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