xref: /openbmc/linux/init/main.c (revision ff4a7481c3898ffc3cc271d6aca431d190c37247)
1 /*
2  *  linux/init/main.c
3  *
4  *  Copyright (C) 1991, 1992  Linus Torvalds
5  *
6  *  GK 2/5/95  -  Changed to support mounting root fs via NFS
7  *  Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
8  *  Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
9  *  Simplified starting of init:  Michael A. Griffith <grif@acm.org>
10  */
11 
12 #define DEBUG		/* Enable initcall_debug */
13 
14 #include <linux/types.h>
15 #include <linux/extable.h>
16 #include <linux/module.h>
17 #include <linux/proc_fs.h>
18 #include <linux/binfmts.h>
19 #include <linux/kernel.h>
20 #include <linux/syscalls.h>
21 #include <linux/stackprotector.h>
22 #include <linux/string.h>
23 #include <linux/ctype.h>
24 #include <linux/delay.h>
25 #include <linux/ioport.h>
26 #include <linux/init.h>
27 #include <linux/initrd.h>
28 #include <linux/memblock.h>
29 #include <linux/acpi.h>
30 #include <linux/console.h>
31 #include <linux/nmi.h>
32 #include <linux/percpu.h>
33 #include <linux/kmod.h>
34 #include <linux/vmalloc.h>
35 #include <linux/kernel_stat.h>
36 #include <linux/start_kernel.h>
37 #include <linux/security.h>
38 #include <linux/smp.h>
39 #include <linux/profile.h>
40 #include <linux/rcupdate.h>
41 #include <linux/moduleparam.h>
42 #include <linux/kallsyms.h>
43 #include <linux/writeback.h>
44 #include <linux/cpu.h>
45 #include <linux/cpuset.h>
46 #include <linux/cgroup.h>
47 #include <linux/efi.h>
48 #include <linux/tick.h>
49 #include <linux/sched/isolation.h>
50 #include <linux/interrupt.h>
51 #include <linux/taskstats_kern.h>
52 #include <linux/delayacct.h>
53 #include <linux/unistd.h>
54 #include <linux/utsname.h>
55 #include <linux/rmap.h>
56 #include <linux/mempolicy.h>
57 #include <linux/key.h>
58 #include <linux/buffer_head.h>
59 #include <linux/page_ext.h>
60 #include <linux/debug_locks.h>
61 #include <linux/debugobjects.h>
62 #include <linux/lockdep.h>
63 #include <linux/kmemleak.h>
64 #include <linux/pid_namespace.h>
65 #include <linux/device.h>
66 #include <linux/kthread.h>
67 #include <linux/sched.h>
68 #include <linux/sched/init.h>
69 #include <linux/signal.h>
70 #include <linux/idr.h>
71 #include <linux/kgdb.h>
72 #include <linux/ftrace.h>
73 #include <linux/async.h>
74 #include <linux/sfi.h>
75 #include <linux/shmem_fs.h>
76 #include <linux/slab.h>
77 #include <linux/perf_event.h>
78 #include <linux/ptrace.h>
79 #include <linux/pti.h>
80 #include <linux/blkdev.h>
81 #include <linux/elevator.h>
82 #include <linux/sched/clock.h>
83 #include <linux/sched/task.h>
84 #include <linux/sched/task_stack.h>
85 #include <linux/context_tracking.h>
86 #include <linux/random.h>
87 #include <linux/list.h>
88 #include <linux/integrity.h>
89 #include <linux/proc_ns.h>
90 #include <linux/io.h>
91 #include <linux/cache.h>
92 #include <linux/rodata_test.h>
93 #include <linux/jump_label.h>
94 #include <linux/mem_encrypt.h>
95 
96 #include <asm/io.h>
97 #include <asm/bugs.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 fork_init(void);
109 extern void radix_tree_init(void);
110 
111 /*
112  * Debug helper: via this flag we know that we are in 'early bootup code'
113  * where only the boot processor is running with IRQ disabled.  This means
114  * two things - IRQ must not be enabled before the flag is cleared and some
115  * operations which are not allowed with IRQ disabled are allowed while the
116  * flag is set.
117  */
118 bool early_boot_irqs_disabled __read_mostly;
119 
120 enum system_states system_state __read_mostly;
121 EXPORT_SYMBOL(system_state);
122 
123 /*
124  * Boot command-line arguments
125  */
126 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
127 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
128 
129 extern void time_init(void);
130 /* Default late time init is NULL. archs can override this later. */
131 void (*__initdata late_time_init)(void);
132 
133 /* Untouched command line saved by arch-specific code. */
134 char __initdata boot_command_line[COMMAND_LINE_SIZE];
135 /* Untouched saved command line (eg. for /proc) */
136 char *saved_command_line;
137 /* Command line for parameter parsing */
138 static char *static_command_line;
139 /* Command line for per-initcall parameter parsing */
140 static char *initcall_command_line;
141 
142 static char *execute_command;
143 static char *ramdisk_execute_command;
144 
145 /*
146  * Used to generate warnings if static_key manipulation functions are used
147  * before jump_label_init is called.
148  */
149 bool static_key_initialized __read_mostly;
150 EXPORT_SYMBOL_GPL(static_key_initialized);
151 
152 /*
153  * If set, this is an indication to the drivers that reset the underlying
154  * device before going ahead with the initialization otherwise driver might
155  * rely on the BIOS and skip the reset operation.
156  *
157  * This is useful if kernel is booting in an unreliable environment.
158  * For ex. kdump situation where previous kernel has crashed, BIOS has been
159  * skipped and devices will be in unknown state.
160  */
161 unsigned int reset_devices;
162 EXPORT_SYMBOL(reset_devices);
163 
164 static int __init set_reset_devices(char *str)
165 {
166 	reset_devices = 1;
167 	return 1;
168 }
169 
170 __setup("reset_devices", set_reset_devices);
171 
172 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
173 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
174 static const char *panic_later, *panic_param;
175 
176 extern const struct obs_kernel_param __setup_start[], __setup_end[];
177 
178 static bool __init obsolete_checksetup(char *line)
179 {
180 	const struct obs_kernel_param *p;
181 	bool had_early_param = false;
182 
183 	p = __setup_start;
184 	do {
185 		int n = strlen(p->str);
186 		if (parameqn(line, p->str, n)) {
187 			if (p->early) {
188 				/* Already done in parse_early_param?
189 				 * (Needs exact match on param part).
190 				 * Keep iterating, as we can have early
191 				 * params and __setups of same names 8( */
192 				if (line[n] == '\0' || line[n] == '=')
193 					had_early_param = true;
194 			} else if (!p->setup_func) {
195 				pr_warn("Parameter %s is obsolete, ignored\n",
196 					p->str);
197 				return true;
198 			} else if (p->setup_func(line + n))
199 				return true;
200 		}
201 		p++;
202 	} while (p < __setup_end);
203 
204 	return had_early_param;
205 }
206 
207 /*
208  * This should be approx 2 Bo*oMips to start (note initial shift), and will
209  * still work even if initially too large, it will just take slightly longer
210  */
211 unsigned long loops_per_jiffy = (1<<12);
212 EXPORT_SYMBOL(loops_per_jiffy);
213 
214 static int __init debug_kernel(char *str)
215 {
216 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
217 	return 0;
218 }
219 
220 static int __init quiet_kernel(char *str)
221 {
222 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
223 	return 0;
224 }
225 
226 early_param("debug", debug_kernel);
227 early_param("quiet", quiet_kernel);
228 
229 static int __init loglevel(char *str)
230 {
231 	int newlevel;
232 
233 	/*
234 	 * Only update loglevel value when a correct setting was passed,
235 	 * to prevent blind crashes (when loglevel being set to 0) that
236 	 * are quite hard to debug
237 	 */
238 	if (get_option(&str, &newlevel)) {
239 		console_loglevel = newlevel;
240 		return 0;
241 	}
242 
243 	return -EINVAL;
244 }
245 
246 early_param("loglevel", loglevel);
247 
248 /* Change NUL term back to "=", to make "param" the whole string. */
249 static int __init repair_env_string(char *param, char *val,
250 				    const char *unused, void *arg)
251 {
252 	if (val) {
253 		/* param=val or param="val"? */
254 		if (val == param+strlen(param)+1)
255 			val[-1] = '=';
256 		else if (val == param+strlen(param)+2) {
257 			val[-2] = '=';
258 			memmove(val-1, val, strlen(val)+1);
259 			val--;
260 		} else
261 			BUG();
262 	}
263 	return 0;
264 }
265 
266 /* Anything after -- gets handed straight to init. */
267 static int __init set_init_arg(char *param, char *val,
268 			       const char *unused, void *arg)
269 {
270 	unsigned int i;
271 
272 	if (panic_later)
273 		return 0;
274 
275 	repair_env_string(param, val, unused, NULL);
276 
277 	for (i = 0; argv_init[i]; i++) {
278 		if (i == MAX_INIT_ARGS) {
279 			panic_later = "init";
280 			panic_param = param;
281 			return 0;
282 		}
283 	}
284 	argv_init[i] = param;
285 	return 0;
286 }
287 
288 /*
289  * Unknown boot options get handed to init, unless they look like
290  * unused parameters (modprobe will find them in /proc/cmdline).
291  */
292 static int __init unknown_bootoption(char *param, char *val,
293 				     const char *unused, void *arg)
294 {
295 	repair_env_string(param, val, unused, NULL);
296 
297 	/* Handle obsolete-style parameters */
298 	if (obsolete_checksetup(param))
299 		return 0;
300 
301 	/* Unused module parameter. */
302 	if (strchr(param, '.') && (!val || strchr(param, '.') < val))
303 		return 0;
304 
305 	if (panic_later)
306 		return 0;
307 
308 	if (val) {
309 		/* Environment option */
310 		unsigned int i;
311 		for (i = 0; envp_init[i]; i++) {
312 			if (i == MAX_INIT_ENVS) {
313 				panic_later = "env";
314 				panic_param = param;
315 			}
316 			if (!strncmp(param, envp_init[i], val - param))
317 				break;
318 		}
319 		envp_init[i] = param;
320 	} else {
321 		/* Command line option */
322 		unsigned int i;
323 		for (i = 0; argv_init[i]; i++) {
324 			if (i == MAX_INIT_ARGS) {
325 				panic_later = "init";
326 				panic_param = param;
327 			}
328 		}
329 		argv_init[i] = param;
330 	}
331 	return 0;
332 }
333 
334 static int __init init_setup(char *str)
335 {
336 	unsigned int i;
337 
338 	execute_command = str;
339 	/*
340 	 * In case LILO is going to boot us with default command line,
341 	 * it prepends "auto" before the whole cmdline which makes
342 	 * the shell think it should execute a script with such name.
343 	 * So we ignore all arguments entered _before_ init=... [MJ]
344 	 */
345 	for (i = 1; i < MAX_INIT_ARGS; i++)
346 		argv_init[i] = NULL;
347 	return 1;
348 }
349 __setup("init=", init_setup);
350 
351 static int __init rdinit_setup(char *str)
352 {
353 	unsigned int i;
354 
355 	ramdisk_execute_command = str;
356 	/* See "auto" comment in init_setup */
357 	for (i = 1; i < MAX_INIT_ARGS; i++)
358 		argv_init[i] = NULL;
359 	return 1;
360 }
361 __setup("rdinit=", rdinit_setup);
362 
363 #ifndef CONFIG_SMP
364 static const unsigned int setup_max_cpus = NR_CPUS;
365 static inline void setup_nr_cpu_ids(void) { }
366 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
367 #endif
368 
369 /*
370  * We need to store the untouched command line for future reference.
371  * We also need to store the touched command line since the parameter
372  * parsing is performed in place, and we should allow a component to
373  * store reference of name/value for future reference.
374  */
375 static void __init setup_command_line(char *command_line)
376 {
377 	saved_command_line =
378 		memblock_alloc(strlen(boot_command_line) + 1, SMP_CACHE_BYTES);
379 	initcall_command_line =
380 		memblock_alloc(strlen(boot_command_line) + 1, SMP_CACHE_BYTES);
381 	static_command_line = memblock_alloc(strlen(command_line) + 1,
382 					     SMP_CACHE_BYTES);
383 	strcpy(saved_command_line, boot_command_line);
384 	strcpy(static_command_line, command_line);
385 }
386 
387 /*
388  * We need to finalize in a non-__init function or else race conditions
389  * between the root thread and the init thread may cause start_kernel to
390  * be reaped by free_initmem before the root thread has proceeded to
391  * cpu_idle.
392  *
393  * gcc-3.4 accidentally inlines this function, so use noinline.
394  */
395 
396 static __initdata DECLARE_COMPLETION(kthreadd_done);
397 
398 noinline void __ref rest_init(void)
399 {
400 	struct task_struct *tsk;
401 	int pid;
402 
403 	rcu_scheduler_starting();
404 	/*
405 	 * We need to spawn init first so that it obtains pid 1, however
406 	 * the init task will end up wanting to create kthreads, which, if
407 	 * we schedule it before we create kthreadd, will OOPS.
408 	 */
409 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
410 	/*
411 	 * Pin init on the boot CPU. Task migration is not properly working
412 	 * until sched_init_smp() has been run. It will set the allowed
413 	 * CPUs for init to the non isolated CPUs.
414 	 */
415 	rcu_read_lock();
416 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
417 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
418 	rcu_read_unlock();
419 
420 	numa_default_policy();
421 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
422 	rcu_read_lock();
423 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
424 	rcu_read_unlock();
425 
426 	/*
427 	 * Enable might_sleep() and smp_processor_id() checks.
428 	 * They cannot be enabled earlier because with CONFIG_PREEMPT=y
429 	 * kernel_thread() would trigger might_sleep() splats. With
430 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
431 	 * already, but it's stuck on the kthreadd_done completion.
432 	 */
433 	system_state = SYSTEM_SCHEDULING;
434 
435 	complete(&kthreadd_done);
436 
437 	/*
438 	 * The boot idle thread must execute schedule()
439 	 * at least once to get things moving:
440 	 */
441 	schedule_preempt_disabled();
442 	/* Call into cpu_idle with preempt disabled */
443 	cpu_startup_entry(CPUHP_ONLINE);
444 }
445 
446 /* Check for early params. */
447 static int __init do_early_param(char *param, char *val,
448 				 const char *unused, void *arg)
449 {
450 	const struct obs_kernel_param *p;
451 
452 	for (p = __setup_start; p < __setup_end; p++) {
453 		if ((p->early && parameq(param, p->str)) ||
454 		    (strcmp(param, "console") == 0 &&
455 		     strcmp(p->str, "earlycon") == 0)
456 		) {
457 			if (p->setup_func(val) != 0)
458 				pr_warn("Malformed early option '%s'\n", param);
459 		}
460 	}
461 	/* We accept everything at this stage. */
462 	return 0;
463 }
464 
465 void __init parse_early_options(char *cmdline)
466 {
467 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
468 		   do_early_param);
469 }
470 
471 /* Arch code calls this early on, or if not, just before other parsing. */
472 void __init parse_early_param(void)
473 {
474 	static int done __initdata;
475 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
476 
477 	if (done)
478 		return;
479 
480 	/* All fall through to do_early_param. */
481 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
482 	parse_early_options(tmp_cmdline);
483 	done = 1;
484 }
485 
486 void __init __weak arch_post_acpi_subsys_init(void) { }
487 
488 void __init __weak smp_setup_processor_id(void)
489 {
490 }
491 
492 # if THREAD_SIZE >= PAGE_SIZE
493 void __init __weak thread_stack_cache_init(void)
494 {
495 }
496 #endif
497 
498 void __init __weak mem_encrypt_init(void) { }
499 
500 bool initcall_debug;
501 core_param(initcall_debug, initcall_debug, bool, 0644);
502 
503 #ifdef TRACEPOINTS_ENABLED
504 static void __init initcall_debug_enable(void);
505 #else
506 static inline void initcall_debug_enable(void)
507 {
508 }
509 #endif
510 
511 /*
512  * Set up kernel memory allocators
513  */
514 static void __init mm_init(void)
515 {
516 	/*
517 	 * page_ext requires contiguous pages,
518 	 * bigger than MAX_ORDER unless SPARSEMEM.
519 	 */
520 	page_ext_init_flatmem();
521 	mem_init();
522 	kmem_cache_init();
523 	pgtable_init();
524 	vmalloc_init();
525 	ioremap_huge_init();
526 	/* Should be run before the first non-init thread is created */
527 	init_espfix_bsp();
528 	/* Should be run after espfix64 is set up. */
529 	pti_init();
530 }
531 
532 void __init __weak arch_call_rest_init(void)
533 {
534 	rest_init();
535 }
536 
537 asmlinkage __visible void __init start_kernel(void)
538 {
539 	char *command_line;
540 	char *after_dashes;
541 
542 	set_task_stack_end_magic(&init_task);
543 	smp_setup_processor_id();
544 	debug_objects_early_init();
545 
546 	cgroup_init_early();
547 
548 	local_irq_disable();
549 	early_boot_irqs_disabled = true;
550 
551 	/*
552 	 * Interrupts are still disabled. Do necessary setups, then
553 	 * enable them.
554 	 */
555 	boot_cpu_init();
556 	page_address_init();
557 	pr_notice("%s", linux_banner);
558 	setup_arch(&command_line);
559 	/*
560 	 * Set up the the initial canary and entropy after arch
561 	 * and after adding latent and command line entropy.
562 	 */
563 	add_latent_entropy();
564 	add_device_randomness(command_line, strlen(command_line));
565 	boot_init_stack_canary();
566 	mm_init_cpumask(&init_mm);
567 	setup_command_line(command_line);
568 	setup_nr_cpu_ids();
569 	setup_per_cpu_areas();
570 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
571 	boot_cpu_hotplug_init();
572 
573 	build_all_zonelists(NULL);
574 	page_alloc_init();
575 
576 	pr_notice("Kernel command line: %s\n", boot_command_line);
577 	parse_early_param();
578 	after_dashes = parse_args("Booting kernel",
579 				  static_command_line, __start___param,
580 				  __stop___param - __start___param,
581 				  -1, -1, NULL, &unknown_bootoption);
582 	if (!IS_ERR_OR_NULL(after_dashes))
583 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
584 			   NULL, set_init_arg);
585 
586 	jump_label_init();
587 
588 	/*
589 	 * These use large bootmem allocations and must precede
590 	 * kmem_cache_init()
591 	 */
592 	setup_log_buf(0);
593 	vfs_caches_init_early();
594 	sort_main_extable();
595 	trap_init();
596 	mm_init();
597 
598 	ftrace_init();
599 
600 	/* trace_printk can be enabled here */
601 	early_trace_init();
602 
603 	/*
604 	 * Set up the scheduler prior starting any interrupts (such as the
605 	 * timer interrupt). Full topology setup happens at smp_init()
606 	 * time - but meanwhile we still have a functioning scheduler.
607 	 */
608 	sched_init();
609 	/*
610 	 * Disable preemption - early bootup scheduling is extremely
611 	 * fragile until we cpu_idle() for the first time.
612 	 */
613 	preempt_disable();
614 	if (WARN(!irqs_disabled(),
615 		 "Interrupts were enabled *very* early, fixing it\n"))
616 		local_irq_disable();
617 	radix_tree_init();
618 
619 	/*
620 	 * Set up housekeeping before setting up workqueues to allow the unbound
621 	 * workqueue to take non-housekeeping into account.
622 	 */
623 	housekeeping_init();
624 
625 	/*
626 	 * Allow workqueue creation and work item queueing/cancelling
627 	 * early.  Work item execution depends on kthreads and starts after
628 	 * workqueue_init().
629 	 */
630 	workqueue_init_early();
631 
632 	rcu_init();
633 
634 	/* Trace events are available after this */
635 	trace_init();
636 
637 	if (initcall_debug)
638 		initcall_debug_enable();
639 
640 	context_tracking_init();
641 	/* init some links before init_ISA_irqs() */
642 	early_irq_init();
643 	init_IRQ();
644 	tick_init();
645 	rcu_init_nohz();
646 	init_timers();
647 	hrtimers_init();
648 	softirq_init();
649 	timekeeping_init();
650 	time_init();
651 	printk_safe_init();
652 	perf_event_init();
653 	profile_init();
654 	call_function_init();
655 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
656 
657 	early_boot_irqs_disabled = false;
658 	local_irq_enable();
659 
660 	kmem_cache_init_late();
661 
662 	/*
663 	 * HACK ALERT! This is early. We're enabling the console before
664 	 * we've done PCI setups etc, and console_init() must be aware of
665 	 * this. But we do want output early, in case something goes wrong.
666 	 */
667 	console_init();
668 	if (panic_later)
669 		panic("Too many boot %s vars at `%s'", panic_later,
670 		      panic_param);
671 
672 	lockdep_init();
673 
674 	/*
675 	 * Need to run this when irqs are enabled, because it wants
676 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
677 	 * too:
678 	 */
679 	locking_selftest();
680 
681 	/*
682 	 * This needs to be called before any devices perform DMA
683 	 * operations that might use the SWIOTLB bounce buffers. It will
684 	 * mark the bounce buffers as decrypted so that their usage will
685 	 * not cause "plain-text" data to be decrypted when accessed.
686 	 */
687 	mem_encrypt_init();
688 
689 #ifdef CONFIG_BLK_DEV_INITRD
690 	if (initrd_start && !initrd_below_start_ok &&
691 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
692 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
693 		    page_to_pfn(virt_to_page((void *)initrd_start)),
694 		    min_low_pfn);
695 		initrd_start = 0;
696 	}
697 #endif
698 	page_ext_init();
699 	kmemleak_init();
700 	debug_objects_mem_init();
701 	setup_per_cpu_pageset();
702 	numa_policy_init();
703 	acpi_early_init();
704 	if (late_time_init)
705 		late_time_init();
706 	sched_clock_init();
707 	calibrate_delay();
708 	pid_idr_init();
709 	anon_vma_init();
710 #ifdef CONFIG_X86
711 	if (efi_enabled(EFI_RUNTIME_SERVICES))
712 		efi_enter_virtual_mode();
713 #endif
714 	thread_stack_cache_init();
715 	cred_init();
716 	fork_init();
717 	proc_caches_init();
718 	uts_ns_init();
719 	buffer_init();
720 	key_init();
721 	security_init();
722 	dbg_late_init();
723 	vfs_caches_init();
724 	pagecache_init();
725 	signals_init();
726 	seq_file_init();
727 	proc_root_init();
728 	nsfs_init();
729 	cpuset_init();
730 	cgroup_init();
731 	taskstats_init_early();
732 	delayacct_init();
733 
734 	check_bugs();
735 
736 	acpi_subsystem_init();
737 	arch_post_acpi_subsys_init();
738 	sfi_init_late();
739 
740 	/* Do the rest non-__init'ed, we're now alive */
741 	arch_call_rest_init();
742 }
743 
744 /* Call all constructor functions linked into the kernel. */
745 static void __init do_ctors(void)
746 {
747 #ifdef CONFIG_CONSTRUCTORS
748 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
749 
750 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
751 		(*fn)();
752 #endif
753 }
754 
755 #ifdef CONFIG_KALLSYMS
756 struct blacklist_entry {
757 	struct list_head next;
758 	char *buf;
759 };
760 
761 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
762 
763 static int __init initcall_blacklist(char *str)
764 {
765 	char *str_entry;
766 	struct blacklist_entry *entry;
767 
768 	/* str argument is a comma-separated list of functions */
769 	do {
770 		str_entry = strsep(&str, ",");
771 		if (str_entry) {
772 			pr_debug("blacklisting initcall %s\n", str_entry);
773 			entry = memblock_alloc(sizeof(*entry),
774 					       SMP_CACHE_BYTES);
775 			entry->buf = memblock_alloc(strlen(str_entry) + 1,
776 						    SMP_CACHE_BYTES);
777 			strcpy(entry->buf, str_entry);
778 			list_add(&entry->next, &blacklisted_initcalls);
779 		}
780 	} while (str_entry);
781 
782 	return 0;
783 }
784 
785 static bool __init_or_module initcall_blacklisted(initcall_t fn)
786 {
787 	struct blacklist_entry *entry;
788 	char fn_name[KSYM_SYMBOL_LEN];
789 	unsigned long addr;
790 
791 	if (list_empty(&blacklisted_initcalls))
792 		return false;
793 
794 	addr = (unsigned long) dereference_function_descriptor(fn);
795 	sprint_symbol_no_offset(fn_name, addr);
796 
797 	/*
798 	 * fn will be "function_name [module_name]" where [module_name] is not
799 	 * displayed for built-in init functions.  Strip off the [module_name].
800 	 */
801 	strreplace(fn_name, ' ', '\0');
802 
803 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
804 		if (!strcmp(fn_name, entry->buf)) {
805 			pr_debug("initcall %s blacklisted\n", fn_name);
806 			return true;
807 		}
808 	}
809 
810 	return false;
811 }
812 #else
813 static int __init initcall_blacklist(char *str)
814 {
815 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
816 	return 0;
817 }
818 
819 static bool __init_or_module initcall_blacklisted(initcall_t fn)
820 {
821 	return false;
822 }
823 #endif
824 __setup("initcall_blacklist=", initcall_blacklist);
825 
826 static __init_or_module void
827 trace_initcall_start_cb(void *data, initcall_t fn)
828 {
829 	ktime_t *calltime = (ktime_t *)data;
830 
831 	printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
832 	*calltime = ktime_get();
833 }
834 
835 static __init_or_module void
836 trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
837 {
838 	ktime_t *calltime = (ktime_t *)data;
839 	ktime_t delta, rettime;
840 	unsigned long long duration;
841 
842 	rettime = ktime_get();
843 	delta = ktime_sub(rettime, *calltime);
844 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
845 	printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
846 		 fn, ret, duration);
847 }
848 
849 static ktime_t initcall_calltime;
850 
851 #ifdef TRACEPOINTS_ENABLED
852 static void __init initcall_debug_enable(void)
853 {
854 	int ret;
855 
856 	ret = register_trace_initcall_start(trace_initcall_start_cb,
857 					    &initcall_calltime);
858 	ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
859 					      &initcall_calltime);
860 	WARN(ret, "Failed to register initcall tracepoints\n");
861 }
862 # define do_trace_initcall_start	trace_initcall_start
863 # define do_trace_initcall_finish	trace_initcall_finish
864 #else
865 static inline void do_trace_initcall_start(initcall_t fn)
866 {
867 	if (!initcall_debug)
868 		return;
869 	trace_initcall_start_cb(&initcall_calltime, fn);
870 }
871 static inline void do_trace_initcall_finish(initcall_t fn, int ret)
872 {
873 	if (!initcall_debug)
874 		return;
875 	trace_initcall_finish_cb(&initcall_calltime, fn, ret);
876 }
877 #endif /* !TRACEPOINTS_ENABLED */
878 
879 int __init_or_module do_one_initcall(initcall_t fn)
880 {
881 	int count = preempt_count();
882 	char msgbuf[64];
883 	int ret;
884 
885 	if (initcall_blacklisted(fn))
886 		return -EPERM;
887 
888 	do_trace_initcall_start(fn);
889 	ret = fn();
890 	do_trace_initcall_finish(fn, ret);
891 
892 	msgbuf[0] = 0;
893 
894 	if (preempt_count() != count) {
895 		sprintf(msgbuf, "preemption imbalance ");
896 		preempt_count_set(count);
897 	}
898 	if (irqs_disabled()) {
899 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
900 		local_irq_enable();
901 	}
902 	WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
903 
904 	add_latent_entropy();
905 	return ret;
906 }
907 
908 
909 extern initcall_entry_t __initcall_start[];
910 extern initcall_entry_t __initcall0_start[];
911 extern initcall_entry_t __initcall1_start[];
912 extern initcall_entry_t __initcall2_start[];
913 extern initcall_entry_t __initcall3_start[];
914 extern initcall_entry_t __initcall4_start[];
915 extern initcall_entry_t __initcall5_start[];
916 extern initcall_entry_t __initcall6_start[];
917 extern initcall_entry_t __initcall7_start[];
918 extern initcall_entry_t __initcall_end[];
919 
920 static initcall_entry_t *initcall_levels[] __initdata = {
921 	__initcall0_start,
922 	__initcall1_start,
923 	__initcall2_start,
924 	__initcall3_start,
925 	__initcall4_start,
926 	__initcall5_start,
927 	__initcall6_start,
928 	__initcall7_start,
929 	__initcall_end,
930 };
931 
932 /* Keep these in sync with initcalls in include/linux/init.h */
933 static char *initcall_level_names[] __initdata = {
934 	"pure",
935 	"core",
936 	"postcore",
937 	"arch",
938 	"subsys",
939 	"fs",
940 	"device",
941 	"late",
942 };
943 
944 static void __init do_initcall_level(int level)
945 {
946 	initcall_entry_t *fn;
947 
948 	strcpy(initcall_command_line, saved_command_line);
949 	parse_args(initcall_level_names[level],
950 		   initcall_command_line, __start___param,
951 		   __stop___param - __start___param,
952 		   level, level,
953 		   NULL, &repair_env_string);
954 
955 	trace_initcall_level(initcall_level_names[level]);
956 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
957 		do_one_initcall(initcall_from_entry(fn));
958 }
959 
960 static void __init do_initcalls(void)
961 {
962 	int level;
963 
964 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
965 		do_initcall_level(level);
966 }
967 
968 /*
969  * Ok, the machine is now initialized. None of the devices
970  * have been touched yet, but the CPU subsystem is up and
971  * running, and memory and process management works.
972  *
973  * Now we can finally start doing some real work..
974  */
975 static void __init do_basic_setup(void)
976 {
977 	cpuset_init_smp();
978 	shmem_init();
979 	driver_init();
980 	init_irq_proc();
981 	do_ctors();
982 	usermodehelper_enable();
983 	do_initcalls();
984 }
985 
986 static void __init do_pre_smp_initcalls(void)
987 {
988 	initcall_entry_t *fn;
989 
990 	trace_initcall_level("early");
991 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
992 		do_one_initcall(initcall_from_entry(fn));
993 }
994 
995 /*
996  * This function requests modules which should be loaded by default and is
997  * called twice right after initrd is mounted and right before init is
998  * exec'd.  If such modules are on either initrd or rootfs, they will be
999  * loaded before control is passed to userland.
1000  */
1001 void __init load_default_modules(void)
1002 {
1003 	load_default_elevator_module();
1004 }
1005 
1006 static int run_init_process(const char *init_filename)
1007 {
1008 	argv_init[0] = init_filename;
1009 	pr_info("Run %s as init process\n", init_filename);
1010 	return do_execve(getname_kernel(init_filename),
1011 		(const char __user *const __user *)argv_init,
1012 		(const char __user *const __user *)envp_init);
1013 }
1014 
1015 static int try_to_run_init_process(const char *init_filename)
1016 {
1017 	int ret;
1018 
1019 	ret = run_init_process(init_filename);
1020 
1021 	if (ret && ret != -ENOENT) {
1022 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1023 		       init_filename, ret);
1024 	}
1025 
1026 	return ret;
1027 }
1028 
1029 static noinline void __init kernel_init_freeable(void);
1030 
1031 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1032 bool rodata_enabled __ro_after_init = true;
1033 static int __init set_debug_rodata(char *str)
1034 {
1035 	return strtobool(str, &rodata_enabled);
1036 }
1037 __setup("rodata=", set_debug_rodata);
1038 #endif
1039 
1040 #ifdef CONFIG_STRICT_KERNEL_RWX
1041 static void mark_readonly(void)
1042 {
1043 	if (rodata_enabled) {
1044 		/*
1045 		 * load_module() results in W+X mappings, which are cleaned
1046 		 * up with call_rcu().  Let's make sure that queued work is
1047 		 * flushed so that we don't hit false positives looking for
1048 		 * insecure pages which are W+X.
1049 		 */
1050 		rcu_barrier();
1051 		mark_rodata_ro();
1052 		rodata_test();
1053 	} else
1054 		pr_info("Kernel memory protection disabled.\n");
1055 }
1056 #else
1057 static inline void mark_readonly(void)
1058 {
1059 	pr_warn("This architecture does not have kernel memory protection.\n");
1060 }
1061 #endif
1062 
1063 static int __ref kernel_init(void *unused)
1064 {
1065 	int ret;
1066 
1067 	kernel_init_freeable();
1068 	/* need to finish all async __init code before freeing the memory */
1069 	async_synchronize_full();
1070 	ftrace_free_init_mem();
1071 	free_initmem();
1072 	mark_readonly();
1073 
1074 	/*
1075 	 * Kernel mappings are now finalized - update the userspace page-table
1076 	 * to finalize PTI.
1077 	 */
1078 	pti_finalize();
1079 
1080 	system_state = SYSTEM_RUNNING;
1081 	numa_default_policy();
1082 
1083 	rcu_end_inkernel_boot();
1084 
1085 	if (ramdisk_execute_command) {
1086 		ret = run_init_process(ramdisk_execute_command);
1087 		if (!ret)
1088 			return 0;
1089 		pr_err("Failed to execute %s (error %d)\n",
1090 		       ramdisk_execute_command, ret);
1091 	}
1092 
1093 	/*
1094 	 * We try each of these until one succeeds.
1095 	 *
1096 	 * The Bourne shell can be used instead of init if we are
1097 	 * trying to recover a really broken machine.
1098 	 */
1099 	if (execute_command) {
1100 		ret = run_init_process(execute_command);
1101 		if (!ret)
1102 			return 0;
1103 		panic("Requested init %s failed (error %d).",
1104 		      execute_command, ret);
1105 	}
1106 	if (!try_to_run_init_process("/sbin/init") ||
1107 	    !try_to_run_init_process("/etc/init") ||
1108 	    !try_to_run_init_process("/bin/init") ||
1109 	    !try_to_run_init_process("/bin/sh"))
1110 		return 0;
1111 
1112 	panic("No working init found.  Try passing init= option to kernel. "
1113 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1114 }
1115 
1116 static noinline void __init kernel_init_freeable(void)
1117 {
1118 	/*
1119 	 * Wait until kthreadd is all set-up.
1120 	 */
1121 	wait_for_completion(&kthreadd_done);
1122 
1123 	/* Now the scheduler is fully set up and can do blocking allocations */
1124 	gfp_allowed_mask = __GFP_BITS_MASK;
1125 
1126 	/*
1127 	 * init can allocate pages on any node
1128 	 */
1129 	set_mems_allowed(node_states[N_MEMORY]);
1130 
1131 	cad_pid = task_pid(current);
1132 
1133 	smp_prepare_cpus(setup_max_cpus);
1134 
1135 	workqueue_init();
1136 
1137 	init_mm_internals();
1138 
1139 	do_pre_smp_initcalls();
1140 	lockup_detector_init();
1141 
1142 	smp_init();
1143 	sched_init_smp();
1144 
1145 	page_alloc_init_late();
1146 
1147 	do_basic_setup();
1148 
1149 	/* Open the /dev/console on the rootfs, this should never fail */
1150 	if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1151 		pr_err("Warning: unable to open an initial console.\n");
1152 
1153 	(void) ksys_dup(0);
1154 	(void) ksys_dup(0);
1155 	/*
1156 	 * check if there is an early userspace init.  If yes, let it do all
1157 	 * the work
1158 	 */
1159 
1160 	if (!ramdisk_execute_command)
1161 		ramdisk_execute_command = "/init";
1162 
1163 	if (ksys_access((const char __user *)
1164 			ramdisk_execute_command, 0) != 0) {
1165 		ramdisk_execute_command = NULL;
1166 		prepare_namespace();
1167 	}
1168 
1169 	/*
1170 	 * Ok, we have completed the initial bootup, and
1171 	 * we're essentially up and running. Get rid of the
1172 	 * initmem segments and start the user-mode stuff..
1173 	 *
1174 	 * rootfs is available now, try loading the public keys
1175 	 * and default modules
1176 	 */
1177 
1178 	integrity_load_keys();
1179 	load_default_modules();
1180 }
1181