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/console.h>
32 #include <linux/nmi.h>
33 #include <linux/percpu.h>
34 #include <linux/kmod.h>
35 #include <linux/kprobes.h>
36 #include <linux/vmalloc.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/start_kernel.h>
39 #include <linux/security.h>
40 #include <linux/smp.h>
41 #include <linux/profile.h>
42 #include <linux/rcupdate.h>
43 #include <linux/moduleparam.h>
44 #include <linux/kallsyms.h>
45 #include <linux/writeback.h>
46 #include <linux/cpu.h>
47 #include <linux/cpuset.h>
48 #include <linux/cgroup.h>
49 #include <linux/efi.h>
50 #include <linux/tick.h>
51 #include <linux/sched/isolation.h>
52 #include <linux/interrupt.h>
53 #include <linux/taskstats_kern.h>
54 #include <linux/delayacct.h>
55 #include <linux/unistd.h>
56 #include <linux/utsname.h>
57 #include <linux/rmap.h>
58 #include <linux/mempolicy.h>
59 #include <linux/key.h>
60 #include <linux/buffer_head.h>
61 #include <linux/page_ext.h>
62 #include <linux/debug_locks.h>
63 #include <linux/debugobjects.h>
64 #include <linux/lockdep.h>
65 #include <linux/kmemleak.h>
66 #include <linux/pid_namespace.h>
67 #include <linux/device.h>
68 #include <linux/kthread.h>
69 #include <linux/sched.h>
70 #include <linux/sched/init.h>
71 #include <linux/signal.h>
72 #include <linux/idr.h>
73 #include <linux/kgdb.h>
74 #include <linux/ftrace.h>
75 #include <linux/async.h>
76 #include <linux/sfi.h>
77 #include <linux/shmem_fs.h>
78 #include <linux/slab.h>
79 #include <linux/perf_event.h>
80 #include <linux/ptrace.h>
81 #include <linux/pti.h>
82 #include <linux/blkdev.h>
83 #include <linux/elevator.h>
84 #include <linux/sched/clock.h>
85 #include <linux/sched/task.h>
86 #include <linux/sched/task_stack.h>
87 #include <linux/context_tracking.h>
88 #include <linux/random.h>
89 #include <linux/list.h>
90 #include <linux/integrity.h>
91 #include <linux/proc_ns.h>
92 #include <linux/io.h>
93 #include <linux/cache.h>
94 #include <linux/rodata_test.h>
95 #include <linux/jump_label.h>
96
97 #include <asm/io.h>
98 #include <asm/setup.h>
99 #include <asm/sections.h>
100 #include <asm/cacheflush.h>
101
102 #define CREATE_TRACE_POINTS
103 #include <trace/events/initcall.h>
104
105 static int kernel_init(void *);
106
107 extern void init_IRQ(void);
108 extern void radix_tree_init(void);
109
110 /*
111 * Debug helper: via this flag we know that we are in 'early bootup code'
112 * where only the boot processor is running with IRQ disabled. This means
113 * two things - IRQ must not be enabled before the flag is cleared and some
114 * operations which are not allowed with IRQ disabled are allowed while the
115 * flag is set.
116 */
117 bool early_boot_irqs_disabled __read_mostly;
118
119 enum system_states system_state __read_mostly;
120 EXPORT_SYMBOL(system_state);
121
122 /*
123 * Boot command-line arguments
124 */
125 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
126 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
127
128 extern void time_init(void);
129 /* Default late time init is NULL. archs can override this later. */
130 void (*__initdata late_time_init)(void);
131
132 /* Untouched command line saved by arch-specific code. */
133 char __initdata boot_command_line[COMMAND_LINE_SIZE];
134 /* Untouched saved command line (eg. for /proc) */
135 char *saved_command_line;
136 /* Command line for parameter parsing */
137 static char *static_command_line;
138 /* Command line for per-initcall parameter parsing */
139 static char *initcall_command_line;
140
141 static char *execute_command;
142 static char *ramdisk_execute_command;
143
144 /*
145 * Used to generate warnings if static_key manipulation functions are used
146 * before jump_label_init is called.
147 */
148 bool static_key_initialized __read_mostly;
149 EXPORT_SYMBOL_GPL(static_key_initialized);
150
151 /*
152 * If set, this is an indication to the drivers that reset the underlying
153 * device before going ahead with the initialization otherwise driver might
154 * rely on the BIOS and skip the reset operation.
155 *
156 * This is useful if kernel is booting in an unreliable environment.
157 * For ex. kdump situation where previous kernel has crashed, BIOS has been
158 * skipped and devices will be in unknown state.
159 */
160 unsigned int reset_devices;
161 EXPORT_SYMBOL(reset_devices);
162
set_reset_devices(char * str)163 static int __init set_reset_devices(char *str)
164 {
165 reset_devices = 1;
166 return 1;
167 }
168
169 __setup("reset_devices", set_reset_devices);
170
171 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
172 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
173 static const char *panic_later, *panic_param;
174
175 extern const struct obs_kernel_param __setup_start[], __setup_end[];
176
obsolete_checksetup(char * line)177 static bool __init obsolete_checksetup(char *line)
178 {
179 const struct obs_kernel_param *p;
180 bool had_early_param = false;
181
182 p = __setup_start;
183 do {
184 int n = strlen(p->str);
185 if (parameqn(line, p->str, n)) {
186 if (p->early) {
187 /* Already done in parse_early_param?
188 * (Needs exact match on param part).
189 * Keep iterating, as we can have early
190 * params and __setups of same names 8( */
191 if (line[n] == '\0' || line[n] == '=')
192 had_early_param = true;
193 } else if (!p->setup_func) {
194 pr_warn("Parameter %s is obsolete, ignored\n",
195 p->str);
196 return true;
197 } else if (p->setup_func(line + n))
198 return true;
199 }
200 p++;
201 } while (p < __setup_end);
202
203 return had_early_param;
204 }
205
206 /*
207 * This should be approx 2 Bo*oMips to start (note initial shift), and will
208 * still work even if initially too large, it will just take slightly longer
209 */
210 unsigned long loops_per_jiffy = (1<<12);
211 EXPORT_SYMBOL(loops_per_jiffy);
212
debug_kernel(char * str)213 static int __init debug_kernel(char *str)
214 {
215 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
216 return 0;
217 }
218
quiet_kernel(char * str)219 static int __init quiet_kernel(char *str)
220 {
221 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
222 return 0;
223 }
224
225 early_param("debug", debug_kernel);
226 early_param("quiet", quiet_kernel);
227
loglevel(char * str)228 static int __init loglevel(char *str)
229 {
230 int newlevel;
231
232 /*
233 * Only update loglevel value when a correct setting was passed,
234 * to prevent blind crashes (when loglevel being set to 0) that
235 * are quite hard to debug
236 */
237 if (get_option(&str, &newlevel)) {
238 console_loglevel = newlevel;
239 return 0;
240 }
241
242 return -EINVAL;
243 }
244
245 early_param("loglevel", loglevel);
246
247 /* Change NUL term back to "=", to make "param" the whole string. */
repair_env_string(char * param,char * val,const char * unused,void * arg)248 static int __init repair_env_string(char *param, char *val,
249 const char *unused, void *arg)
250 {
251 if (val) {
252 /* param=val or param="val"? */
253 if (val == param+strlen(param)+1)
254 val[-1] = '=';
255 else if (val == param+strlen(param)+2) {
256 val[-2] = '=';
257 memmove(val-1, val, strlen(val)+1);
258 val--;
259 } else
260 BUG();
261 }
262 return 0;
263 }
264
265 /* Anything after -- gets handed straight to init. */
set_init_arg(char * param,char * val,const char * unused,void * arg)266 static int __init set_init_arg(char *param, char *val,
267 const char *unused, void *arg)
268 {
269 unsigned int i;
270
271 if (panic_later)
272 return 0;
273
274 repair_env_string(param, val, unused, NULL);
275
276 for (i = 0; argv_init[i]; i++) {
277 if (i == MAX_INIT_ARGS) {
278 panic_later = "init";
279 panic_param = param;
280 return 0;
281 }
282 }
283 argv_init[i] = param;
284 return 0;
285 }
286
287 /*
288 * Unknown boot options get handed to init, unless they look like
289 * unused parameters (modprobe will find them in /proc/cmdline).
290 */
unknown_bootoption(char * param,char * val,const char * unused,void * arg)291 static int __init unknown_bootoption(char *param, char *val,
292 const char *unused, void *arg)
293 {
294 repair_env_string(param, val, unused, NULL);
295
296 /* Handle obsolete-style parameters */
297 if (obsolete_checksetup(param))
298 return 0;
299
300 /* Unused module parameter. */
301 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
302 return 0;
303
304 if (panic_later)
305 return 0;
306
307 if (val) {
308 /* Environment option */
309 unsigned int i;
310 for (i = 0; envp_init[i]; i++) {
311 if (i == MAX_INIT_ENVS) {
312 panic_later = "env";
313 panic_param = param;
314 }
315 if (!strncmp(param, envp_init[i], val - param))
316 break;
317 }
318 envp_init[i] = param;
319 } else {
320 /* Command line option */
321 unsigned int i;
322 for (i = 0; argv_init[i]; i++) {
323 if (i == MAX_INIT_ARGS) {
324 panic_later = "init";
325 panic_param = param;
326 }
327 }
328 argv_init[i] = param;
329 }
330 return 0;
331 }
332
init_setup(char * str)333 static int __init init_setup(char *str)
334 {
335 unsigned int i;
336
337 execute_command = str;
338 /*
339 * In case LILO is going to boot us with default command line,
340 * it prepends "auto" before the whole cmdline which makes
341 * the shell think it should execute a script with such name.
342 * So we ignore all arguments entered _before_ init=... [MJ]
343 */
344 for (i = 1; i < MAX_INIT_ARGS; i++)
345 argv_init[i] = NULL;
346 return 1;
347 }
348 __setup("init=", init_setup);
349
rdinit_setup(char * str)350 static int __init rdinit_setup(char *str)
351 {
352 unsigned int i;
353
354 ramdisk_execute_command = str;
355 /* See "auto" comment in init_setup */
356 for (i = 1; i < MAX_INIT_ARGS; i++)
357 argv_init[i] = NULL;
358 return 1;
359 }
360 __setup("rdinit=", rdinit_setup);
361
362 #ifndef CONFIG_SMP
363 static const unsigned int setup_max_cpus = NR_CPUS;
setup_nr_cpu_ids(void)364 static inline void setup_nr_cpu_ids(void) { }
smp_prepare_cpus(unsigned int maxcpus)365 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
366 #endif
367
368 /*
369 * We need to store the untouched command line for future reference.
370 * We also need to store the touched command line since the parameter
371 * parsing is performed in place, and we should allow a component to
372 * store reference of name/value for future reference.
373 */
setup_command_line(char * command_line)374 static void __init setup_command_line(char *command_line)
375 {
376 size_t len = strlen(boot_command_line) + 1;
377
378 saved_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
379 if (!saved_command_line)
380 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
381
382 initcall_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
383 if (!initcall_command_line)
384 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
385
386 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
387 if (!static_command_line)
388 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
389
390 strcpy(saved_command_line, boot_command_line);
391 strcpy(static_command_line, command_line);
392 }
393
394 /*
395 * We need to finalize in a non-__init function or else race conditions
396 * between the root thread and the init thread may cause start_kernel to
397 * be reaped by free_initmem before the root thread has proceeded to
398 * cpu_idle.
399 *
400 * gcc-3.4 accidentally inlines this function, so use noinline.
401 */
402
403 static __initdata DECLARE_COMPLETION(kthreadd_done);
404
rest_init(void)405 noinline void __ref rest_init(void)
406 {
407 struct task_struct *tsk;
408 int pid;
409
410 rcu_scheduler_starting();
411 /*
412 * We need to spawn init first so that it obtains pid 1, however
413 * the init task will end up wanting to create kthreads, which, if
414 * we schedule it before we create kthreadd, will OOPS.
415 */
416 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
417 /*
418 * Pin init on the boot CPU. Task migration is not properly working
419 * until sched_init_smp() has been run. It will set the allowed
420 * CPUs for init to the non isolated CPUs.
421 */
422 rcu_read_lock();
423 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
424 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
425 rcu_read_unlock();
426
427 numa_default_policy();
428 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
429 rcu_read_lock();
430 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
431 rcu_read_unlock();
432
433 /*
434 * Enable might_sleep() and smp_processor_id() checks.
435 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
436 * kernel_thread() would trigger might_sleep() splats. With
437 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
438 * already, but it's stuck on the kthreadd_done completion.
439 */
440 system_state = SYSTEM_SCHEDULING;
441
442 complete(&kthreadd_done);
443
444 /*
445 * The boot idle thread must execute schedule()
446 * at least once to get things moving:
447 */
448 schedule_preempt_disabled();
449 /* Call into cpu_idle with preempt disabled */
450 cpu_startup_entry(CPUHP_ONLINE);
451 }
452
453 /* Check for early params. */
do_early_param(char * param,char * val,const char * unused,void * arg)454 static int __init do_early_param(char *param, char *val,
455 const char *unused, void *arg)
456 {
457 const struct obs_kernel_param *p;
458
459 for (p = __setup_start; p < __setup_end; p++) {
460 if ((p->early && parameq(param, p->str)) ||
461 (strcmp(param, "console") == 0 &&
462 strcmp(p->str, "earlycon") == 0)
463 ) {
464 if (p->setup_func(val) != 0)
465 pr_warn("Malformed early option '%s'\n", param);
466 }
467 }
468 /* We accept everything at this stage. */
469 return 0;
470 }
471
parse_early_options(char * cmdline)472 void __init parse_early_options(char *cmdline)
473 {
474 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
475 do_early_param);
476 }
477
478 /* Arch code calls this early on, or if not, just before other parsing. */
parse_early_param(void)479 void __init parse_early_param(void)
480 {
481 static int done __initdata;
482 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
483
484 if (done)
485 return;
486
487 /* All fall through to do_early_param. */
488 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
489 parse_early_options(tmp_cmdline);
490 done = 1;
491 }
492
arch_post_acpi_subsys_init(void)493 void __init __weak arch_post_acpi_subsys_init(void) { }
494
smp_setup_processor_id(void)495 void __init __weak smp_setup_processor_id(void)
496 {
497 }
498
499 # if THREAD_SIZE >= PAGE_SIZE
thread_stack_cache_init(void)500 void __init __weak thread_stack_cache_init(void)
501 {
502 }
503 #endif
504
poking_init(void)505 void __init __weak poking_init(void) { }
506
pgtable_cache_init(void)507 void __init __weak pgtable_cache_init(void) { }
508
509 bool initcall_debug;
510 core_param(initcall_debug, initcall_debug, bool, 0644);
511
512 #ifdef TRACEPOINTS_ENABLED
513 static void __init initcall_debug_enable(void);
514 #else
initcall_debug_enable(void)515 static inline void initcall_debug_enable(void)
516 {
517 }
518 #endif
519
520 /* Report memory auto-initialization states for this boot. */
report_meminit(void)521 static void __init report_meminit(void)
522 {
523 const char *stack;
524
525 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
526 stack = "all(pattern)";
527 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
528 stack = "all(zero)";
529 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
530 stack = "byref_all(zero)";
531 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
532 stack = "byref(zero)";
533 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
534 stack = "__user(zero)";
535 else
536 stack = "off";
537
538 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
539 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
540 want_init_on_free() ? "on" : "off");
541 if (want_init_on_free())
542 pr_info("mem auto-init: clearing system memory may take some time...\n");
543 }
544
545 /*
546 * Set up kernel memory allocators
547 */
mm_init(void)548 static void __init mm_init(void)
549 {
550 /*
551 * page_ext requires contiguous pages,
552 * bigger than MAX_ORDER unless SPARSEMEM.
553 */
554 page_ext_init_flatmem();
555 init_debug_pagealloc();
556 report_meminit();
557 mem_init();
558 kmem_cache_init();
559 kmemleak_init();
560 pgtable_init();
561 debug_objects_mem_init();
562 vmalloc_init();
563 ioremap_huge_init();
564 /* Should be run before the first non-init thread is created */
565 init_espfix_bsp();
566 /* Should be run after espfix64 is set up. */
567 pti_init();
568 mm_cache_init();
569 }
570
arch_call_rest_init(void)571 void __init __weak arch_call_rest_init(void)
572 {
573 rest_init();
574 }
575
start_kernel(void)576 asmlinkage __visible void __init start_kernel(void)
577 {
578 char *command_line;
579 char *after_dashes;
580
581 set_task_stack_end_magic(&init_task);
582 smp_setup_processor_id();
583 debug_objects_early_init();
584
585 cgroup_init_early();
586
587 local_irq_disable();
588 early_boot_irqs_disabled = true;
589
590 /*
591 * Interrupts are still disabled. Do necessary setups, then
592 * enable them.
593 */
594 boot_cpu_init();
595 page_address_init();
596 pr_notice("%s", linux_banner);
597 early_security_init();
598 setup_arch(&command_line);
599 setup_command_line(command_line);
600 setup_nr_cpu_ids();
601 setup_per_cpu_areas();
602 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
603 boot_cpu_hotplug_init();
604
605 build_all_zonelists(NULL);
606 page_alloc_init();
607
608 pr_notice("Kernel command line: %s\n", boot_command_line);
609 /* parameters may set static keys */
610 jump_label_init();
611 parse_early_param();
612 after_dashes = parse_args("Booting kernel",
613 static_command_line, __start___param,
614 __stop___param - __start___param,
615 -1, -1, NULL, &unknown_bootoption);
616 if (!IS_ERR_OR_NULL(after_dashes))
617 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
618 NULL, set_init_arg);
619
620 /*
621 * These use large bootmem allocations and must precede
622 * kmem_cache_init()
623 */
624 setup_log_buf(0);
625 vfs_caches_init_early();
626 sort_main_extable();
627 trap_init();
628 mm_init();
629 poking_init();
630 ftrace_init();
631
632 /* trace_printk can be enabled here */
633 early_trace_init();
634
635 /*
636 * Set up the scheduler prior starting any interrupts (such as the
637 * timer interrupt). Full topology setup happens at smp_init()
638 * time - but meanwhile we still have a functioning scheduler.
639 */
640 sched_init();
641 /*
642 * Disable preemption - early bootup scheduling is extremely
643 * fragile until we cpu_idle() for the first time.
644 */
645 preempt_disable();
646 if (WARN(!irqs_disabled(),
647 "Interrupts were enabled *very* early, fixing it\n"))
648 local_irq_disable();
649 radix_tree_init();
650
651 /*
652 * Set up housekeeping before setting up workqueues to allow the unbound
653 * workqueue to take non-housekeeping into account.
654 */
655 housekeeping_init();
656
657 /*
658 * Allow workqueue creation and work item queueing/cancelling
659 * early. Work item execution depends on kthreads and starts after
660 * workqueue_init().
661 */
662 workqueue_init_early();
663
664 rcu_init();
665
666 /* Trace events are available after this */
667 trace_init();
668
669 if (initcall_debug)
670 initcall_debug_enable();
671
672 context_tracking_init();
673 /* init some links before init_ISA_irqs() */
674 early_irq_init();
675 init_IRQ();
676 tick_init();
677 rcu_init_nohz();
678 init_timers();
679 hrtimers_init();
680 softirq_init();
681 timekeeping_init();
682 time_init();
683
684 /*
685 * For best initial stack canary entropy, prepare it after:
686 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
687 * - timekeeping_init() for ktime entropy used in random_init()
688 * - time_init() for making random_get_entropy() work on some platforms
689 * - random_init() to initialize the RNG from from early entropy sources
690 */
691 random_init(command_line);
692 boot_init_stack_canary();
693
694 perf_event_init();
695 profile_init();
696 call_function_init();
697 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
698
699 early_boot_irqs_disabled = false;
700 local_irq_enable();
701
702 kmem_cache_init_late();
703
704 /*
705 * HACK ALERT! This is early. We're enabling the console before
706 * we've done PCI setups etc, and console_init() must be aware of
707 * this. But we do want output early, in case something goes wrong.
708 */
709 console_init();
710 if (panic_later)
711 panic("Too many boot %s vars at `%s'", panic_later,
712 panic_param);
713
714 lockdep_init();
715
716 /*
717 * Need to run this when irqs are enabled, because it wants
718 * to self-test [hard/soft]-irqs on/off lock inversion bugs
719 * too:
720 */
721 locking_selftest();
722
723 #ifdef CONFIG_BLK_DEV_INITRD
724 if (initrd_start && !initrd_below_start_ok &&
725 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
726 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
727 page_to_pfn(virt_to_page((void *)initrd_start)),
728 min_low_pfn);
729 initrd_start = 0;
730 }
731 #endif
732 setup_per_cpu_pageset();
733 numa_policy_init();
734 acpi_early_init();
735 if (late_time_init)
736 late_time_init();
737 sched_clock_init();
738 calibrate_delay();
739
740 arch_cpu_finalize_init();
741
742 pid_idr_init();
743 anon_vma_init();
744 #ifdef CONFIG_X86
745 if (efi_enabled(EFI_RUNTIME_SERVICES))
746 efi_enter_virtual_mode();
747 #endif
748 thread_stack_cache_init();
749 cred_init();
750 fork_init();
751 proc_caches_init();
752 uts_ns_init();
753 buffer_init();
754 key_init();
755 security_init();
756 dbg_late_init();
757 vfs_caches_init();
758 pagecache_init();
759 signals_init();
760 seq_file_init();
761 proc_root_init();
762 nsfs_init();
763 cpuset_init();
764 cgroup_init();
765 taskstats_init_early();
766 delayacct_init();
767
768 acpi_subsystem_init();
769 arch_post_acpi_subsys_init();
770 sfi_init_late();
771
772 /* Do the rest non-__init'ed, we're now alive */
773 arch_call_rest_init();
774
775 prevent_tail_call_optimization();
776 }
777
778 /* Call all constructor functions linked into the kernel. */
do_ctors(void)779 static void __init do_ctors(void)
780 {
781 #ifdef CONFIG_CONSTRUCTORS
782 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
783
784 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
785 (*fn)();
786 #endif
787 }
788
789 #ifdef CONFIG_KALLSYMS
790 struct blacklist_entry {
791 struct list_head next;
792 char *buf;
793 };
794
795 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
796
initcall_blacklist(char * str)797 static int __init initcall_blacklist(char *str)
798 {
799 char *str_entry;
800 struct blacklist_entry *entry;
801
802 /* str argument is a comma-separated list of functions */
803 do {
804 str_entry = strsep(&str, ",");
805 if (str_entry) {
806 pr_debug("blacklisting initcall %s\n", str_entry);
807 entry = memblock_alloc(sizeof(*entry),
808 SMP_CACHE_BYTES);
809 if (!entry)
810 panic("%s: Failed to allocate %zu bytes\n",
811 __func__, sizeof(*entry));
812 entry->buf = memblock_alloc(strlen(str_entry) + 1,
813 SMP_CACHE_BYTES);
814 if (!entry->buf)
815 panic("%s: Failed to allocate %zu bytes\n",
816 __func__, strlen(str_entry) + 1);
817 strcpy(entry->buf, str_entry);
818 list_add(&entry->next, &blacklisted_initcalls);
819 }
820 } while (str_entry);
821
822 return 1;
823 }
824
initcall_blacklisted(initcall_t fn)825 static bool __init_or_module initcall_blacklisted(initcall_t fn)
826 {
827 struct blacklist_entry *entry;
828 char fn_name[KSYM_SYMBOL_LEN];
829 unsigned long addr;
830
831 if (list_empty(&blacklisted_initcalls))
832 return false;
833
834 addr = (unsigned long) dereference_function_descriptor(fn);
835 sprint_symbol_no_offset(fn_name, addr);
836
837 /*
838 * fn will be "function_name [module_name]" where [module_name] is not
839 * displayed for built-in init functions. Strip off the [module_name].
840 */
841 strreplace(fn_name, ' ', '\0');
842
843 list_for_each_entry(entry, &blacklisted_initcalls, next) {
844 if (!strcmp(fn_name, entry->buf)) {
845 pr_debug("initcall %s blacklisted\n", fn_name);
846 return true;
847 }
848 }
849
850 return false;
851 }
852 #else
initcall_blacklist(char * str)853 static int __init initcall_blacklist(char *str)
854 {
855 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
856 return 0;
857 }
858
initcall_blacklisted(initcall_t fn)859 static bool __init_or_module initcall_blacklisted(initcall_t fn)
860 {
861 return false;
862 }
863 #endif
864 __setup("initcall_blacklist=", initcall_blacklist);
865
866 static __init_or_module void
trace_initcall_start_cb(void * data,initcall_t fn)867 trace_initcall_start_cb(void *data, initcall_t fn)
868 {
869 ktime_t *calltime = (ktime_t *)data;
870
871 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
872 *calltime = ktime_get();
873 }
874
875 static __init_or_module void
trace_initcall_finish_cb(void * data,initcall_t fn,int ret)876 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
877 {
878 ktime_t *calltime = (ktime_t *)data;
879 ktime_t delta, rettime;
880 unsigned long long duration;
881
882 rettime = ktime_get();
883 delta = ktime_sub(rettime, *calltime);
884 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
885 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
886 fn, ret, duration);
887 }
888
889 static ktime_t initcall_calltime;
890
891 #ifdef TRACEPOINTS_ENABLED
initcall_debug_enable(void)892 static void __init initcall_debug_enable(void)
893 {
894 int ret;
895
896 ret = register_trace_initcall_start(trace_initcall_start_cb,
897 &initcall_calltime);
898 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
899 &initcall_calltime);
900 WARN(ret, "Failed to register initcall tracepoints\n");
901 }
902 # define do_trace_initcall_start trace_initcall_start
903 # define do_trace_initcall_finish trace_initcall_finish
904 #else
do_trace_initcall_start(initcall_t fn)905 static inline void do_trace_initcall_start(initcall_t fn)
906 {
907 if (!initcall_debug)
908 return;
909 trace_initcall_start_cb(&initcall_calltime, fn);
910 }
do_trace_initcall_finish(initcall_t fn,int ret)911 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
912 {
913 if (!initcall_debug)
914 return;
915 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
916 }
917 #endif /* !TRACEPOINTS_ENABLED */
918
do_one_initcall(initcall_t fn)919 int __init_or_module do_one_initcall(initcall_t fn)
920 {
921 int count = preempt_count();
922 char msgbuf[64];
923 int ret;
924
925 if (initcall_blacklisted(fn))
926 return -EPERM;
927
928 do_trace_initcall_start(fn);
929 ret = fn();
930 do_trace_initcall_finish(fn, ret);
931
932 msgbuf[0] = 0;
933
934 if (preempt_count() != count) {
935 sprintf(msgbuf, "preemption imbalance ");
936 preempt_count_set(count);
937 }
938 if (irqs_disabled()) {
939 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
940 local_irq_enable();
941 }
942 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
943
944 add_latent_entropy();
945 return ret;
946 }
947
948
949 extern initcall_entry_t __initcall_start[];
950 extern initcall_entry_t __initcall0_start[];
951 extern initcall_entry_t __initcall1_start[];
952 extern initcall_entry_t __initcall2_start[];
953 extern initcall_entry_t __initcall3_start[];
954 extern initcall_entry_t __initcall4_start[];
955 extern initcall_entry_t __initcall5_start[];
956 extern initcall_entry_t __initcall6_start[];
957 extern initcall_entry_t __initcall7_start[];
958 extern initcall_entry_t __initcall_end[];
959
960 static initcall_entry_t *initcall_levels[] __initdata = {
961 __initcall0_start,
962 __initcall1_start,
963 __initcall2_start,
964 __initcall3_start,
965 __initcall4_start,
966 __initcall5_start,
967 __initcall6_start,
968 __initcall7_start,
969 __initcall_end,
970 };
971
972 /* Keep these in sync with initcalls in include/linux/init.h */
973 static const char *initcall_level_names[] __initdata = {
974 "pure",
975 "core",
976 "postcore",
977 "arch",
978 "subsys",
979 "fs",
980 "device",
981 "late",
982 };
983
do_initcall_level(int level)984 static void __init do_initcall_level(int level)
985 {
986 initcall_entry_t *fn;
987
988 strcpy(initcall_command_line, saved_command_line);
989 parse_args(initcall_level_names[level],
990 initcall_command_line, __start___param,
991 __stop___param - __start___param,
992 level, level,
993 NULL, &repair_env_string);
994
995 trace_initcall_level(initcall_level_names[level]);
996 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
997 do_one_initcall(initcall_from_entry(fn));
998 }
999
do_initcalls(void)1000 static void __init do_initcalls(void)
1001 {
1002 int level;
1003
1004 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
1005 do_initcall_level(level);
1006 }
1007
1008 /*
1009 * Ok, the machine is now initialized. None of the devices
1010 * have been touched yet, but the CPU subsystem is up and
1011 * running, and memory and process management works.
1012 *
1013 * Now we can finally start doing some real work..
1014 */
do_basic_setup(void)1015 static void __init do_basic_setup(void)
1016 {
1017 cpuset_init_smp();
1018 driver_init();
1019 init_irq_proc();
1020 do_ctors();
1021 usermodehelper_enable();
1022 do_initcalls();
1023 }
1024
do_pre_smp_initcalls(void)1025 static void __init do_pre_smp_initcalls(void)
1026 {
1027 initcall_entry_t *fn;
1028
1029 trace_initcall_level("early");
1030 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1031 do_one_initcall(initcall_from_entry(fn));
1032 }
1033
run_init_process(const char * init_filename)1034 static int run_init_process(const char *init_filename)
1035 {
1036 argv_init[0] = init_filename;
1037 pr_info("Run %s as init process\n", init_filename);
1038 return do_execve(getname_kernel(init_filename),
1039 (const char __user *const __user *)argv_init,
1040 (const char __user *const __user *)envp_init);
1041 }
1042
try_to_run_init_process(const char * init_filename)1043 static int try_to_run_init_process(const char *init_filename)
1044 {
1045 int ret;
1046
1047 ret = run_init_process(init_filename);
1048
1049 if (ret && ret != -ENOENT) {
1050 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1051 init_filename, ret);
1052 }
1053
1054 return ret;
1055 }
1056
1057 static noinline void __init kernel_init_freeable(void);
1058
1059 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1060 bool rodata_enabled __ro_after_init = true;
set_debug_rodata(char * str)1061 static int __init set_debug_rodata(char *str)
1062 {
1063 if (strtobool(str, &rodata_enabled))
1064 pr_warn("Invalid option string for rodata: '%s'\n", str);
1065 return 1;
1066 }
1067 __setup("rodata=", set_debug_rodata);
1068 #endif
1069
1070 #ifdef CONFIG_STRICT_KERNEL_RWX
mark_readonly(void)1071 static void mark_readonly(void)
1072 {
1073 if (rodata_enabled) {
1074 /*
1075 * load_module() results in W+X mappings, which are cleaned
1076 * up with call_rcu(). Let's make sure that queued work is
1077 * flushed so that we don't hit false positives looking for
1078 * insecure pages which are W+X.
1079 */
1080 rcu_barrier();
1081 mark_rodata_ro();
1082 rodata_test();
1083 } else
1084 pr_info("Kernel memory protection disabled.\n");
1085 }
1086 #else
mark_readonly(void)1087 static inline void mark_readonly(void)
1088 {
1089 pr_warn("This architecture does not have kernel memory protection.\n");
1090 }
1091 #endif
1092
free_initmem(void)1093 void __weak free_initmem(void)
1094 {
1095 free_initmem_default(POISON_FREE_INITMEM);
1096 }
1097
kernel_init(void * unused)1098 static int __ref kernel_init(void *unused)
1099 {
1100 int ret;
1101
1102 kernel_init_freeable();
1103 /* need to finish all async __init code before freeing the memory */
1104 async_synchronize_full();
1105 kprobe_free_init_mem();
1106 ftrace_free_init_mem();
1107 free_initmem();
1108 mark_readonly();
1109
1110 /*
1111 * Kernel mappings are now finalized - update the userspace page-table
1112 * to finalize PTI.
1113 */
1114 pti_finalize();
1115
1116 system_state = SYSTEM_RUNNING;
1117 numa_default_policy();
1118
1119 rcu_end_inkernel_boot();
1120
1121 if (ramdisk_execute_command) {
1122 ret = run_init_process(ramdisk_execute_command);
1123 if (!ret)
1124 return 0;
1125 pr_err("Failed to execute %s (error %d)\n",
1126 ramdisk_execute_command, ret);
1127 }
1128
1129 /*
1130 * We try each of these until one succeeds.
1131 *
1132 * The Bourne shell can be used instead of init if we are
1133 * trying to recover a really broken machine.
1134 */
1135 if (execute_command) {
1136 ret = run_init_process(execute_command);
1137 if (!ret)
1138 return 0;
1139 panic("Requested init %s failed (error %d).",
1140 execute_command, ret);
1141 }
1142 if (!try_to_run_init_process("/sbin/init") ||
1143 !try_to_run_init_process("/etc/init") ||
1144 !try_to_run_init_process("/bin/init") ||
1145 !try_to_run_init_process("/bin/sh"))
1146 return 0;
1147
1148 panic("No working init found. Try passing init= option to kernel. "
1149 "See Linux Documentation/admin-guide/init.rst for guidance.");
1150 }
1151
kernel_init_freeable(void)1152 static noinline void __init kernel_init_freeable(void)
1153 {
1154 /*
1155 * Wait until kthreadd is all set-up.
1156 */
1157 wait_for_completion(&kthreadd_done);
1158
1159 /* Now the scheduler is fully set up and can do blocking allocations */
1160 gfp_allowed_mask = __GFP_BITS_MASK;
1161
1162 /*
1163 * init can allocate pages on any node
1164 */
1165 set_mems_allowed(node_states[N_MEMORY]);
1166
1167 cad_pid = get_pid(task_pid(current));
1168
1169 smp_prepare_cpus(setup_max_cpus);
1170
1171 workqueue_init();
1172
1173 init_mm_internals();
1174
1175 do_pre_smp_initcalls();
1176 lockup_detector_init();
1177
1178 smp_init();
1179 sched_init_smp();
1180
1181 page_alloc_init_late();
1182 /* Initialize page ext after all struct pages are initialized. */
1183 page_ext_init();
1184
1185 do_basic_setup();
1186
1187 /* Open the /dev/console on the rootfs, this should never fail */
1188 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1189 pr_err("Warning: unable to open an initial console.\n");
1190
1191 (void) ksys_dup(0);
1192 (void) ksys_dup(0);
1193 /*
1194 * check if there is an early userspace init. If yes, let it do all
1195 * the work
1196 */
1197
1198 if (!ramdisk_execute_command)
1199 ramdisk_execute_command = "/init";
1200
1201 if (ksys_access((const char __user *)
1202 ramdisk_execute_command, 0) != 0) {
1203 ramdisk_execute_command = NULL;
1204 prepare_namespace();
1205 }
1206
1207 /*
1208 * Ok, we have completed the initial bootup, and
1209 * we're essentially up and running. Get rid of the
1210 * initmem segments and start the user-mode stuff..
1211 *
1212 * rootfs is available now, try loading the public keys
1213 * and default modules
1214 */
1215
1216 integrity_load_keys();
1217 }
1218