
526 return 0;
527
528 repair_env_string(param, val);
529
530 /* Handle bootloader identifier */
531 for (int i = 0; bootloader[i]; i++) {
532 if (strstarts(param, bootloader[i]))
533 return 0;
534 }
535
536 /* Handle obsolete-style parameters */
537 if (obsolete_checksetup(param))
538 return 0;
539
540 /* Unused module parameter. */
541 if (strnchr(param, len, '.'))
542 return 0;
543
544 if (panic_later)
545 return 0;
546
547 if (val) {
548 /* Environment option */
549 unsigned int i;
550 for (i = 0; envp_init[i]; i++) {
551 if (i == MAX_INIT_ENVS) {
552 panic_later = "env";
553 panic_param = param;
554 }
555 if (!strncmp(param, envp_init[i], len+1))
556 break;
557 }
558 envp_init[i] = param;
559 } else {
560 /* Command line option */
561 unsigned int i;
562 for (i = 0; argv_init[i]; i++) {
563 if (i == MAX_INIT_ARGS) {
564 panic_later = "init";
565 panic_param = param;
566 }
567 }
568 argv_init[i] = param;
569 }
570 return 0;
571 }
572
init_setup(char * str)573 static int __init init_setup(char *str)
574 {
575 unsigned int i;
576
577 execute_command = str;
578 /*
579 * In case LILO is going to boot us with default command line,
580 * it prepends "auto" before the whole cmdline which makes
581 * the shell think it should execute a script with such name.
582 * So we ignore all arguments entered _before_ init=... [MJ]
583 */
584 for (i = 1; i < MAX_INIT_ARGS; i++)
585 argv_init[i] = NULL;
586 return 1;
587 }
588 __setup("init=", init_setup);
589
rdinit_setup(char * str)590 static int __init rdinit_setup(char *str)
591 {
592 unsigned int i;
593
594 ramdisk_execute_command = str;
595 ramdisk_execute_command_set = true;
596 /* See "auto" comment in init_setup */
597 for (i = 1; i < MAX_INIT_ARGS; i++)
598 argv_init[i] = NULL;
599 return 1;
600 }
601 __setup("rdinit=", rdinit_setup);
602
603 #ifndef CONFIG_SMP
setup_nr_cpu_ids(void)604 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)605 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
606 #endif
607
608 /*
609 * We need to store the untouched command line for future reference.
610 * We also need to store the touched command line since the parameter
611 * parsing is performed in place, and we should allow a component to
612 * store reference of name/value for future reference.
613 */
setup_command_line(char * command_line)614 static void __init setup_command_line(char *command_line)
615 {
616 size_t len, xlen = 0, ilen = 0;
617
618 if (extra_command_line)
619 xlen = strlen(extra_command_line);
620 if (extra_init_args) {
621 extra_init_args = strim(extra_init_args); /* remove trailing space */
622 ilen = strlen(extra_init_args) + 4; /* for " -- " */
623 }
624
625 len = xlen + strlen(boot_command_line) + ilen + 1;
626
627 saved_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
628
629 len = xlen + strlen(command_line) + 1;
630
631 static_command_line = memblock_alloc_or_panic(len, SMP_CACHE_BYTES);
632
633 if (xlen) {
634 /*
635 * We have to put extra_command_line before boot command
636 * lines because there could be dashes (separator of init
637 * command line) in the command lines.
638 */
639 strcpy(saved_command_line, extra_command_line);
640 strcpy(static_command_line, extra_command_line);
641 }
642 strcpy(saved_command_line + xlen, boot_command_line);
643 strcpy(static_command_line + xlen, command_line);
644
645 if (ilen) {
646 /*
647 * Append supplemental init boot args to saved_command_line
648 * so that user can check what command line options passed
649 * to init.
650 * The order should always be
651 * " -- "[bootconfig init-param][cmdline init-param]
652 */
653 if (initargs_offs) {
654 len = xlen + initargs_offs;
655 strcpy(saved_command_line + len, extra_init_args);
656 len += ilen - 4; /* strlen(extra_init_args) */
657 strcpy(saved_command_line + len,
658 boot_command_line + initargs_offs - 1);
659 } else {
660 len = strlen(saved_command_line);
661 strcpy(saved_command_line + len, " -- ");
662 len += 4;
663 strcpy(saved_command_line + len, extra_init_args);
664 }
665 }
666
667 saved_command_line_len = strlen(saved_command_line);
668 }
669
670 /*
671 * We need to finalize in a non-__init function or else race conditions
672 * between the root thread and the init thread may cause start_kernel to
673 * be reaped by free_initmem before the root thread has proceeded to
674 * cpu_idle.
675 *
676 * gcc-3.4 accidentally inlines this function, so use noinline.
677 */
678
679 static __initdata DECLARE_COMPLETION(kthreadd_done);
680
rest_init(void)681 static noinline void __ref __noreturn rest_init(void)
682 {
683 struct kernel_clone_args init_args = {
684 .flags = (CLONE_VM | CLONE_UNTRACED),
685 .fn = kernel_init,
686 .fn_arg = NULL,
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 = kernel_clone(&init_args);
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 if (p->setup_func(val) != 0)
744 pr_warn("Malformed early option '%s'\n", param);
745 }
746 }
747 /* We accept everything at this stage. */
748 return 0;
749 }
750
parse_early_options(char * cmdline)751 void __init parse_early_options(char *cmdline)
752 {
753 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
754 do_early_param);
755 }
756
757 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)758 void __init parse_early_param(void)
759 {
760 static int done __initdata;
761 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
762
763 if (done)
764 return;
765
766 /* All fall through to do_early_param. */
767 strscpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
768 parse_early_options(tmp_cmdline);
769 done = 1;
770 }
771
arch_post_acpi_subsys_init(void)772 void __init __weak arch_post_acpi_subsys_init(void) { }
773
smp_setup_processor_id(void)774 void __init __weak smp_setup_processor_id(void)
775 {
776 }
777
smp_prepare_boot_cpu(void)778 void __init __weak smp_prepare_boot_cpu(void)
779 {
780 }
781
782 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)783 void __init __weak thread_stack_cache_init(void)
784 {
785 }
786 #endif
787
poking_init(void)788 void __init __weak poking_init(void) { }
789
pgtable_cache_init(void)790 void __init __weak pgtable_cache_init(void) { }
791
trap_init(void)792 void __init __weak trap_init(void) { }
793
794 bool initcall_debug;
795 core_param(initcall_debug, initcall_debug, bool, 0644);
796
797 #ifdef TRACEPOINTS_ENABLED
798 static void __init initcall_debug_enable(void);
799 #else
initcall_debug_enable(void)800 static inline void initcall_debug_enable(void)
801 {
802 }
803 #endif
804
805 #ifdef CONFIG_RANDOMIZE_KSTACK_OFFSET
806 DEFINE_STATIC_KEY_MAYBE_RO(CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT,
807 randomize_kstack_offset);
808 DEFINE_PER_CPU(struct rnd_state, kstack_rnd_state);
809
random_kstack_init(void)810 static int __init random_kstack_init(void)
811 {
812 prandom_seed_full_state(&kstack_rnd_state);
813 return 0;
814 }
815 late_initcall(random_kstack_init);
816
early_randomize_kstack_offset(char * buf)817 static int __init early_randomize_kstack_offset(char *buf)
818 {
819 int ret;
820 bool bool_result;
821
822 ret = kstrtobool(buf, &bool_result);
823 if (ret)
824 return ret;
825
826 if (bool_result)
827 static_branch_enable(&randomize_kstack_offset);
828 else
829 static_branch_disable(&randomize_kstack_offset);
830 return 0;
831 }
832 early_param("randomize_kstack_offset", early_randomize_kstack_offset);
833 #endif
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
early_numa_node_init(void)878 static void __init early_numa_node_init(void)
879 {
880 #ifdef CONFIG_USE_PERCPU_NUMA_NODE_ID
881 #ifndef cpu_to_node
882 int cpu;
883
884 /* The early_cpu_to_node() should be ready here. */
885 for_each_possible_cpu(cpu)
886 set_cpu_numa_node(cpu, early_cpu_to_node(cpu));
887 #endif
888 #endif
889 }
890
891 #define KERNEL_CMDLINE_PREFIX "Kernel command line: "
892 #define KERNEL_CMDLINE_PREFIX_LEN (sizeof(KERNEL_CMDLINE_PREFIX) - 1)
893 #define KERNEL_CMDLINE_CONTINUATION " \\"
894 #define KERNEL_CMDLINE_CONTINUATION_LEN (sizeof(KERNEL_CMDLINE_CONTINUATION) - 1)
895
896 #define MIN_CMDLINE_LOG_WRAP_IDEAL_LEN (KERNEL_CMDLINE_PREFIX_LEN + \
897 KERNEL_CMDLINE_CONTINUATION_LEN)
898 #define CMDLINE_LOG_WRAP_IDEAL_LEN (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN > \
899 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN ? \
900 CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN : \
901 MIN_CMDLINE_LOG_WRAP_IDEAL_LEN)
902
903 #define IDEAL_CMDLINE_LEN (CMDLINE_LOG_WRAP_IDEAL_LEN - KERNEL_CMDLINE_PREFIX_LEN)
904 #define IDEAL_CMDLINE_SPLIT_LEN (IDEAL_CMDLINE_LEN - KERNEL_CMDLINE_CONTINUATION_LEN)
905
906 /**
907 * print_kernel_cmdline() - Print the kernel cmdline with wrapping.
908 * @cmdline: The cmdline to print.
909 *
910 * Print the kernel command line, trying to wrap based on the Kconfig knob
911 * CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN.
912 *
913 * Wrapping is based on spaces, ignoring quotes. All lines are prefixed
914 * with "Kernel command line: " and lines that are not the last line have
915 * a " \" suffix added to them. The prefix and suffix count towards the
916 * line length for wrapping purposes. The ideal length will be exceeded
917 * if no appropriate place to wrap is found.
918 *
919 * Example output if CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN is 40:
920 * Kernel command line: loglevel=7 \
921 * Kernel command line: init=/sbin/init \
922 * Kernel command line: root=PARTUUID=8c3efc1a-768b-6642-8d0c-89eb782f19f0/PARTNROFF=1 \
923 * Kernel command line: rootwait ro \
924 * Kernel command line: my_quoted_arg="The \
925 * Kernel command line: quick brown fox \
926 * Kernel command line: jumps over the \
927 * Kernel command line: lazy dog."
928 */
print_kernel_cmdline(const char * cmdline)929 static void __init print_kernel_cmdline(const char *cmdline)
930 {
931 size_t len;
932
933 /* Config option of 0 or anything longer than the max disables wrapping */
934 if (CONFIG_CMDLINE_LOG_WRAP_IDEAL_LEN == 0 ||
935 IDEAL_CMDLINE_LEN >= COMMAND_LINE_SIZE - 1) {
936 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline);
937 return;
938 }
939
940 len = strlen(cmdline);
941 while (len > IDEAL_CMDLINE_LEN) {
942 const char *first_space;
943 const char *prev_cutoff;
944 const char *cutoff;
945 int to_print;
946 size_t used;
947
948 /* Find the last ' ' that wouldn't make the line too long */
949 prev_cutoff = NULL;
950 cutoff = cmdline;
951 while (true) {
952 cutoff = strchr(cutoff + 1, ' ');
953 if (!cutoff || cutoff - cmdline > IDEAL_CMDLINE_SPLIT_LEN)
954 break;
955 prev_cutoff = cutoff;
956 }
957 if (prev_cutoff)
958 cutoff = prev_cutoff;
959 else if (!cutoff)
960 break;
961
962 /* Find the beginning and end of the string of spaces */
963 first_space = cutoff;
964 while (first_space > cmdline && first_space[-1] == ' ')
965 first_space--;
966 to_print = first_space - cmdline;
967 while (*cutoff == ' ')
968 cutoff++;
969 used = cutoff - cmdline;
970
971 /* If the whole string is used, break and do the final printout */
972 if (len == used)
973 break;
974
975 if (to_print)
976 pr_notice("%s%.*s%s\n", KERNEL_CMDLINE_PREFIX,
977 to_print, cmdline, KERNEL_CMDLINE_CONTINUATION);
978
979 len -= used;
980 cmdline += used;
981 }
982 if (len)
983 pr_notice("%s%s\n", KERNEL_CMDLINE_PREFIX, cmdline);
984 }
985
986 asmlinkage __visible __init __no_sanitize_address __noreturn __no_stack_protector
start_kernel(void)987 void start_kernel(void)
988 {
989 char *command_line;
990 char *after_dashes;
991
992 set_task_stack_end_magic(&init_task);
993 smp_setup_processor_id();
994 debug_objects_early_init();
995 init_vmlinux_build_id();
996
997 cgroup_init_early();
998
999 local_irq_disable();
1000 early_boot_irqs_disabled = true;
1001
1002 /*
1003 * Interrupts are still disabled. Do necessary setups, then
1004 * enable them.
1005 */
1006 boot_cpu_init();
1007 page_address_init();
1008 pr_notice("%s", linux_banner);
1009 setup_arch(&command_line);
1010 mm_core_init_early();
1011 /* Static keys and static calls are needed by LSMs */
1012 jump_label_init();
1013 static_call_init();
1014 early_security_init();
1015 setup_boot_config();
1016 setup_command_line(command_line);
1017 setup_nr_cpu_ids();
1018 setup_per_cpu_areas();
1019 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
1020 early_numa_node_init();
1021 boot_cpu_hotplug_init();
1022
1023 print_kernel_cmdline(saved_command_line);
1024 /* parameters may set static keys */
1025 parse_early_param();
1026 after_dashes = parse_args("Booting kernel",
1027 static_command_line, __start___param,
1028 __stop___param - __start___param,
1029 -1, -1, NULL, &unknown_bootoption);
1030 print_unknown_bootoptions();
1031 if (!IS_ERR_OR_NULL(after_dashes))
1032 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
1033 NULL, set_init_arg);
1034 if (extra_init_args)
1035 parse_args("Setting extra init args", extra_init_args,
1036 NULL, 0, -1, -1, NULL, set_init_arg);
1037
1038 /* Architectural and non-timekeeping rng init, before allocator init */
1039 random_init_early(command_line);
1040
1041 /*
1042 * These use large bootmem allocations and must precede
1043 * initalization of page allocator
1044 */
1045 setup_log_buf(0);
1046 vfs_caches_init_early();
1047 sort_main_extable();
1048 trap_init();
1049 mm_core_init();
1050 maple_tree_init();
1051 poking_init();
1052 ftrace_init();
1053
1054 /* trace_printk can be enabled here */
1055 early_trace_init();
1056
1057 /*
1058 * Set up the scheduler prior starting any interrupts (such as the
1059 * timer interrupt). Full topology setup happens at smp_init()
1060 * time - but meanwhile we still have a functioning scheduler.
1061 */
1062 sched_init();
1063
1064 if (WARN(!irqs_disabled(),
1065 "Interrupts were enabled *very* early, fixing it\n"))
1066 local_irq_disable();
1067 radix_tree_init();
1068
1069 /*
1070 * Set up housekeeping before setting up workqueues to allow the unbound
1071 * workqueue to take non-housekeeping into account.
1072 */
1073 housekeeping_init();
1074
1075 /*
1076 * Allow workqueue creation and work item queueing/cancelling
1077 * early. Work item execution depends on kthreads and starts after
1078 * workqueue_init().
1079 */
1080 workqueue_init_early();
1081
1082 rcu_init();
1083 kvfree_rcu_init();
1084
1085 /* Trace events are available after this */
1086 trace_init();
1087
1088 if (initcall_debug)
1089 initcall_debug_enable();
1090
1091 context_tracking_init();
1092 /* init some links before init_ISA_irqs() */
1093 early_irq_init();
1094 init_IRQ();
1095 tick_init();
1096 rcu_init_nohz();
1097 timers_init();
1098 srcu_init();
1099 hrtimers_init();
1100 softirq_init();
1101 vdso_setup_data_pages();
1102 timekeeping_init();
1103 time_init();
1104
1105 /* This must be after timekeeping is initialized */
1106 random_init();
1107
1108 /* These make use of the fully initialized rng */
1109 kfence_init();
1110 boot_init_stack_canary();
1111
1112 perf_event_init();
1113 profile_init();
1114 call_function_init();
1115 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
1116
1117 early_boot_irqs_disabled = false;
1118 local_irq_enable();
1119
1120 kmem_cache_init_late();
1121
1122 /*
1123 * HACK ALERT! This is early. We're enabling the console before
1124 * we've done PCI setups etc, and console_init() must be aware of
1125 * this. But we do want output early, in case something goes wrong.
1126 */
1127 console_init();
1128 if (panic_later)
1129 panic("Too many boot %s vars at `%s'", panic_later,
1130 panic_param);
1131
1132 lockdep_init();
1133
1134 /*
1135 * Need to run this when irqs are enabled, because it wants
1136 * to self-test [hard/soft]-irqs on/off lock inversion bugs
1137 * too:
1138 */
1139 locking_selftest();
1140
1141 #ifdef CONFIG_BLK_DEV_INITRD
1142 if (initrd_start && !initrd_below_start_ok &&
1143 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1144 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1145 page_to_pfn(virt_to_page((void *)initrd_start)),
1146 min_low_pfn);
1147 initrd_start = 0;
1148 }
1149 #endif
1150 setup_per_cpu_pageset();
1151 numa_policy_init();
1152 acpi_early_init();
1153 if (late_time_init)
1154 late_time_init();
1155 sched_clock_init();
1156 calibrate_delay();
1157
1158 arch_cpu_finalize_init();
1159
1160 pid_idr_init();
1161 anon_vma_init();
1162 thread_stack_cache_init();
1163 cred_init();
1164 fork_init();
1165 proc_caches_init();
1166 uts_ns_init();
1167 time_ns_init();
1168 key_init();
1169 security_init();
1170 dbg_late_init();
1171 net_ns_init();
1172 vfs_caches_init();
1173 pagecache_init();
1174 signals_init();
1175 seq_file_init();
1176 proc_root_init();
1177 nsfs_init();
1178 pidfs_init();
1179 cpuset_init();
1180 mem_cgroup_init();
1181 cgroup_init();
1182 taskstats_init_early();
1183 delayacct_init();
1184
1185 acpi_subsystem_init();
1186 arch_post_acpi_subsys_init();
1187 kcsan_init();
1188
1189 /* Do the rest non-__init'ed, we're now alive */
1190 rest_init();
1191
1192 /*
1193 * Avoid stack canaries in callers of boot_init_stack_canary for gcc-10
1194 * and older.
1195 */
1196 #if !__has_attribute(__no_stack_protector__)
1197 prevent_tail_call_optimization();
1198 #endif
1199 }
1200
1201 /* Call all constructor functions linked into the kernel. */
do_ctors(void)1202 static void __init do_ctors(void)
1203 {
1204 /*
1205 * For UML, the constructors have already been called by the
1206 * normal setup code as it's just a normal ELF binary, so we
1207 * cannot do it again - but we do need CONFIG_CONSTRUCTORS
1208 * even on UML for modules.
1209 */
1210 #if defined(CONFIG_CONSTRUCTORS) && !defined(CONFIG_UML)
1211 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1212
1213 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1214 (*fn)();
1215 #endif
1216 }
1217
1218 #ifdef CONFIG_KALLSYMS
1219 struct blacklist_entry {
1220 struct list_head next;
1221 char *buf;
1222 };
1223
1224 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1225
initcall_blacklist(char * str)1226 static int __init initcall_blacklist(char *str)
1227 {
1228 char *str_entry;
1229 struct blacklist_entry *entry;
1230
1231 /* str argument is a comma-separated list of functions */
1232 do {
1233 str_entry = strsep(&str, ",");
1234 if (str_entry) {
1235 pr_debug("blacklisting initcall %s\n", str_entry);
1236 entry = memblock_alloc_or_panic(sizeof(*entry),
1237 SMP_CACHE_BYTES);
1238 entry->buf = memblock_alloc_or_panic(strlen(str_entry) + 1,
1239 SMP_CACHE_BYTES);
1240 strcpy(entry->buf, str_entry);
1241 list_add(&entry->next, &blacklisted_initcalls);
1242 }
1243 } while (str_entry);
1244
1245 return 1;
1246 }
1247
initcall_blacklisted(initcall_t fn)1248 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1249 {
1250 struct blacklist_entry *entry;
1251 char fn_name[KSYM_SYMBOL_LEN];
1252 unsigned long addr;
1253
1254 if (list_empty(&blacklisted_initcalls))
1255 return false;
1256
1257 addr = (unsigned long) dereference_function_descriptor(fn);
1258 sprint_symbol_no_offset(fn_name, addr);
1259
1260 /*
1261 * fn will be "function_name [module_name]" where [module_name] is not
1262 * displayed for built-in init functions. Strip off the [module_name].
1263 */
1264 strreplace(fn_name, ' ', '\0');
1265
1266 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1267 if (!strcmp(fn_name, entry->buf)) {
1268 pr_debug("initcall %s blacklisted\n", fn_name);
1269 return true;
1270 }
1271 }
1272
1273 return false;
1274 }
1275 #else
initcall_blacklist(char * str)1276 static int __init initcall_blacklist(char *str)
1277 {
1278 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1279 return 0;
1280 }
1281
initcall_blacklisted(initcall_t fn)1282 static bool __init_or_module initcall_blacklisted(initcall_t fn)
1283 {
1284 return false;
1285 }
1286 #endif
1287 __setup("initcall_blacklist=", initcall_blacklist);
1288
1289 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)1290 trace_initcall_start_cb(void *data, initcall_t fn)
1291 {
1292 ktime_t *calltime = data;
1293
1294 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1295 *calltime = ktime_get();
1296 }
1297
1298 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)1299 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1300 {
1301 ktime_t rettime, *calltime = data;
1302
1303 rettime = ktime_get();
1304 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1305 fn, ret, (unsigned long long)ktime_us_delta(rettime, *calltime));
1306 }
1307
1308 static __init_or_module void
trace_initcall_level_cb(void * data,const char * level)1309 trace_initcall_level_cb(void *data, const char *level)
1310 {
1311 printk(KERN_DEBUG "entering initcall level: %s\n", level);
1312 }
1313
1314 static ktime_t initcall_calltime;
1315
1316 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)1317 static void __init initcall_debug_enable(void)
1318 {
1319 int ret;
1320
1321 ret = register_trace_initcall_start(trace_initcall_start_cb,
1322 &initcall_calltime);
1323 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1324 &initcall_calltime);
1325 ret |= register_trace_initcall_level(trace_initcall_level_cb, NULL);
1326 WARN(ret, "Failed to register initcall tracepoints\n");
1327 }
1328 # define do_trace_initcall_start trace_initcall_start
1329 # define do_trace_initcall_finish trace_initcall_finish
1330 # define do_trace_initcall_level trace_initcall_level
1331 #else
do_trace_initcall_start(initcall_t fn)1332 static inline void do_trace_initcall_start(initcall_t fn)
1333 {
1334 if (!initcall_debug)
1335 return;
1336 trace_initcall_start_cb(&initcall_calltime, fn);
1337 }
do_trace_initcall_finish(initcall_t fn,int ret)1338 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1339 {
1340 if (!initcall_debug)
1341 return;
1342 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1343 }
do_trace_initcall_level(const char * level)1344 static inline void do_trace_initcall_level(const char *level)
1345 {
1346 if (!initcall_debug)
1347 return;
1348 trace_initcall_level_cb(NULL, level);
1349 }
1350 #endif /* !TRACEPOINTS_ENABLED */
1351
do_one_initcall(initcall_t fn)1352 int __init_or_module do_one_initcall(initcall_t fn)
1353 {
1354 int count = preempt_count();
1355 char msgbuf[64];
1356 int ret;
1357
1358 if (initcall_blacklisted(fn))
1359 return -EPERM;
1360
1361 do_trace_initcall_start(fn);
1362 ret = fn();
1363 do_trace_initcall_finish(fn, ret);
1364
1365 msgbuf[0] = 0;
1366
1367 if (preempt_count() != count) {
1368 sprintf(msgbuf, "preemption imbalance ");
1369 preempt_count_set(count);
1370 }
1371 if (irqs_disabled()) {
1372 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1373 local_irq_enable();
1374 }
1375 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1376
1377 add_latent_entropy();
1378 return ret;
1379 }
1380
1381
1382 static initcall_entry_t *initcall_levels[] __initdata = {
1383 __initcall0_start,
1384 __initcall1_start,
1385 __initcall2_start,
1386 __initcall3_start,
1387 __initcall4_start,
1388 __initcall5_start,
1389 __initcall6_start,
1390 __initcall7_start,
1391 __initcall_end,
1392 };
1393
1394 /* Keep these in sync with initcalls in include/linux/init.h */
1395 static const char *initcall_level_names[] __initdata = {
1396 "pure",
1397 "core",
1398 "postcore",
1399 "arch",
1400 "subsys",
1401 "fs",
1402 "device",
1403 "late",
1404 };
1405
ignore_unknown_bootoption(char * param,char * val,const char * unused,void * arg)1406 static int __init ignore_unknown_bootoption(char *param, char *val,
1407 const char *unused, void *arg)
1408 {
1409 return 0;
1410 }
1411
do_initcall_level(int level,char * command_line)1412 static void __init do_initcall_level(int level, char *command_line)
1413 {
1414 initcall_entry_t *fn;
1415
1416 parse_args(initcall_level_names[level],
1417 command_line, __start___param,
1418 __stop___param - __start___param,
1419 level, level,
1420 NULL, ignore_unknown_bootoption);
1421
1422 do_trace_initcall_level(initcall_level_names[level]);
1423 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1424 do_one_initcall(initcall_from_entry(fn));
1425 }
1426
do_initcalls(void)1427 static void __init do_initcalls(void)
1428 {
1429 int level;
1430 size_t len = saved_command_line_len + 1;
1431 char *command_line;
1432
1433 command_line = kzalloc(len, GFP_KERNEL);
1434 if (!command_line)
1435 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1436
1437 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1438 /* Parser modifies command_line, restore it each time */
1439 strcpy(command_line, saved_command_line);
1440 do_initcall_level(level, command_line);
1441 }
1442
1443 kfree(command_line);
1444 }
1445
1446 /*
1447 * Ok, the machine is now initialized. None of the devices
1448 * have been touched yet, but the CPU subsystem is up and
1449 * running, and memory and process management works.
1450 *
1451 * Now we can finally start doing some real work..
1452 */
do_basic_setup(void)1453 static void __init do_basic_setup(void)
1454 {
1455 cpuset_init_smp();
1456 ksysfs_init();
1457 driver_init();
1458 init_irq_proc();
1459 do_ctors();
1460 do_initcalls();
1461 }
1462
do_pre_smp_initcalls(void)1463 static void __init do_pre_smp_initcalls(void)
1464 {
1465 initcall_entry_t *fn;
1466
1467 do_trace_initcall_level("early");
1468 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1469 do_one_initcall(initcall_from_entry(fn));
1470 }
1471
run_init_process(const char * init_filename)1472 static int run_init_process(const char *init_filename)
1473 {
1474 const char *const *p;
1475
1476 argv_init[0] = init_filename;
1477 pr_info("Run %s as init process\n", init_filename);
1478 pr_debug(" with arguments:\n");
1479 for (p = argv_init; *p; p++)
1480 pr_debug(" %s\n", *p);
1481 pr_debug(" with environment:\n");
1482 for (p = envp_init; *p; p++)
1483 pr_debug(" %s\n", *p);
1484 return kernel_execve(init_filename, argv_init, envp_init);
1485 }
1486
try_to_run_init_process(const char * init_filename)1487 static int try_to_run_init_process(const char *init_filename)
1488 {
1489 int ret;
1490
1491 ret = run_init_process(init_filename);
1492
1493 if (ret && ret != -ENOENT) {
1494 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1495 init_filename, ret);
1496 }
1497
1498 return ret;
1499 }
1500
1501 static noinline void __init kernel_init_freeable(void);
1502
1503 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1504 bool rodata_enabled __ro_after_init = true;
1505
1506 #ifndef arch_parse_debug_rodata
arch_parse_debug_rodata(char * str)1507 static inline bool arch_parse_debug_rodata(char *str) { return false; }
1508 #endif
1509
set_debug_rodata(char * str)1510 static int __init set_debug_rodata(char *str)
1511 {
1512 if (arch_parse_debug_rodata(str))
1513 return 0;
1514
1515 if (str && !strcmp(str, "on"))
1516 rodata_enabled = true;
1517 else if (str && !strcmp(str, "off"))
1518 rodata_enabled = false;
1519 else
1520 pr_warn("Invalid option string for rodata: '%s'\n", str);
1521 return 0;
1522 }
1523 early_param("rodata", set_debug_rodata);
1524 #endif
1525
mark_readonly(void)1526 static void mark_readonly(void)
1527 {
1528 if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && rodata_enabled) {
1529 /*
1530 * load_module() results in W+X mappings, which are cleaned
1531 * up with init_free_wq. Let's make sure that queued work is
1532 * flushed so that we don't hit false positives looking for
1533 * insecure pages which are W+X.
1534 */
1535 flush_module_init_free_work();
1536 jump_label_init_ro();
1537 mark_rodata_ro();
1538 debug_checkwx();
1539 rodata_test();
1540 } else if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX)) {
1541 pr_info("Kernel memory protection disabled.\n");
1542 } else if (IS_ENABLED(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)) {
1543 pr_warn("Kernel memory protection not selected by kernel config.\n");
1544 } else {
1545 pr_warn("This architecture does not have kernel memory protection.\n");
1546 }
1547 }
1548
free_initmem(void)1549 void __weak free_initmem(void)
1550 {
1551 free_initmem_default(POISON_FREE_INITMEM);
1552 }
1553
kernel_init(void * unused)1554 static int __ref kernel_init(void *unused)
1555 {
1556 int ret;
1557
1558 init_userspace_fs();
1559
1560 /*
1561 * Wait until kthreadd is all set-up.
1562 */
1563 wait_for_completion(&kthreadd_done);
1564
1565 kernel_init_freeable();
1566 /* need to finish all async __init code before freeing the memory */
1567 async_synchronize_full();
1568
1569 system_state = SYSTEM_FREEING_INITMEM;
1570 kprobe_free_init_mem();
1571 ftrace_free_init_mem();
1572 kgdb_free_init_mem();
1573 exit_boot_config();
1574 free_initmem();
1575 mark_readonly();
1576
1577 /*
1578 * Kernel mappings are now finalized - update the userspace page-table
1579 * to finalize PTI.
1580 */
1581 pti_finalize();
1582
1583 system_state = SYSTEM_RUNNING;
1584 numa_default_policy();
1585
1586 rcu_end_inkernel_boot();
1587
1588 do_sysctl_args();
1589
1590 if (ramdisk_execute_command) {
1591 ret = run_init_process(ramdisk_execute_command);
1592 if (!ret)
1593 return 0;
1594 pr_err("Failed to execute %s (error %d)\n",
1595 ramdisk_execute_command, ret);
1596 }
1597
1598 /*
1599 * We try each of these until one succeeds.
1600 *
1601 * The Bourne shell can be used instead of init if we are
1602 * trying to recover a really broken machine.
1603 */
1604 if (execute_command) {
1605 ret = run_init_process(execute_command);
1606 if (!ret)
1607 return 0;
1608 panic("Requested init %s failed (error %d).",
1609 execute_command, ret);
1610 }
1611
1612 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1613 ret = run_init_process(CONFIG_DEFAULT_INIT);
1614 if (ret)
1615 pr_err("Default init %s failed (error %d)\n",
1616 CONFIG_DEFAULT_INIT, ret);
1617 else
1618 return 0;
1619 }
1620
1621 if (!try_to_run_init_process("/sbin/init") ||
1622 !try_to_run_init_process("/etc/init") ||
1623 !try_to_run_init_process("/bin/init") ||
1624 !try_to_run_init_process("/bin/sh"))
1625 return 0;
1626
1627 panic("No working init found. Try passing init= option to kernel. "
1628 "See Linux Documentation/admin-guide/init.rst for guidance.");
1629 }
1630
1631 /* Open /dev/console, for stdin/stdout/stderr, this should never fail */
console_on_rootfs(void)1632 void __init console_on_rootfs(void)
1633 {
1634 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1635
1636 if (IS_ERR(file)) {
1637 pr_err("Warning: unable to open an initial console.\n");
1638 return;
1639 }
1640 init_dup(file);
1641 init_dup(file);
1642 init_dup(file);
1643 fput(file);
1644 }
1645
kernel_init_freeable(void)1646 static noinline void __init kernel_init_freeable(void)
1647 {
1648 /* Now the scheduler is fully set up and can do blocking allocations */
1649 gfp_allowed_mask = __GFP_BITS_MASK;
1650
1651 /*
1652 * init can allocate pages on any node
1653 */
1654 set_mems_allowed(node_states[N_MEMORY]);
1655
1656 rcu_assign_pointer(cad_pid, get_pid(task_pid(current)));
1657
1658 smp_prepare_cpus(setup_max_cpus);
1659
1660 workqueue_init();
1661
1662 init_mm_internals();
1663
1664 do_pre_smp_initcalls();
1665 lockup_detector_init();
1666
1667 smp_init();
1668 sched_init_smp();
1669
1670 workqueue_init_topology();
1671 async_init();
1672 padata_init();
1673 page_alloc_init_late();
1674
1675 do_basic_setup();
1676
1677 kunit_run_all_tests();
1678
1679 wait_for_initramfs();
1680 console_on_rootfs();
1681
1682 /*
1683 * check if there is an early userspace init. If yes, let it do all
1684 * the work
1685 */
1686 int ramdisk_command_access;
1687 ramdisk_command_access = init_eaccess(ramdisk_execute_command);
1688 if (ramdisk_command_access != 0) {
1689 if (ramdisk_execute_command_set)
1690 pr_warn("check access for rdinit=%s failed: %i, ignoring\n",
1691 ramdisk_execute_command, ramdisk_command_access);
1692 ramdisk_execute_command = NULL;
1693 prepare_namespace();
1694 }
1695
1696 /*
1697 * Ok, we have completed the initial bootup, and
1698 * we're essentially up and running. Get rid of the
1699 * initmem segments and start the user-mode stuff..
1700 *
1701 * rootfs is available now, try loading the public keys
1702 * and default modules
1703 */
1704
1705 integrity_load_keys();
1706 }
1707