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
105static int kernel_init(void *);
106
107extern void init_IRQ(void);
108extern 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 */
117bool early_boot_irqs_disabled __read_mostly;
118
119enum system_states system_state __read_mostly;
120EXPORT_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
128extern void time_init(void);
129/* Default late time init is NULL. archs can override this later. */
130void (*__initdata late_time_init)(void);
131
132/* Untouched command line saved by arch-specific code. */
133char __initdata boot_command_line[COMMAND_LINE_SIZE];
134/* Untouched saved command line (eg. for /proc) */
135char *saved_command_line;
136/* Command line for parameter parsing */
137static char *static_command_line;
138/* Command line for per-initcall parameter parsing */
139static char *initcall_command_line;
140
141static char *execute_command;
142static 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 */
148bool static_key_initialized __read_mostly;
149EXPORT_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 */
160unsigned int reset_devices;
161EXPORT_SYMBOL(reset_devices);
162
163static 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
171static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
172const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
173static const char *panic_later, *panic_param;
174
175extern const struct obs_kernel_param __setup_start[], __setup_end[];
176
177static 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 */
210unsigned long loops_per_jiffy = (1<<12);
211EXPORT_SYMBOL(loops_per_jiffy);
212
213static int __init debug_kernel(char *str)
214{
215 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
216 return 0;
217}
218
219static int __init quiet_kernel(char *str)
220{
221 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
222 return 0;
223}
224
225early_param("debug", debug_kernel);
226early_param("quiet", quiet_kernel);
227
228static 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
245early_param("loglevel", loglevel);
246
247/* Change NUL term back to "=", to make "param" the whole string. */
248static 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. */
266static 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 */
291static 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
333static 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
350static 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
363static const unsigned int setup_max_cpus = NR_CPUS;
364static inline void setup_nr_cpu_ids(void) { }
365static 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 */
374static 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
403static __initdata DECLARE_COMPLETION(kthreadd_done);
404
405noinline 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_PREEMPT=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. */
454static 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
472void __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. */
479void __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
493void __init __weak arch_post_acpi_subsys_init(void) { }
494
495void __init __weak smp_setup_processor_id(void)
496{
497}
498
499# if THREAD_SIZE >= PAGE_SIZE
500void __init __weak thread_stack_cache_init(void)
501{
502}
503#endif
504
505void __init __weak mem_encrypt_init(void) { }
506
507bool initcall_debug;
508core_param(initcall_debug, initcall_debug, bool, 0644);
509
510#ifdef TRACEPOINTS_ENABLED
511static void __init initcall_debug_enable(void);
512#else
513static inline void initcall_debug_enable(void)
514{
515}
516#endif
517
518/*
519 * Set up kernel memory allocators
520 */
521static void __init mm_init(void)
522{
523 /*
524 * page_ext requires contiguous pages,
525 * bigger than MAX_ORDER unless SPARSEMEM.
526 */
527 page_ext_init_flatmem();
528 mem_init();
529 kmem_cache_init();
530 pgtable_init();
531 debug_objects_mem_init();
532 vmalloc_init();
533 ioremap_huge_init();
534 /* Should be run before the first non-init thread is created */
535 init_espfix_bsp();
536 /* Should be run after espfix64 is set up. */
537 pti_init();
538}
539
540void __init __weak arch_call_rest_init(void)
541{
542 rest_init();
543}
544
545asmlinkage __visible void __init start_kernel(void)
546{
547 char *command_line;
548 char *after_dashes;
549
550 set_task_stack_end_magic(&init_task);
551 smp_setup_processor_id();
552 debug_objects_early_init();
553
554 cgroup_init_early();
555
556 local_irq_disable();
557 early_boot_irqs_disabled = true;
558
559 /*
560 * Interrupts are still disabled. Do necessary setups, then
561 * enable them.
562 */
563 boot_cpu_init();
564 page_address_init();
565 pr_notice("%s", linux_banner);
566 setup_arch(&command_line);
567 /*
568 * Set up the the initial canary and entropy after arch
569 * and after adding latent and command line entropy.
570 */
571 add_latent_entropy();
572 add_device_randomness(command_line, strlen(command_line));
573 boot_init_stack_canary();
574 mm_init_cpumask(&init_mm);
575 setup_command_line(command_line);
576 setup_nr_cpu_ids();
577 setup_per_cpu_areas();
578 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
579 boot_cpu_hotplug_init();
580
581 build_all_zonelists(NULL);
582 page_alloc_init();
583
584 pr_notice("Kernel command line: %s\n", boot_command_line);
585 parse_early_param();
586 after_dashes = parse_args("Booting kernel",
587 static_command_line, __start___param,
588 __stop___param - __start___param,
589 -1, -1, NULL, &unknown_bootoption);
590 if (!IS_ERR_OR_NULL(after_dashes))
591 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
592 NULL, set_init_arg);
593
594 jump_label_init();
595
596 /*
597 * These use large bootmem allocations and must precede
598 * kmem_cache_init()
599 */
600 setup_log_buf(0);
601 vfs_caches_init_early();
602 sort_main_extable();
603 trap_init();
604 mm_init();
605
606 ftrace_init();
607
608 /* trace_printk can be enabled here */
609 early_trace_init();
610
611 /*
612 * Set up the scheduler prior starting any interrupts (such as the
613 * timer interrupt). Full topology setup happens at smp_init()
614 * time - but meanwhile we still have a functioning scheduler.
615 */
616 sched_init();
617 /*
618 * Disable preemption - early bootup scheduling is extremely
619 * fragile until we cpu_idle() for the first time.
620 */
621 preempt_disable();
622 if (WARN(!irqs_disabled(),
623 "Interrupts were enabled *very* early, fixing it\n"))
624 local_irq_disable();
625 radix_tree_init();
626
627 /*
628 * Set up housekeeping before setting up workqueues to allow the unbound
629 * workqueue to take non-housekeeping into account.
630 */
631 housekeeping_init();
632
633 /*
634 * Allow workqueue creation and work item queueing/cancelling
635 * early. Work item execution depends on kthreads and starts after
636 * workqueue_init().
637 */
638 workqueue_init_early();
639
640 rcu_init();
641
642 /* Trace events are available after this */
643 trace_init();
644
645 if (initcall_debug)
646 initcall_debug_enable();
647
648 context_tracking_init();
649 /* init some links before init_ISA_irqs() */
650 early_irq_init();
651 init_IRQ();
652 tick_init();
653 rcu_init_nohz();
654 init_timers();
655 hrtimers_init();
656 softirq_init();
657 timekeeping_init();
658 time_init();
659 printk_safe_init();
660 perf_event_init();
661 profile_init();
662 call_function_init();
663 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
664
665 early_boot_irqs_disabled = false;
666 local_irq_enable();
667
668 kmem_cache_init_late();
669
670 /*
671 * HACK ALERT! This is early. We're enabling the console before
672 * we've done PCI setups etc, and console_init() must be aware of
673 * this. But we do want output early, in case something goes wrong.
674 */
675 console_init();
676 if (panic_later)
677 panic("Too many boot %s vars at `%s'", panic_later,
678 panic_param);
679
680 lockdep_init();
681
682 /*
683 * Need to run this when irqs are enabled, because it wants
684 * to self-test [hard/soft]-irqs on/off lock inversion bugs
685 * too:
686 */
687 locking_selftest();
688
689 /*
690 * This needs to be called before any devices perform DMA
691 * operations that might use the SWIOTLB bounce buffers. It will
692 * mark the bounce buffers as decrypted so that their usage will
693 * not cause "plain-text" data to be decrypted when accessed.
694 */
695 mem_encrypt_init();
696
697#ifdef CONFIG_BLK_DEV_INITRD
698 if (initrd_start && !initrd_below_start_ok &&
699 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
700 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
701 page_to_pfn(virt_to_page((void *)initrd_start)),
702 min_low_pfn);
703 initrd_start = 0;
704 }
705#endif
706 kmemleak_init();
707 setup_per_cpu_pageset();
708 numa_policy_init();
709 acpi_early_init();
710 if (late_time_init)
711 late_time_init();
712 sched_clock_init();
713 calibrate_delay();
714 pid_idr_init();
715 anon_vma_init();
716#ifdef CONFIG_X86
717 if (efi_enabled(EFI_RUNTIME_SERVICES))
718 efi_enter_virtual_mode();
719#endif
720 thread_stack_cache_init();
721 cred_init();
722 fork_init();
723 proc_caches_init();
724 uts_ns_init();
725 buffer_init();
726 key_init();
727 security_init();
728 dbg_late_init();
729 vfs_caches_init();
730 pagecache_init();
731 signals_init();
732 seq_file_init();
733 proc_root_init();
734 nsfs_init();
735 cpuset_init();
736 cgroup_init();
737 taskstats_init_early();
738 delayacct_init();
739
740 check_bugs();
741
742 acpi_subsystem_init();
743 arch_post_acpi_subsys_init();
744 sfi_init_late();
745
746 /* Do the rest non-__init'ed, we're now alive */
747 arch_call_rest_init();
748}
749
750/* Call all constructor functions linked into the kernel. */
751static void __init do_ctors(void)
752{
753#ifdef CONFIG_CONSTRUCTORS
754 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
755
756 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
757 (*fn)();
758#endif
759}
760
761#ifdef CONFIG_KALLSYMS
762struct blacklist_entry {
763 struct list_head next;
764 char *buf;
765};
766
767static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
768
769static int __init initcall_blacklist(char *str)
770{
771 char *str_entry;
772 struct blacklist_entry *entry;
773
774 /* str argument is a comma-separated list of functions */
775 do {
776 str_entry = strsep(&str, ",");
777 if (str_entry) {
778 pr_debug("blacklisting initcall %s\n", str_entry);
779 entry = memblock_alloc(sizeof(*entry),
780 SMP_CACHE_BYTES);
781 if (!entry)
782 panic("%s: Failed to allocate %zu bytes\n",
783 __func__, sizeof(*entry));
784 entry->buf = memblock_alloc(strlen(str_entry) + 1,
785 SMP_CACHE_BYTES);
786 if (!entry->buf)
787 panic("%s: Failed to allocate %zu bytes\n",
788 __func__, strlen(str_entry) + 1);
789 strcpy(entry->buf, str_entry);
790 list_add(&entry->next, &blacklisted_initcalls);
791 }
792 } while (str_entry);
793
794 return 0;
795}
796
797static bool __init_or_module initcall_blacklisted(initcall_t fn)
798{
799 struct blacklist_entry *entry;
800 char fn_name[KSYM_SYMBOL_LEN];
801 unsigned long addr;
802
803 if (list_empty(&blacklisted_initcalls))
804 return false;
805
806 addr = (unsigned long) dereference_function_descriptor(fn);
807 sprint_symbol_no_offset(fn_name, addr);
808
809 /*
810 * fn will be "function_name [module_name]" where [module_name] is not
811 * displayed for built-in init functions. Strip off the [module_name].
812 */
813 strreplace(fn_name, ' ', '\0');
814
815 list_for_each_entry(entry, &blacklisted_initcalls, next) {
816 if (!strcmp(fn_name, entry->buf)) {
817 pr_debug("initcall %s blacklisted\n", fn_name);
818 return true;
819 }
820 }
821
822 return false;
823}
824#else
825static int __init initcall_blacklist(char *str)
826{
827 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
828 return 0;
829}
830
831static bool __init_or_module initcall_blacklisted(initcall_t fn)
832{
833 return false;
834}
835#endif
836__setup("initcall_blacklist=", initcall_blacklist);
837
838static __init_or_module void
839trace_initcall_start_cb(void *data, initcall_t fn)
840{
841 ktime_t *calltime = (ktime_t *)data;
842
843 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
844 *calltime = ktime_get();
845}
846
847static __init_or_module void
848trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
849{
850 ktime_t *calltime = (ktime_t *)data;
851 ktime_t delta, rettime;
852 unsigned long long duration;
853
854 rettime = ktime_get();
855 delta = ktime_sub(rettime, *calltime);
856 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
857 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n",
858 fn, ret, duration);
859}
860
861static ktime_t initcall_calltime;
862
863#ifdef TRACEPOINTS_ENABLED
864static void __init initcall_debug_enable(void)
865{
866 int ret;
867
868 ret = register_trace_initcall_start(trace_initcall_start_cb,
869 &initcall_calltime);
870 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
871 &initcall_calltime);
872 WARN(ret, "Failed to register initcall tracepoints\n");
873}
874# define do_trace_initcall_start trace_initcall_start
875# define do_trace_initcall_finish trace_initcall_finish
876#else
877static inline void do_trace_initcall_start(initcall_t fn)
878{
879 if (!initcall_debug)
880 return;
881 trace_initcall_start_cb(&initcall_calltime, fn);
882}
883static inline void do_trace_initcall_finish(initcall_t fn, int ret)
884{
885 if (!initcall_debug)
886 return;
887 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
888}
889#endif /* !TRACEPOINTS_ENABLED */
890
891int __init_or_module do_one_initcall(initcall_t fn)
892{
893 int count = preempt_count();
894 char msgbuf[64];
895 int ret;
896
897 if (initcall_blacklisted(fn))
898 return -EPERM;
899
900 do_trace_initcall_start(fn);
901 ret = fn();
902 do_trace_initcall_finish(fn, ret);
903
904 msgbuf[0] = 0;
905
906 if (preempt_count() != count) {
907 sprintf(msgbuf, "preemption imbalance ");
908 preempt_count_set(count);
909 }
910 if (irqs_disabled()) {
911 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
912 local_irq_enable();
913 }
914 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
915
916 add_latent_entropy();
917 return ret;
918}
919
920
921extern initcall_entry_t __initcall_start[];
922extern initcall_entry_t __initcall0_start[];
923extern initcall_entry_t __initcall1_start[];
924extern initcall_entry_t __initcall2_start[];
925extern initcall_entry_t __initcall3_start[];
926extern initcall_entry_t __initcall4_start[];
927extern initcall_entry_t __initcall5_start[];
928extern initcall_entry_t __initcall6_start[];
929extern initcall_entry_t __initcall7_start[];
930extern initcall_entry_t __initcall_end[];
931
932static initcall_entry_t *initcall_levels[] __initdata = {
933 __initcall0_start,
934 __initcall1_start,
935 __initcall2_start,
936 __initcall3_start,
937 __initcall4_start,
938 __initcall5_start,
939 __initcall6_start,
940 __initcall7_start,
941 __initcall_end,
942};
943
944/* Keep these in sync with initcalls in include/linux/init.h */
945static const char *initcall_level_names[] __initdata = {
946 "pure",
947 "core",
948 "postcore",
949 "arch",
950 "subsys",
951 "fs",
952 "device",
953 "late",
954};
955
956static void __init do_initcall_level(int level)
957{
958 initcall_entry_t *fn;
959
960 strcpy(initcall_command_line, saved_command_line);
961 parse_args(initcall_level_names[level],
962 initcall_command_line, __start___param,
963 __stop___param - __start___param,
964 level, level,
965 NULL, &repair_env_string);
966
967 trace_initcall_level(initcall_level_names[level]);
968 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
969 do_one_initcall(initcall_from_entry(fn));
970}
971
972static void __init do_initcalls(void)
973{
974 int level;
975
976 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
977 do_initcall_level(level);
978}
979
980/*
981 * Ok, the machine is now initialized. None of the devices
982 * have been touched yet, but the CPU subsystem is up and
983 * running, and memory and process management works.
984 *
985 * Now we can finally start doing some real work..
986 */
987static void __init do_basic_setup(void)
988{
989 cpuset_init_smp();
990 shmem_init();
991 driver_init();
992 init_irq_proc();
993 do_ctors();
994 usermodehelper_enable();
995 do_initcalls();
996}
997
998static void __init do_pre_smp_initcalls(void)
999{
1000 initcall_entry_t *fn;
1001
1002 trace_initcall_level("early");
1003 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1004 do_one_initcall(initcall_from_entry(fn));
1005}
1006
1007static int run_init_process(const char *init_filename)
1008{
1009 argv_init[0] = init_filename;
1010 pr_info("Run %s as init process\n", init_filename);
1011 return do_execve(getname_kernel(init_filename),
1012 (const char __user *const __user *)argv_init,
1013 (const char __user *const __user *)envp_init);
1014}
1015
1016static int try_to_run_init_process(const char *init_filename)
1017{
1018 int ret;
1019
1020 ret = run_init_process(init_filename);
1021
1022 if (ret && ret != -ENOENT) {
1023 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1024 init_filename, ret);
1025 }
1026
1027 return ret;
1028}
1029
1030static noinline void __init kernel_init_freeable(void);
1031
1032#if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1033bool rodata_enabled __ro_after_init = true;
1034static int __init set_debug_rodata(char *str)
1035{
1036 return strtobool(str, &rodata_enabled);
1037}
1038__setup("rodata=", set_debug_rodata);
1039#endif
1040
1041#ifdef CONFIG_STRICT_KERNEL_RWX
1042static void mark_readonly(void)
1043{
1044 if (rodata_enabled) {
1045 /*
1046 * load_module() results in W+X mappings, which are cleaned
1047 * up with call_rcu(). Let's make sure that queued work is
1048 * flushed so that we don't hit false positives looking for
1049 * insecure pages which are W+X.
1050 */
1051 rcu_barrier();
1052 mark_rodata_ro();
1053 rodata_test();
1054 } else
1055 pr_info("Kernel memory protection disabled.\n");
1056}
1057#else
1058static inline void mark_readonly(void)
1059{
1060 pr_warn("This architecture does not have kernel memory protection.\n");
1061}
1062#endif
1063
1064static int __ref kernel_init(void *unused)
1065{
1066 int ret;
1067
1068 kernel_init_freeable();
1069 /* need to finish all async __init code before freeing the memory */
1070 async_synchronize_full();
1071 ftrace_free_init_mem();
1072 free_initmem();
1073 mark_readonly();
1074
1075 /*
1076 * Kernel mappings are now finalized - update the userspace page-table
1077 * to finalize PTI.
1078 */
1079 pti_finalize();
1080
1081 system_state = SYSTEM_RUNNING;
1082 numa_default_policy();
1083
1084 rcu_end_inkernel_boot();
1085
1086 if (ramdisk_execute_command) {
1087 ret = run_init_process(ramdisk_execute_command);
1088 if (!ret)
1089 return 0;
1090 pr_err("Failed to execute %s (error %d)\n",
1091 ramdisk_execute_command, ret);
1092 }
1093
1094 /*
1095 * We try each of these until one succeeds.
1096 *
1097 * The Bourne shell can be used instead of init if we are
1098 * trying to recover a really broken machine.
1099 */
1100 if (execute_command) {
1101 ret = run_init_process(execute_command);
1102 if (!ret)
1103 return 0;
1104 panic("Requested init %s failed (error %d).",
1105 execute_command, ret);
1106 }
1107 if (!try_to_run_init_process("/sbin/init") ||
1108 !try_to_run_init_process("/etc/init") ||
1109 !try_to_run_init_process("/bin/init") ||
1110 !try_to_run_init_process("/bin/sh"))
1111 return 0;
1112
1113 panic("No working init found. Try passing init= option to kernel. "
1114 "See Linux Documentation/admin-guide/init.rst for guidance.");
1115}
1116
1117static noinline void __init kernel_init_freeable(void)
1118{
1119 /*
1120 * Wait until kthreadd is all set-up.
1121 */
1122 wait_for_completion(&kthreadd_done);
1123
1124 /* Now the scheduler is fully set up and can do blocking allocations */
1125 gfp_allowed_mask = __GFP_BITS_MASK;
1126
1127 /*
1128 * init can allocate pages on any node
1129 */
1130 set_mems_allowed(node_states[N_MEMORY]);
1131
1132 cad_pid = task_pid(current);
1133
1134 smp_prepare_cpus(setup_max_cpus);
1135
1136 workqueue_init();
1137
1138 init_mm_internals();
1139
1140 do_pre_smp_initcalls();
1141 lockup_detector_init();
1142
1143 smp_init();
1144 sched_init_smp();
1145
1146 page_alloc_init_late();
1147 /* Initialize page ext after all struct pages are initialized. */
1148 page_ext_init();
1149
1150 do_basic_setup();
1151
1152 /* Open the /dev/console on the rootfs, this should never fail */
1153 if (ksys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
1154 pr_err("Warning: unable to open an initial console.\n");
1155
1156 (void) ksys_dup(0);
1157 (void) ksys_dup(0);
1158 /*
1159 * check if there is an early userspace init. If yes, let it do all
1160 * the work
1161 */
1162
1163 if (!ramdisk_execute_command)
1164 ramdisk_execute_command = "/init";
1165
1166 if (ksys_access((const char __user *)
1167 ramdisk_execute_command, 0) != 0) {
1168 ramdisk_execute_command = NULL;
1169 prepare_namespace();
1170 }
1171
1172 /*
1173 * Ok, we have completed the initial bootup, and
1174 * we're essentially up and running. Get rid of the
1175 * initmem segments and start the user-mode stuff..
1176 *
1177 * rootfs is available now, try loading the public keys
1178 * and default modules
1179 */
1180
1181 integrity_load_keys();
1182}
1183