xref: /openbmc/linux/init/main.c (revision d0807da78e11d46f18399cbf8c4028c731346766)
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/bootmem.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/rmap.h>
55 #include <linux/mempolicy.h>
56 #include <linux/key.h>
57 #include <linux/buffer_head.h>
58 #include <linux/page_ext.h>
59 #include <linux/debug_locks.h>
60 #include <linux/debugobjects.h>
61 #include <linux/lockdep.h>
62 #include <linux/kmemleak.h>
63 #include <linux/pid_namespace.h>
64 #include <linux/device.h>
65 #include <linux/kthread.h>
66 #include <linux/sched.h>
67 #include <linux/sched/init.h>
68 #include <linux/signal.h>
69 #include <linux/idr.h>
70 #include <linux/kgdb.h>
71 #include <linux/ftrace.h>
72 #include <linux/async.h>
73 #include <linux/kmemcheck.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/blkdev.h>
80 #include <linux/elevator.h>
81 #include <linux/sched_clock.h>
82 #include <linux/sched/task.h>
83 #include <linux/sched/task_stack.h>
84 #include <linux/context_tracking.h>
85 #include <linux/random.h>
86 #include <linux/list.h>
87 #include <linux/integrity.h>
88 #include <linux/proc_ns.h>
89 #include <linux/io.h>
90 #include <linux/cache.h>
91 #include <linux/rodata_test.h>
92 
93 #include <asm/io.h>
94 #include <asm/bugs.h>
95 #include <asm/setup.h>
96 #include <asm/sections.h>
97 #include <asm/cacheflush.h>
98 
99 static int kernel_init(void *);
100 
101 extern void init_IRQ(void);
102 extern void fork_init(void);
103 extern void radix_tree_init(void);
104 
105 /*
106  * Debug helper: via this flag we know that we are in 'early bootup code'
107  * where only the boot processor is running with IRQ disabled.  This means
108  * two things - IRQ must not be enabled before the flag is cleared and some
109  * operations which are not allowed with IRQ disabled are allowed while the
110  * flag is set.
111  */
112 bool early_boot_irqs_disabled __read_mostly;
113 
114 enum system_states system_state __read_mostly;
115 EXPORT_SYMBOL(system_state);
116 
117 /*
118  * Boot command-line arguments
119  */
120 #define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
121 #define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
122 
123 extern void time_init(void);
124 /* Default late time init is NULL. archs can override this later. */
125 void (*__initdata late_time_init)(void);
126 
127 /* Untouched command line saved by arch-specific code. */
128 char __initdata boot_command_line[COMMAND_LINE_SIZE];
129 /* Untouched saved command line (eg. for /proc) */
130 char *saved_command_line;
131 /* Command line for parameter parsing */
132 static char *static_command_line;
133 /* Command line for per-initcall parameter parsing */
134 static char *initcall_command_line;
135 
136 static char *execute_command;
137 static char *ramdisk_execute_command;
138 
139 /*
140  * Used to generate warnings if static_key manipulation functions are used
141  * before jump_label_init is called.
142  */
143 bool static_key_initialized __read_mostly;
144 EXPORT_SYMBOL_GPL(static_key_initialized);
145 
146 /*
147  * If set, this is an indication to the drivers that reset the underlying
148  * device before going ahead with the initialization otherwise driver might
149  * rely on the BIOS and skip the reset operation.
150  *
151  * This is useful if kernel is booting in an unreliable environment.
152  * For ex. kdump situation where previous kernel has crashed, BIOS has been
153  * skipped and devices will be in unknown state.
154  */
155 unsigned int reset_devices;
156 EXPORT_SYMBOL(reset_devices);
157 
158 static int __init set_reset_devices(char *str)
159 {
160 	reset_devices = 1;
161 	return 1;
162 }
163 
164 __setup("reset_devices", set_reset_devices);
165 
166 static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
167 const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
168 static const char *panic_later, *panic_param;
169 
170 extern const struct obs_kernel_param __setup_start[], __setup_end[];
171 
172 static bool __init obsolete_checksetup(char *line)
173 {
174 	const struct obs_kernel_param *p;
175 	bool had_early_param = false;
176 
177 	p = __setup_start;
178 	do {
179 		int n = strlen(p->str);
180 		if (parameqn(line, p->str, n)) {
181 			if (p->early) {
182 				/* Already done in parse_early_param?
183 				 * (Needs exact match on param part).
184 				 * Keep iterating, as we can have early
185 				 * params and __setups of same names 8( */
186 				if (line[n] == '\0' || line[n] == '=')
187 					had_early_param = true;
188 			} else if (!p->setup_func) {
189 				pr_warn("Parameter %s is obsolete, ignored\n",
190 					p->str);
191 				return true;
192 			} else if (p->setup_func(line + n))
193 				return true;
194 		}
195 		p++;
196 	} while (p < __setup_end);
197 
198 	return had_early_param;
199 }
200 
201 /*
202  * This should be approx 2 Bo*oMips to start (note initial shift), and will
203  * still work even if initially too large, it will just take slightly longer
204  */
205 unsigned long loops_per_jiffy = (1<<12);
206 EXPORT_SYMBOL(loops_per_jiffy);
207 
208 static int __init debug_kernel(char *str)
209 {
210 	console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
211 	return 0;
212 }
213 
214 static int __init quiet_kernel(char *str)
215 {
216 	console_loglevel = CONSOLE_LOGLEVEL_QUIET;
217 	return 0;
218 }
219 
220 early_param("debug", debug_kernel);
221 early_param("quiet", quiet_kernel);
222 
223 static int __init loglevel(char *str)
224 {
225 	int newlevel;
226 
227 	/*
228 	 * Only update loglevel value when a correct setting was passed,
229 	 * to prevent blind crashes (when loglevel being set to 0) that
230 	 * are quite hard to debug
231 	 */
232 	if (get_option(&str, &newlevel)) {
233 		console_loglevel = newlevel;
234 		return 0;
235 	}
236 
237 	return -EINVAL;
238 }
239 
240 early_param("loglevel", loglevel);
241 
242 /* Change NUL term back to "=", to make "param" the whole string. */
243 static int __init repair_env_string(char *param, char *val,
244 				    const char *unused, void *arg)
245 {
246 	if (val) {
247 		/* param=val or param="val"? */
248 		if (val == param+strlen(param)+1)
249 			val[-1] = '=';
250 		else if (val == param+strlen(param)+2) {
251 			val[-2] = '=';
252 			memmove(val-1, val, strlen(val)+1);
253 			val--;
254 		} else
255 			BUG();
256 	}
257 	return 0;
258 }
259 
260 /* Anything after -- gets handed straight to init. */
261 static int __init set_init_arg(char *param, char *val,
262 			       const char *unused, void *arg)
263 {
264 	unsigned int i;
265 
266 	if (panic_later)
267 		return 0;
268 
269 	repair_env_string(param, val, unused, NULL);
270 
271 	for (i = 0; argv_init[i]; i++) {
272 		if (i == MAX_INIT_ARGS) {
273 			panic_later = "init";
274 			panic_param = param;
275 			return 0;
276 		}
277 	}
278 	argv_init[i] = param;
279 	return 0;
280 }
281 
282 /*
283  * Unknown boot options get handed to init, unless they look like
284  * unused parameters (modprobe will find them in /proc/cmdline).
285  */
286 static int __init unknown_bootoption(char *param, char *val,
287 				     const char *unused, void *arg)
288 {
289 	repair_env_string(param, val, unused, NULL);
290 
291 	/* Handle obsolete-style parameters */
292 	if (obsolete_checksetup(param))
293 		return 0;
294 
295 	/* Unused module parameter. */
296 	if (strchr(param, '.') && (!val || strchr(param, '.') < val))
297 		return 0;
298 
299 	if (panic_later)
300 		return 0;
301 
302 	if (val) {
303 		/* Environment option */
304 		unsigned int i;
305 		for (i = 0; envp_init[i]; i++) {
306 			if (i == MAX_INIT_ENVS) {
307 				panic_later = "env";
308 				panic_param = param;
309 			}
310 			if (!strncmp(param, envp_init[i], val - param))
311 				break;
312 		}
313 		envp_init[i] = param;
314 	} else {
315 		/* Command line option */
316 		unsigned int i;
317 		for (i = 0; argv_init[i]; i++) {
318 			if (i == MAX_INIT_ARGS) {
319 				panic_later = "init";
320 				panic_param = param;
321 			}
322 		}
323 		argv_init[i] = param;
324 	}
325 	return 0;
326 }
327 
328 static int __init init_setup(char *str)
329 {
330 	unsigned int i;
331 
332 	execute_command = str;
333 	/*
334 	 * In case LILO is going to boot us with default command line,
335 	 * it prepends "auto" before the whole cmdline which makes
336 	 * the shell think it should execute a script with such name.
337 	 * So we ignore all arguments entered _before_ init=... [MJ]
338 	 */
339 	for (i = 1; i < MAX_INIT_ARGS; i++)
340 		argv_init[i] = NULL;
341 	return 1;
342 }
343 __setup("init=", init_setup);
344 
345 static int __init rdinit_setup(char *str)
346 {
347 	unsigned int i;
348 
349 	ramdisk_execute_command = str;
350 	/* See "auto" comment in init_setup */
351 	for (i = 1; i < MAX_INIT_ARGS; i++)
352 		argv_init[i] = NULL;
353 	return 1;
354 }
355 __setup("rdinit=", rdinit_setup);
356 
357 #ifndef CONFIG_SMP
358 static const unsigned int setup_max_cpus = NR_CPUS;
359 static inline void setup_nr_cpu_ids(void) { }
360 static inline void smp_prepare_cpus(unsigned int maxcpus) { }
361 #endif
362 
363 /*
364  * We need to store the untouched command line for future reference.
365  * We also need to store the touched command line since the parameter
366  * parsing is performed in place, and we should allow a component to
367  * store reference of name/value for future reference.
368  */
369 static void __init setup_command_line(char *command_line)
370 {
371 	saved_command_line =
372 		memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
373 	initcall_command_line =
374 		memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
375 	static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
376 	strcpy(saved_command_line, boot_command_line);
377 	strcpy(static_command_line, command_line);
378 }
379 
380 /*
381  * We need to finalize in a non-__init function or else race conditions
382  * between the root thread and the init thread may cause start_kernel to
383  * be reaped by free_initmem before the root thread has proceeded to
384  * cpu_idle.
385  *
386  * gcc-3.4 accidentally inlines this function, so use noinline.
387  */
388 
389 static __initdata DECLARE_COMPLETION(kthreadd_done);
390 
391 static noinline void __ref rest_init(void)
392 {
393 	struct task_struct *tsk;
394 	int pid;
395 
396 	rcu_scheduler_starting();
397 	/*
398 	 * We need to spawn init first so that it obtains pid 1, however
399 	 * the init task will end up wanting to create kthreads, which, if
400 	 * we schedule it before we create kthreadd, will OOPS.
401 	 */
402 	pid = kernel_thread(kernel_init, NULL, CLONE_FS);
403 	/*
404 	 * Pin init on the boot CPU. Task migration is not properly working
405 	 * until sched_init_smp() has been run. It will set the allowed
406 	 * CPUs for init to the non isolated CPUs.
407 	 */
408 	rcu_read_lock();
409 	tsk = find_task_by_pid_ns(pid, &init_pid_ns);
410 	set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
411 	rcu_read_unlock();
412 
413 	numa_default_policy();
414 	pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
415 	rcu_read_lock();
416 	kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
417 	rcu_read_unlock();
418 
419 	/*
420 	 * Enable might_sleep() and smp_processor_id() checks.
421 	 * They cannot be enabled earlier because with CONFIG_PRREMPT=y
422 	 * kernel_thread() would trigger might_sleep() splats. With
423 	 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
424 	 * already, but it's stuck on the kthreadd_done completion.
425 	 */
426 	system_state = SYSTEM_SCHEDULING;
427 
428 	complete(&kthreadd_done);
429 
430 	/*
431 	 * The boot idle thread must execute schedule()
432 	 * at least once to get things moving:
433 	 */
434 	schedule_preempt_disabled();
435 	/* Call into cpu_idle with preempt disabled */
436 	cpu_startup_entry(CPUHP_ONLINE);
437 }
438 
439 /* Check for early params. */
440 static int __init do_early_param(char *param, char *val,
441 				 const char *unused, void *arg)
442 {
443 	const struct obs_kernel_param *p;
444 
445 	for (p = __setup_start; p < __setup_end; p++) {
446 		if ((p->early && parameq(param, p->str)) ||
447 		    (strcmp(param, "console") == 0 &&
448 		     strcmp(p->str, "earlycon") == 0)
449 		) {
450 			if (p->setup_func(val) != 0)
451 				pr_warn("Malformed early option '%s'\n", param);
452 		}
453 	}
454 	/* We accept everything at this stage. */
455 	return 0;
456 }
457 
458 void __init parse_early_options(char *cmdline)
459 {
460 	parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
461 		   do_early_param);
462 }
463 
464 /* Arch code calls this early on, or if not, just before other parsing. */
465 void __init parse_early_param(void)
466 {
467 	static int done __initdata;
468 	static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
469 
470 	if (done)
471 		return;
472 
473 	/* All fall through to do_early_param. */
474 	strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
475 	parse_early_options(tmp_cmdline);
476 	done = 1;
477 }
478 
479 void __init __weak arch_post_acpi_subsys_init(void) { }
480 
481 void __init __weak smp_setup_processor_id(void)
482 {
483 }
484 
485 # if THREAD_SIZE >= PAGE_SIZE
486 void __init __weak thread_stack_cache_init(void)
487 {
488 }
489 #endif
490 
491 void __init __weak mem_encrypt_init(void) { }
492 
493 /*
494  * Set up kernel memory allocators
495  */
496 static void __init mm_init(void)
497 {
498 	/*
499 	 * page_ext requires contiguous pages,
500 	 * bigger than MAX_ORDER unless SPARSEMEM.
501 	 */
502 	page_ext_init_flatmem();
503 	mem_init();
504 	kmem_cache_init();
505 	pgtable_init();
506 	vmalloc_init();
507 	ioremap_huge_init();
508 }
509 
510 asmlinkage __visible void __init start_kernel(void)
511 {
512 	char *command_line;
513 	char *after_dashes;
514 
515 	set_task_stack_end_magic(&init_task);
516 	smp_setup_processor_id();
517 	debug_objects_early_init();
518 
519 	cgroup_init_early();
520 
521 	local_irq_disable();
522 	early_boot_irqs_disabled = true;
523 
524 	/*
525 	 * Interrupts are still disabled. Do necessary setups, then
526 	 * enable them.
527 	 */
528 	boot_cpu_init();
529 	page_address_init();
530 	pr_notice("%s", linux_banner);
531 	setup_arch(&command_line);
532 	/*
533 	 * Set up the the initial canary and entropy after arch
534 	 * and after adding latent and command line entropy.
535 	 */
536 	add_latent_entropy();
537 	add_device_randomness(command_line, strlen(command_line));
538 	boot_init_stack_canary();
539 	mm_init_cpumask(&init_mm);
540 	setup_command_line(command_line);
541 	setup_nr_cpu_ids();
542 	setup_per_cpu_areas();
543 	boot_cpu_state_init();
544 	smp_prepare_boot_cpu();	/* arch-specific boot-cpu hooks */
545 
546 	build_all_zonelists(NULL);
547 	page_alloc_init();
548 
549 	pr_notice("Kernel command line: %s\n", boot_command_line);
550 	parse_early_param();
551 	after_dashes = parse_args("Booting kernel",
552 				  static_command_line, __start___param,
553 				  __stop___param - __start___param,
554 				  -1, -1, NULL, &unknown_bootoption);
555 	if (!IS_ERR_OR_NULL(after_dashes))
556 		parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
557 			   NULL, set_init_arg);
558 
559 	jump_label_init();
560 
561 	/*
562 	 * These use large bootmem allocations and must precede
563 	 * kmem_cache_init()
564 	 */
565 	setup_log_buf(0);
566 	pidhash_init();
567 	vfs_caches_init_early();
568 	sort_main_extable();
569 	trap_init();
570 	mm_init();
571 
572 	ftrace_init();
573 
574 	/* trace_printk can be enabled here */
575 	early_trace_init();
576 
577 	/*
578 	 * Set up the scheduler prior starting any interrupts (such as the
579 	 * timer interrupt). Full topology setup happens at smp_init()
580 	 * time - but meanwhile we still have a functioning scheduler.
581 	 */
582 	sched_init();
583 	/*
584 	 * Disable preemption - early bootup scheduling is extremely
585 	 * fragile until we cpu_idle() for the first time.
586 	 */
587 	preempt_disable();
588 	if (WARN(!irqs_disabled(),
589 		 "Interrupts were enabled *very* early, fixing it\n"))
590 		local_irq_disable();
591 	radix_tree_init();
592 
593 	/*
594 	 * Allow workqueue creation and work item queueing/cancelling
595 	 * early.  Work item execution depends on kthreads and starts after
596 	 * workqueue_init().
597 	 */
598 	workqueue_init_early();
599 
600 	rcu_init();
601 
602 	/* Trace events are available after this */
603 	trace_init();
604 
605 	context_tracking_init();
606 	/* init some links before init_ISA_irqs() */
607 	early_irq_init();
608 	init_IRQ();
609 	tick_init();
610 	housekeeping_init();
611 	rcu_init_nohz();
612 	init_timers();
613 	hrtimers_init();
614 	softirq_init();
615 	timekeeping_init();
616 	time_init();
617 	sched_clock_postinit();
618 	printk_safe_init();
619 	perf_event_init();
620 	profile_init();
621 	call_function_init();
622 	WARN(!irqs_disabled(), "Interrupts were enabled early\n");
623 	early_boot_irqs_disabled = false;
624 	local_irq_enable();
625 
626 	kmem_cache_init_late();
627 
628 	/*
629 	 * HACK ALERT! This is early. We're enabling the console before
630 	 * we've done PCI setups etc, and console_init() must be aware of
631 	 * this. But we do want output early, in case something goes wrong.
632 	 */
633 	console_init();
634 	if (panic_later)
635 		panic("Too many boot %s vars at `%s'", panic_later,
636 		      panic_param);
637 
638 	lockdep_info();
639 
640 	/*
641 	 * Need to run this when irqs are enabled, because it wants
642 	 * to self-test [hard/soft]-irqs on/off lock inversion bugs
643 	 * too:
644 	 */
645 	locking_selftest();
646 
647 	/*
648 	 * This needs to be called before any devices perform DMA
649 	 * operations that might use the SWIOTLB bounce buffers. It will
650 	 * mark the bounce buffers as decrypted so that their usage will
651 	 * not cause "plain-text" data to be decrypted when accessed.
652 	 */
653 	mem_encrypt_init();
654 
655 #ifdef CONFIG_BLK_DEV_INITRD
656 	if (initrd_start && !initrd_below_start_ok &&
657 	    page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
658 		pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
659 		    page_to_pfn(virt_to_page((void *)initrd_start)),
660 		    min_low_pfn);
661 		initrd_start = 0;
662 	}
663 #endif
664 	page_ext_init();
665 	kmemleak_init();
666 	debug_objects_mem_init();
667 	setup_per_cpu_pageset();
668 	numa_policy_init();
669 	acpi_early_init();
670 	if (late_time_init)
671 		late_time_init();
672 	calibrate_delay();
673 	pidmap_init();
674 	anon_vma_init();
675 #ifdef CONFIG_X86
676 	if (efi_enabled(EFI_RUNTIME_SERVICES))
677 		efi_enter_virtual_mode();
678 #endif
679 #ifdef CONFIG_X86_ESPFIX64
680 	/* Should be run before the first non-init thread is created */
681 	init_espfix_bsp();
682 #endif
683 	thread_stack_cache_init();
684 	cred_init();
685 	fork_init();
686 	proc_caches_init();
687 	buffer_init();
688 	key_init();
689 	security_init();
690 	dbg_late_init();
691 	vfs_caches_init();
692 	pagecache_init();
693 	signals_init();
694 	proc_root_init();
695 	nsfs_init();
696 	cpuset_init();
697 	cgroup_init();
698 	taskstats_init_early();
699 	delayacct_init();
700 
701 	check_bugs();
702 
703 	acpi_subsystem_init();
704 	arch_post_acpi_subsys_init();
705 	sfi_init_late();
706 
707 	if (efi_enabled(EFI_RUNTIME_SERVICES)) {
708 		efi_free_boot_services();
709 	}
710 
711 	/* Do the rest non-__init'ed, we're now alive */
712 	rest_init();
713 }
714 
715 /* Call all constructor functions linked into the kernel. */
716 static void __init do_ctors(void)
717 {
718 #ifdef CONFIG_CONSTRUCTORS
719 	ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
720 
721 	for (; fn < (ctor_fn_t *) __ctors_end; fn++)
722 		(*fn)();
723 #endif
724 }
725 
726 bool initcall_debug;
727 core_param(initcall_debug, initcall_debug, bool, 0644);
728 
729 #ifdef CONFIG_KALLSYMS
730 struct blacklist_entry {
731 	struct list_head next;
732 	char *buf;
733 };
734 
735 static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
736 
737 static int __init initcall_blacklist(char *str)
738 {
739 	char *str_entry;
740 	struct blacklist_entry *entry;
741 
742 	/* str argument is a comma-separated list of functions */
743 	do {
744 		str_entry = strsep(&str, ",");
745 		if (str_entry) {
746 			pr_debug("blacklisting initcall %s\n", str_entry);
747 			entry = alloc_bootmem(sizeof(*entry));
748 			entry->buf = alloc_bootmem(strlen(str_entry) + 1);
749 			strcpy(entry->buf, str_entry);
750 			list_add(&entry->next, &blacklisted_initcalls);
751 		}
752 	} while (str_entry);
753 
754 	return 0;
755 }
756 
757 static bool __init_or_module initcall_blacklisted(initcall_t fn)
758 {
759 	struct blacklist_entry *entry;
760 	char fn_name[KSYM_SYMBOL_LEN];
761 	unsigned long addr;
762 
763 	if (list_empty(&blacklisted_initcalls))
764 		return false;
765 
766 	addr = (unsigned long) dereference_function_descriptor(fn);
767 	sprint_symbol_no_offset(fn_name, addr);
768 
769 	/*
770 	 * fn will be "function_name [module_name]" where [module_name] is not
771 	 * displayed for built-in init functions.  Strip off the [module_name].
772 	 */
773 	strreplace(fn_name, ' ', '\0');
774 
775 	list_for_each_entry(entry, &blacklisted_initcalls, next) {
776 		if (!strcmp(fn_name, entry->buf)) {
777 			pr_debug("initcall %s blacklisted\n", fn_name);
778 			return true;
779 		}
780 	}
781 
782 	return false;
783 }
784 #else
785 static int __init initcall_blacklist(char *str)
786 {
787 	pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
788 	return 0;
789 }
790 
791 static bool __init_or_module initcall_blacklisted(initcall_t fn)
792 {
793 	return false;
794 }
795 #endif
796 __setup("initcall_blacklist=", initcall_blacklist);
797 
798 static int __init_or_module do_one_initcall_debug(initcall_t fn)
799 {
800 	ktime_t calltime, delta, rettime;
801 	unsigned long long duration;
802 	int ret;
803 
804 	printk(KERN_DEBUG "calling  %pF @ %i\n", fn, task_pid_nr(current));
805 	calltime = ktime_get();
806 	ret = fn();
807 	rettime = ktime_get();
808 	delta = ktime_sub(rettime, calltime);
809 	duration = (unsigned long long) ktime_to_ns(delta) >> 10;
810 	printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
811 		 fn, ret, duration);
812 
813 	return ret;
814 }
815 
816 int __init_or_module do_one_initcall(initcall_t fn)
817 {
818 	int count = preempt_count();
819 	int ret;
820 	char msgbuf[64];
821 
822 	if (initcall_blacklisted(fn))
823 		return -EPERM;
824 
825 	if (initcall_debug)
826 		ret = do_one_initcall_debug(fn);
827 	else
828 		ret = fn();
829 
830 	msgbuf[0] = 0;
831 
832 	if (preempt_count() != count) {
833 		sprintf(msgbuf, "preemption imbalance ");
834 		preempt_count_set(count);
835 	}
836 	if (irqs_disabled()) {
837 		strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
838 		local_irq_enable();
839 	}
840 	WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
841 
842 	add_latent_entropy();
843 	return ret;
844 }
845 
846 
847 extern initcall_t __initcall_start[];
848 extern initcall_t __initcall0_start[];
849 extern initcall_t __initcall1_start[];
850 extern initcall_t __initcall2_start[];
851 extern initcall_t __initcall3_start[];
852 extern initcall_t __initcall4_start[];
853 extern initcall_t __initcall5_start[];
854 extern initcall_t __initcall6_start[];
855 extern initcall_t __initcall7_start[];
856 extern initcall_t __initcall_end[];
857 
858 static initcall_t *initcall_levels[] __initdata = {
859 	__initcall0_start,
860 	__initcall1_start,
861 	__initcall2_start,
862 	__initcall3_start,
863 	__initcall4_start,
864 	__initcall5_start,
865 	__initcall6_start,
866 	__initcall7_start,
867 	__initcall_end,
868 };
869 
870 /* Keep these in sync with initcalls in include/linux/init.h */
871 static char *initcall_level_names[] __initdata = {
872 	"early",
873 	"core",
874 	"postcore",
875 	"arch",
876 	"subsys",
877 	"fs",
878 	"device",
879 	"late",
880 };
881 
882 static void __init do_initcall_level(int level)
883 {
884 	initcall_t *fn;
885 
886 	strcpy(initcall_command_line, saved_command_line);
887 	parse_args(initcall_level_names[level],
888 		   initcall_command_line, __start___param,
889 		   __stop___param - __start___param,
890 		   level, level,
891 		   NULL, &repair_env_string);
892 
893 	for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
894 		do_one_initcall(*fn);
895 }
896 
897 static void __init do_initcalls(void)
898 {
899 	int level;
900 
901 	for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
902 		do_initcall_level(level);
903 }
904 
905 /*
906  * Ok, the machine is now initialized. None of the devices
907  * have been touched yet, but the CPU subsystem is up and
908  * running, and memory and process management works.
909  *
910  * Now we can finally start doing some real work..
911  */
912 static void __init do_basic_setup(void)
913 {
914 	cpuset_init_smp();
915 	shmem_init();
916 	driver_init();
917 	init_irq_proc();
918 	do_ctors();
919 	usermodehelper_enable();
920 	do_initcalls();
921 }
922 
923 static void __init do_pre_smp_initcalls(void)
924 {
925 	initcall_t *fn;
926 
927 	for (fn = __initcall_start; fn < __initcall0_start; fn++)
928 		do_one_initcall(*fn);
929 }
930 
931 /*
932  * This function requests modules which should be loaded by default and is
933  * called twice right after initrd is mounted and right before init is
934  * exec'd.  If such modules are on either initrd or rootfs, they will be
935  * loaded before control is passed to userland.
936  */
937 void __init load_default_modules(void)
938 {
939 	load_default_elevator_module();
940 }
941 
942 static int run_init_process(const char *init_filename)
943 {
944 	argv_init[0] = init_filename;
945 	return do_execve(getname_kernel(init_filename),
946 		(const char __user *const __user *)argv_init,
947 		(const char __user *const __user *)envp_init);
948 }
949 
950 static int try_to_run_init_process(const char *init_filename)
951 {
952 	int ret;
953 
954 	ret = run_init_process(init_filename);
955 
956 	if (ret && ret != -ENOENT) {
957 		pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
958 		       init_filename, ret);
959 	}
960 
961 	return ret;
962 }
963 
964 static noinline void __init kernel_init_freeable(void);
965 
966 #if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
967 bool rodata_enabled __ro_after_init = true;
968 static int __init set_debug_rodata(char *str)
969 {
970 	return strtobool(str, &rodata_enabled);
971 }
972 __setup("rodata=", set_debug_rodata);
973 #endif
974 
975 #ifdef CONFIG_STRICT_KERNEL_RWX
976 static void mark_readonly(void)
977 {
978 	if (rodata_enabled) {
979 		mark_rodata_ro();
980 		rodata_test();
981 	} else
982 		pr_info("Kernel memory protection disabled.\n");
983 }
984 #else
985 static inline void mark_readonly(void)
986 {
987 	pr_warn("This architecture does not have kernel memory protection.\n");
988 }
989 #endif
990 
991 static int __ref kernel_init(void *unused)
992 {
993 	int ret;
994 
995 	kernel_init_freeable();
996 	/* need to finish all async __init code before freeing the memory */
997 	async_synchronize_full();
998 	ftrace_free_init_mem();
999 	free_initmem();
1000 	mark_readonly();
1001 	system_state = SYSTEM_RUNNING;
1002 	numa_default_policy();
1003 
1004 	rcu_end_inkernel_boot();
1005 
1006 	if (ramdisk_execute_command) {
1007 		ret = run_init_process(ramdisk_execute_command);
1008 		if (!ret)
1009 			return 0;
1010 		pr_err("Failed to execute %s (error %d)\n",
1011 		       ramdisk_execute_command, ret);
1012 	}
1013 
1014 	/*
1015 	 * We try each of these until one succeeds.
1016 	 *
1017 	 * The Bourne shell can be used instead of init if we are
1018 	 * trying to recover a really broken machine.
1019 	 */
1020 	if (execute_command) {
1021 		ret = run_init_process(execute_command);
1022 		if (!ret)
1023 			return 0;
1024 		panic("Requested init %s failed (error %d).",
1025 		      execute_command, ret);
1026 	}
1027 	if (!try_to_run_init_process("/sbin/init") ||
1028 	    !try_to_run_init_process("/etc/init") ||
1029 	    !try_to_run_init_process("/bin/init") ||
1030 	    !try_to_run_init_process("/bin/sh"))
1031 		return 0;
1032 
1033 	panic("No working init found.  Try passing init= option to kernel. "
1034 	      "See Linux Documentation/admin-guide/init.rst for guidance.");
1035 }
1036 
1037 static noinline void __init kernel_init_freeable(void)
1038 {
1039 	/*
1040 	 * Wait until kthreadd is all set-up.
1041 	 */
1042 	wait_for_completion(&kthreadd_done);
1043 
1044 	/* Now the scheduler is fully set up and can do blocking allocations */
1045 	gfp_allowed_mask = __GFP_BITS_MASK;
1046 
1047 	/*
1048 	 * init can allocate pages on any node
1049 	 */
1050 	set_mems_allowed(node_states[N_MEMORY]);
1051 
1052 	cad_pid = task_pid(current);
1053 
1054 	smp_prepare_cpus(setup_max_cpus);
1055 
1056 	workqueue_init();
1057 
1058 	init_mm_internals();
1059 
1060 	do_pre_smp_initcalls();
1061 	lockup_detector_init();
1062 
1063 	smp_init();
1064 	sched_init_smp();
1065 
1066 	page_alloc_init_late();
1067 
1068 	do_basic_setup();
1069 
1070 	/* Open the /dev/console on the rootfs, this should never fail */
1071 	if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1072 		pr_err("Warning: unable to open an initial console.\n");
1073 
1074 	(void) sys_dup(0);
1075 	(void) sys_dup(0);
1076 	/*
1077 	 * check if there is an early userspace init.  If yes, let it do all
1078 	 * the work
1079 	 */
1080 
1081 	if (!ramdisk_execute_command)
1082 		ramdisk_execute_command = "/init";
1083 
1084 	if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
1085 		ramdisk_execute_command = NULL;
1086 		prepare_namespace();
1087 	}
1088 
1089 	/*
1090 	 * Ok, we have completed the initial bootup, and
1091 	 * we're essentially up and running. Get rid of the
1092 	 * initmem segments and start the user-mode stuff..
1093 	 *
1094 	 * rootfs is available now, try loading the public keys
1095 	 * and default modules
1096 	 */
1097 
1098 	integrity_load_keys();
1099 	load_default_modules();
1100 }
1101