Loading...
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#include <linux/types.h>
13#include <linux/module.h>
14#include <linux/proc_fs.h>
15#include <linux/kernel.h>
16#include <linux/syscalls.h>
17#include <linux/stackprotector.h>
18#include <linux/string.h>
19#include <linux/ctype.h>
20#include <linux/delay.h>
21#include <linux/ioport.h>
22#include <linux/init.h>
23#include <linux/initrd.h>
24#include <linux/bootmem.h>
25#include <linux/acpi.h>
26#include <linux/tty.h>
27#include <linux/percpu.h>
28#include <linux/kmod.h>
29#include <linux/vmalloc.h>
30#include <linux/kernel_stat.h>
31#include <linux/start_kernel.h>
32#include <linux/security.h>
33#include <linux/smp.h>
34#include <linux/profile.h>
35#include <linux/rcupdate.h>
36#include <linux/moduleparam.h>
37#include <linux/kallsyms.h>
38#include <linux/writeback.h>
39#include <linux/cpu.h>
40#include <linux/cpuset.h>
41#include <linux/cgroup.h>
42#include <linux/efi.h>
43#include <linux/tick.h>
44#include <linux/interrupt.h>
45#include <linux/taskstats_kern.h>
46#include <linux/delayacct.h>
47#include <linux/unistd.h>
48#include <linux/rmap.h>
49#include <linux/mempolicy.h>
50#include <linux/key.h>
51#include <linux/buffer_head.h>
52#include <linux/page_cgroup.h>
53#include <linux/debug_locks.h>
54#include <linux/debugobjects.h>
55#include <linux/lockdep.h>
56#include <linux/kmemleak.h>
57#include <linux/pid_namespace.h>
58#include <linux/device.h>
59#include <linux/kthread.h>
60#include <linux/sched.h>
61#include <linux/signal.h>
62#include <linux/idr.h>
63#include <linux/kgdb.h>
64#include <linux/ftrace.h>
65#include <linux/async.h>
66#include <linux/kmemcheck.h>
67#include <linux/sfi.h>
68#include <linux/shmem_fs.h>
69#include <linux/slab.h>
70#include <linux/perf_event.h>
71
72#include <asm/io.h>
73#include <asm/bugs.h>
74#include <asm/setup.h>
75#include <asm/sections.h>
76#include <asm/cacheflush.h>
77
78#ifdef CONFIG_X86_LOCAL_APIC
79#include <asm/smp.h>
80#endif
81
82static int kernel_init(void *);
83
84extern void init_IRQ(void);
85extern void fork_init(unsigned long);
86extern void mca_init(void);
87extern void sbus_init(void);
88extern void prio_tree_init(void);
89extern void radix_tree_init(void);
90extern void free_initmem(void);
91#ifndef CONFIG_DEBUG_RODATA
92static inline void mark_rodata_ro(void) { }
93#endif
94
95#ifdef CONFIG_TC
96extern void tc_init(void);
97#endif
98
99/*
100 * Debug helper: via this flag we know that we are in 'early bootup code'
101 * where only the boot processor is running with IRQ disabled. This means
102 * two things - IRQ must not be enabled before the flag is cleared and some
103 * operations which are not allowed with IRQ disabled are allowed while the
104 * flag is set.
105 */
106bool early_boot_irqs_disabled __read_mostly;
107
108enum system_states system_state __read_mostly;
109EXPORT_SYMBOL(system_state);
110
111/*
112 * Boot command-line arguments
113 */
114#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
115#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
116
117extern void time_init(void);
118/* Default late time init is NULL. archs can override this later. */
119void (*__initdata late_time_init)(void);
120extern void softirq_init(void);
121
122/* Untouched command line saved by arch-specific code. */
123char __initdata boot_command_line[COMMAND_LINE_SIZE];
124/* Untouched saved command line (eg. for /proc) */
125char *saved_command_line;
126/* Command line for parameter parsing */
127static char *static_command_line;
128
129static char *execute_command;
130static char *ramdisk_execute_command;
131
132/*
133 * If set, this is an indication to the drivers that reset the underlying
134 * device before going ahead with the initialization otherwise driver might
135 * rely on the BIOS and skip the reset operation.
136 *
137 * This is useful if kernel is booting in an unreliable environment.
138 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
139 * skipped and devices will be in unknown state.
140 */
141unsigned int reset_devices;
142EXPORT_SYMBOL(reset_devices);
143
144static int __init set_reset_devices(char *str)
145{
146 reset_devices = 1;
147 return 1;
148}
149
150__setup("reset_devices", set_reset_devices);
151
152static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
153const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
154static const char *panic_later, *panic_param;
155
156extern const struct obs_kernel_param __setup_start[], __setup_end[];
157
158static int __init obsolete_checksetup(char *line)
159{
160 const struct obs_kernel_param *p;
161 int had_early_param = 0;
162
163 p = __setup_start;
164 do {
165 int n = strlen(p->str);
166 if (!strncmp(line, p->str, n)) {
167 if (p->early) {
168 /* Already done in parse_early_param?
169 * (Needs exact match on param part).
170 * Keep iterating, as we can have early
171 * params and __setups of same names 8( */
172 if (line[n] == '\0' || line[n] == '=')
173 had_early_param = 1;
174 } else if (!p->setup_func) {
175 printk(KERN_WARNING "Parameter %s is obsolete,"
176 " ignored\n", p->str);
177 return 1;
178 } else if (p->setup_func(line + n))
179 return 1;
180 }
181 p++;
182 } while (p < __setup_end);
183
184 return had_early_param;
185}
186
187/*
188 * This should be approx 2 Bo*oMips to start (note initial shift), and will
189 * still work even if initially too large, it will just take slightly longer
190 */
191unsigned long loops_per_jiffy = (1<<12);
192
193EXPORT_SYMBOL(loops_per_jiffy);
194
195static int __init debug_kernel(char *str)
196{
197 console_loglevel = 10;
198 return 0;
199}
200
201static int __init quiet_kernel(char *str)
202{
203 console_loglevel = 4;
204 return 0;
205}
206
207early_param("debug", debug_kernel);
208early_param("quiet", quiet_kernel);
209
210static int __init loglevel(char *str)
211{
212 int newlevel;
213
214 /*
215 * Only update loglevel value when a correct setting was passed,
216 * to prevent blind crashes (when loglevel being set to 0) that
217 * are quite hard to debug
218 */
219 if (get_option(&str, &newlevel)) {
220 console_loglevel = newlevel;
221 return 0;
222 }
223
224 return -EINVAL;
225}
226
227early_param("loglevel", loglevel);
228
229/*
230 * Unknown boot options get handed to init, unless they look like
231 * unused parameters (modprobe will find them in /proc/cmdline).
232 */
233static int __init unknown_bootoption(char *param, char *val)
234{
235 /* Change NUL term back to "=", to make "param" the whole string. */
236 if (val) {
237 /* param=val or param="val"? */
238 if (val == param+strlen(param)+1)
239 val[-1] = '=';
240 else if (val == param+strlen(param)+2) {
241 val[-2] = '=';
242 memmove(val-1, val, strlen(val)+1);
243 val--;
244 } else
245 BUG();
246 }
247
248 /* Handle obsolete-style parameters */
249 if (obsolete_checksetup(param))
250 return 0;
251
252 /* Unused module parameter. */
253 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
254 return 0;
255
256 if (panic_later)
257 return 0;
258
259 if (val) {
260 /* Environment option */
261 unsigned int i;
262 for (i = 0; envp_init[i]; i++) {
263 if (i == MAX_INIT_ENVS) {
264 panic_later = "Too many boot env vars at `%s'";
265 panic_param = param;
266 }
267 if (!strncmp(param, envp_init[i], val - param))
268 break;
269 }
270 envp_init[i] = param;
271 } else {
272 /* Command line option */
273 unsigned int i;
274 for (i = 0; argv_init[i]; i++) {
275 if (i == MAX_INIT_ARGS) {
276 panic_later = "Too many boot init vars at `%s'";
277 panic_param = param;
278 }
279 }
280 argv_init[i] = param;
281 }
282 return 0;
283}
284
285#ifdef CONFIG_DEBUG_PAGEALLOC
286int __read_mostly debug_pagealloc_enabled = 0;
287#endif
288
289static int __init init_setup(char *str)
290{
291 unsigned int i;
292
293 execute_command = str;
294 /*
295 * In case LILO is going to boot us with default command line,
296 * it prepends "auto" before the whole cmdline which makes
297 * the shell think it should execute a script with such name.
298 * So we ignore all arguments entered _before_ init=... [MJ]
299 */
300 for (i = 1; i < MAX_INIT_ARGS; i++)
301 argv_init[i] = NULL;
302 return 1;
303}
304__setup("init=", init_setup);
305
306static int __init rdinit_setup(char *str)
307{
308 unsigned int i;
309
310 ramdisk_execute_command = str;
311 /* See "auto" comment in init_setup */
312 for (i = 1; i < MAX_INIT_ARGS; i++)
313 argv_init[i] = NULL;
314 return 1;
315}
316__setup("rdinit=", rdinit_setup);
317
318#ifndef CONFIG_SMP
319static const unsigned int setup_max_cpus = NR_CPUS;
320#ifdef CONFIG_X86_LOCAL_APIC
321static void __init smp_init(void)
322{
323 APIC_init_uniprocessor();
324}
325#else
326#define smp_init() do { } while (0)
327#endif
328
329static inline void setup_nr_cpu_ids(void) { }
330static inline void smp_prepare_cpus(unsigned int maxcpus) { }
331#endif
332
333/*
334 * We need to store the untouched command line for future reference.
335 * We also need to store the touched command line since the parameter
336 * parsing is performed in place, and we should allow a component to
337 * store reference of name/value for future reference.
338 */
339static void __init setup_command_line(char *command_line)
340{
341 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
342 static_command_line = alloc_bootmem(strlen (command_line)+1);
343 strcpy (saved_command_line, boot_command_line);
344 strcpy (static_command_line, command_line);
345}
346
347/*
348 * We need to finalize in a non-__init function or else race conditions
349 * between the root thread and the init thread may cause start_kernel to
350 * be reaped by free_initmem before the root thread has proceeded to
351 * cpu_idle.
352 *
353 * gcc-3.4 accidentally inlines this function, so use noinline.
354 */
355
356static __initdata DECLARE_COMPLETION(kthreadd_done);
357
358static noinline void __init_refok rest_init(void)
359{
360 int pid;
361
362 rcu_scheduler_starting();
363 /*
364 * We need to spawn init first so that it obtains pid 1, however
365 * the init task will end up wanting to create kthreads, which, if
366 * we schedule it before we create kthreadd, will OOPS.
367 */
368 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
369 numa_default_policy();
370 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
371 rcu_read_lock();
372 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
373 rcu_read_unlock();
374 complete(&kthreadd_done);
375
376 /*
377 * The boot idle thread must execute schedule()
378 * at least once to get things moving:
379 */
380 init_idle_bootup_task(current);
381 preempt_enable_no_resched();
382 schedule();
383
384 /* Call into cpu_idle with preempt disabled */
385 preempt_disable();
386 cpu_idle();
387}
388
389/* Check for early params. */
390static int __init do_early_param(char *param, char *val)
391{
392 const struct obs_kernel_param *p;
393
394 for (p = __setup_start; p < __setup_end; p++) {
395 if ((p->early && strcmp(param, p->str) == 0) ||
396 (strcmp(param, "console") == 0 &&
397 strcmp(p->str, "earlycon") == 0)
398 ) {
399 if (p->setup_func(val) != 0)
400 printk(KERN_WARNING
401 "Malformed early option '%s'\n", param);
402 }
403 }
404 /* We accept everything at this stage. */
405 return 0;
406}
407
408void __init parse_early_options(char *cmdline)
409{
410 parse_args("early options", cmdline, NULL, 0, do_early_param);
411}
412
413/* Arch code calls this early on, or if not, just before other parsing. */
414void __init parse_early_param(void)
415{
416 static __initdata int done = 0;
417 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
418
419 if (done)
420 return;
421
422 /* All fall through to do_early_param. */
423 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
424 parse_early_options(tmp_cmdline);
425 done = 1;
426}
427
428/*
429 * Activate the first processor.
430 */
431
432static void __init boot_cpu_init(void)
433{
434 int cpu = smp_processor_id();
435 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
436 set_cpu_online(cpu, true);
437 set_cpu_active(cpu, true);
438 set_cpu_present(cpu, true);
439 set_cpu_possible(cpu, true);
440}
441
442void __init __weak smp_setup_processor_id(void)
443{
444}
445
446void __init __weak thread_info_cache_init(void)
447{
448}
449
450/*
451 * Set up kernel memory allocators
452 */
453static void __init mm_init(void)
454{
455 /*
456 * page_cgroup requires countinous pages as memmap
457 * and it's bigger than MAX_ORDER unless SPARSEMEM.
458 */
459 page_cgroup_init_flatmem();
460 mem_init();
461 kmem_cache_init();
462 percpu_init_late();
463 pgtable_cache_init();
464 vmalloc_init();
465}
466
467asmlinkage void __init start_kernel(void)
468{
469 char * command_line;
470 extern const struct kernel_param __start___param[], __stop___param[];
471
472 smp_setup_processor_id();
473
474 /*
475 * Need to run as early as possible, to initialize the
476 * lockdep hash:
477 */
478 lockdep_init();
479 debug_objects_early_init();
480
481 /*
482 * Set up the the initial canary ASAP:
483 */
484 boot_init_stack_canary();
485
486 cgroup_init_early();
487
488 local_irq_disable();
489 early_boot_irqs_disabled = true;
490
491/*
492 * Interrupts are still disabled. Do necessary setups, then
493 * enable them
494 */
495 tick_init();
496 boot_cpu_init();
497 page_address_init();
498 printk(KERN_NOTICE "%s", linux_banner);
499 setup_arch(&command_line);
500 mm_init_owner(&init_mm, &init_task);
501 mm_init_cpumask(&init_mm);
502 setup_command_line(command_line);
503 setup_nr_cpu_ids();
504 setup_per_cpu_areas();
505 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
506
507 build_all_zonelists(NULL);
508 page_alloc_init();
509
510 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
511 parse_early_param();
512 parse_args("Booting kernel", static_command_line, __start___param,
513 __stop___param - __start___param,
514 &unknown_bootoption);
515 /*
516 * These use large bootmem allocations and must precede
517 * kmem_cache_init()
518 */
519 setup_log_buf(0);
520 pidhash_init();
521 vfs_caches_init_early();
522 sort_main_extable();
523 trap_init();
524 mm_init();
525
526 /*
527 * Set up the scheduler prior starting any interrupts (such as the
528 * timer interrupt). Full topology setup happens at smp_init()
529 * time - but meanwhile we still have a functioning scheduler.
530 */
531 sched_init();
532 /*
533 * Disable preemption - early bootup scheduling is extremely
534 * fragile until we cpu_idle() for the first time.
535 */
536 preempt_disable();
537 if (!irqs_disabled()) {
538 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
539 "enabled *very* early, fixing it\n");
540 local_irq_disable();
541 }
542 idr_init_cache();
543 perf_event_init();
544 rcu_init();
545 radix_tree_init();
546 /* init some links before init_ISA_irqs() */
547 early_irq_init();
548 init_IRQ();
549 prio_tree_init();
550 init_timers();
551 hrtimers_init();
552 softirq_init();
553 timekeeping_init();
554 time_init();
555 profile_init();
556 call_function_init();
557 if (!irqs_disabled())
558 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
559 "enabled early\n");
560 early_boot_irqs_disabled = false;
561 local_irq_enable();
562
563 /* Interrupts are enabled now so all GFP allocations are safe. */
564 gfp_allowed_mask = __GFP_BITS_MASK;
565
566 kmem_cache_init_late();
567
568 /*
569 * HACK ALERT! This is early. We're enabling the console before
570 * we've done PCI setups etc, and console_init() must be aware of
571 * this. But we do want output early, in case something goes wrong.
572 */
573 console_init();
574 if (panic_later)
575 panic(panic_later, panic_param);
576
577 lockdep_info();
578
579 /*
580 * Need to run this when irqs are enabled, because it wants
581 * to self-test [hard/soft]-irqs on/off lock inversion bugs
582 * too:
583 */
584 locking_selftest();
585
586#ifdef CONFIG_BLK_DEV_INITRD
587 if (initrd_start && !initrd_below_start_ok &&
588 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
589 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
590 "disabling it.\n",
591 page_to_pfn(virt_to_page((void *)initrd_start)),
592 min_low_pfn);
593 initrd_start = 0;
594 }
595#endif
596 page_cgroup_init();
597 enable_debug_pagealloc();
598 debug_objects_mem_init();
599 kmemleak_init();
600 setup_per_cpu_pageset();
601 numa_policy_init();
602 if (late_time_init)
603 late_time_init();
604 sched_clock_init();
605 calibrate_delay();
606 pidmap_init();
607 anon_vma_init();
608#ifdef CONFIG_X86
609 if (efi_enabled)
610 efi_enter_virtual_mode();
611#endif
612 thread_info_cache_init();
613 cred_init();
614 fork_init(totalram_pages);
615 proc_caches_init();
616 buffer_init();
617 key_init();
618 security_init();
619 dbg_late_init();
620 vfs_caches_init(totalram_pages);
621 signals_init();
622 /* rootfs populating might need page-writeback */
623 page_writeback_init();
624#ifdef CONFIG_PROC_FS
625 proc_root_init();
626#endif
627 cgroup_init();
628 cpuset_init();
629 taskstats_init_early();
630 delayacct_init();
631
632 check_bugs();
633
634 acpi_early_init(); /* before LAPIC and SMP init */
635 sfi_init_late();
636
637 ftrace_init();
638
639 /* Do the rest non-__init'ed, we're now alive */
640 rest_init();
641}
642
643/* Call all constructor functions linked into the kernel. */
644static void __init do_ctors(void)
645{
646#ifdef CONFIG_CONSTRUCTORS
647 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
648
649 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
650 (*fn)();
651#endif
652}
653
654int initcall_debug;
655core_param(initcall_debug, initcall_debug, bool, 0644);
656
657static char msgbuf[64];
658
659static int __init_or_module do_one_initcall_debug(initcall_t fn)
660{
661 ktime_t calltime, delta, rettime;
662 unsigned long long duration;
663 int ret;
664
665 printk(KERN_DEBUG "calling %pF @ %i\n", fn, task_pid_nr(current));
666 calltime = ktime_get();
667 ret = fn();
668 rettime = ktime_get();
669 delta = ktime_sub(rettime, calltime);
670 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
671 printk(KERN_DEBUG "initcall %pF returned %d after %lld usecs\n", fn,
672 ret, duration);
673
674 return ret;
675}
676
677int __init_or_module do_one_initcall(initcall_t fn)
678{
679 int count = preempt_count();
680 int ret;
681
682 if (initcall_debug)
683 ret = do_one_initcall_debug(fn);
684 else
685 ret = fn();
686
687 msgbuf[0] = 0;
688
689 if (ret && ret != -ENODEV && initcall_debug)
690 sprintf(msgbuf, "error code %d ", ret);
691
692 if (preempt_count() != count) {
693 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
694 preempt_count() = count;
695 }
696 if (irqs_disabled()) {
697 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
698 local_irq_enable();
699 }
700 if (msgbuf[0]) {
701 printk("initcall %pF returned with %s\n", fn, msgbuf);
702 }
703
704 return ret;
705}
706
707
708extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
709
710static void __init do_initcalls(void)
711{
712 initcall_t *fn;
713
714 for (fn = __early_initcall_end; fn < __initcall_end; fn++)
715 do_one_initcall(*fn);
716}
717
718/*
719 * Ok, the machine is now initialized. None of the devices
720 * have been touched yet, but the CPU subsystem is up and
721 * running, and memory and process management works.
722 *
723 * Now we can finally start doing some real work..
724 */
725static void __init do_basic_setup(void)
726{
727 cpuset_init_smp();
728 usermodehelper_init();
729 shmem_init();
730 driver_init();
731 init_irq_proc();
732 do_ctors();
733 usermodehelper_enable();
734 do_initcalls();
735}
736
737static void __init do_pre_smp_initcalls(void)
738{
739 initcall_t *fn;
740
741 for (fn = __initcall_start; fn < __early_initcall_end; fn++)
742 do_one_initcall(*fn);
743}
744
745static void run_init_process(const char *init_filename)
746{
747 argv_init[0] = init_filename;
748 kernel_execve(init_filename, argv_init, envp_init);
749}
750
751/* This is a non __init function. Force it to be noinline otherwise gcc
752 * makes it inline to init() and it becomes part of init.text section
753 */
754static noinline int init_post(void)
755{
756 /* need to finish all async __init code before freeing the memory */
757 async_synchronize_full();
758 free_initmem();
759 mark_rodata_ro();
760 system_state = SYSTEM_RUNNING;
761 numa_default_policy();
762
763
764 current->signal->flags |= SIGNAL_UNKILLABLE;
765
766 if (ramdisk_execute_command) {
767 run_init_process(ramdisk_execute_command);
768 printk(KERN_WARNING "Failed to execute %s\n",
769 ramdisk_execute_command);
770 }
771
772 /*
773 * We try each of these until one succeeds.
774 *
775 * The Bourne shell can be used instead of init if we are
776 * trying to recover a really broken machine.
777 */
778 if (execute_command) {
779 run_init_process(execute_command);
780 printk(KERN_WARNING "Failed to execute %s. Attempting "
781 "defaults...\n", execute_command);
782 }
783 run_init_process("/sbin/init");
784 run_init_process("/etc/init");
785 run_init_process("/bin/init");
786 run_init_process("/bin/sh");
787
788 panic("No init found. Try passing init= option to kernel. "
789 "See Linux Documentation/init.txt for guidance.");
790}
791
792static int __init kernel_init(void * unused)
793{
794 /*
795 * Wait until kthreadd is all set-up.
796 */
797 wait_for_completion(&kthreadd_done);
798 /*
799 * init can allocate pages on any node
800 */
801 set_mems_allowed(node_states[N_HIGH_MEMORY]);
802 /*
803 * init can run on any cpu.
804 */
805 set_cpus_allowed_ptr(current, cpu_all_mask);
806
807 cad_pid = task_pid(current);
808
809 smp_prepare_cpus(setup_max_cpus);
810
811 do_pre_smp_initcalls();
812 lockup_detector_init();
813
814 smp_init();
815 sched_init_smp();
816
817 do_basic_setup();
818
819 /* Open the /dev/console on the rootfs, this should never fail */
820 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
821 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
822
823 (void) sys_dup(0);
824 (void) sys_dup(0);
825 /*
826 * check if there is an early userspace init. If yes, let it do all
827 * the work
828 */
829
830 if (!ramdisk_execute_command)
831 ramdisk_execute_command = "/init";
832
833 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
834 ramdisk_execute_command = NULL;
835 prepare_namespace();
836 }
837
838 /*
839 * Ok, we have completed the initial bootup, and
840 * we're essentially up and running. Get rid of the
841 * initmem segments and start the user-mode stuff..
842 */
843
844 init_post();
845 return 0;
846}
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * linux/init/main.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * GK 2/5/95 - Changed to support mounting root fs via NFS
8 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
9 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
10 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
11 */
12
13#define DEBUG /* Enable initcall_debug */
14
15#include <linux/types.h>
16#include <linux/extable.h>
17#include <linux/module.h>
18#include <linux/proc_fs.h>
19#include <linux/binfmts.h>
20#include <linux/kernel.h>
21#include <linux/syscalls.h>
22#include <linux/stackprotector.h>
23#include <linux/string.h>
24#include <linux/ctype.h>
25#include <linux/delay.h>
26#include <linux/ioport.h>
27#include <linux/init.h>
28#include <linux/initrd.h>
29#include <linux/memblock.h>
30#include <linux/acpi.h>
31#include <linux/bootconfig.h>
32#include <linux/console.h>
33#include <linux/nmi.h>
34#include <linux/percpu.h>
35#include <linux/kmod.h>
36#include <linux/kprobes.h>
37#include <linux/vmalloc.h>
38#include <linux/kernel_stat.h>
39#include <linux/start_kernel.h>
40#include <linux/security.h>
41#include <linux/smp.h>
42#include <linux/profile.h>
43#include <linux/rcupdate.h>
44#include <linux/moduleparam.h>
45#include <linux/kallsyms.h>
46#include <linux/writeback.h>
47#include <linux/cpu.h>
48#include <linux/cpuset.h>
49#include <linux/cgroup.h>
50#include <linux/efi.h>
51#include <linux/tick.h>
52#include <linux/sched/isolation.h>
53#include <linux/interrupt.h>
54#include <linux/taskstats_kern.h>
55#include <linux/delayacct.h>
56#include <linux/unistd.h>
57#include <linux/utsname.h>
58#include <linux/rmap.h>
59#include <linux/mempolicy.h>
60#include <linux/key.h>
61#include <linux/buffer_head.h>
62#include <linux/page_ext.h>
63#include <linux/debug_locks.h>
64#include <linux/debugobjects.h>
65#include <linux/lockdep.h>
66#include <linux/kmemleak.h>
67#include <linux/padata.h>
68#include <linux/pid_namespace.h>
69#include <linux/device/driver.h>
70#include <linux/kthread.h>
71#include <linux/sched.h>
72#include <linux/sched/init.h>
73#include <linux/signal.h>
74#include <linux/idr.h>
75#include <linux/kgdb.h>
76#include <linux/ftrace.h>
77#include <linux/async.h>
78#include <linux/sfi.h>
79#include <linux/shmem_fs.h>
80#include <linux/slab.h>
81#include <linux/perf_event.h>
82#include <linux/ptrace.h>
83#include <linux/pti.h>
84#include <linux/blkdev.h>
85#include <linux/elevator.h>
86#include <linux/sched/clock.h>
87#include <linux/sched/task.h>
88#include <linux/sched/task_stack.h>
89#include <linux/context_tracking.h>
90#include <linux/random.h>
91#include <linux/list.h>
92#include <linux/integrity.h>
93#include <linux/proc_ns.h>
94#include <linux/io.h>
95#include <linux/cache.h>
96#include <linux/rodata_test.h>
97#include <linux/jump_label.h>
98#include <linux/mem_encrypt.h>
99#include <linux/kcsan.h>
100#include <linux/init_syscalls.h>
101
102#include <asm/io.h>
103#include <asm/bugs.h>
104#include <asm/setup.h>
105#include <asm/sections.h>
106#include <asm/cacheflush.h>
107
108#define CREATE_TRACE_POINTS
109#include <trace/events/initcall.h>
110
111static int kernel_init(void *);
112
113extern void init_IRQ(void);
114extern void radix_tree_init(void);
115
116/*
117 * Debug helper: via this flag we know that we are in 'early bootup code'
118 * where only the boot processor is running with IRQ disabled. This means
119 * two things - IRQ must not be enabled before the flag is cleared and some
120 * operations which are not allowed with IRQ disabled are allowed while the
121 * flag is set.
122 */
123bool early_boot_irqs_disabled __read_mostly;
124
125enum system_states system_state __read_mostly;
126EXPORT_SYMBOL(system_state);
127
128/*
129 * Boot command-line arguments
130 */
131#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
132#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
133
134extern void time_init(void);
135/* Default late time init is NULL. archs can override this later. */
136void (*__initdata late_time_init)(void);
137
138/* Untouched command line saved by arch-specific code. */
139char __initdata boot_command_line[COMMAND_LINE_SIZE];
140/* Untouched saved command line (eg. for /proc) */
141char *saved_command_line;
142/* Command line for parameter parsing */
143static char *static_command_line;
144/* Untouched extra command line */
145static char *extra_command_line;
146/* Extra init arguments */
147static char *extra_init_args;
148
149#ifdef CONFIG_BOOT_CONFIG
150/* Is bootconfig on command line? */
151static bool bootconfig_found;
152static bool initargs_found;
153#else
154# define bootconfig_found false
155# define initargs_found false
156#endif
157
158static char *execute_command;
159static char *ramdisk_execute_command = "/init";
160
161/*
162 * Used to generate warnings if static_key manipulation functions are used
163 * before jump_label_init is called.
164 */
165bool static_key_initialized __read_mostly;
166EXPORT_SYMBOL_GPL(static_key_initialized);
167
168/*
169 * If set, this is an indication to the drivers that reset the underlying
170 * device before going ahead with the initialization otherwise driver might
171 * rely on the BIOS and skip the reset operation.
172 *
173 * This is useful if kernel is booting in an unreliable environment.
174 * For ex. kdump situation where previous kernel has crashed, BIOS has been
175 * skipped and devices will be in unknown state.
176 */
177unsigned int reset_devices;
178EXPORT_SYMBOL(reset_devices);
179
180static int __init set_reset_devices(char *str)
181{
182 reset_devices = 1;
183 return 1;
184}
185
186__setup("reset_devices", set_reset_devices);
187
188static const char *argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
189const char *envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
190static const char *panic_later, *panic_param;
191
192extern const struct obs_kernel_param __setup_start[], __setup_end[];
193
194static bool __init obsolete_checksetup(char *line)
195{
196 const struct obs_kernel_param *p;
197 bool had_early_param = false;
198
199 p = __setup_start;
200 do {
201 int n = strlen(p->str);
202 if (parameqn(line, p->str, n)) {
203 if (p->early) {
204 /* Already done in parse_early_param?
205 * (Needs exact match on param part).
206 * Keep iterating, as we can have early
207 * params and __setups of same names 8( */
208 if (line[n] == '\0' || line[n] == '=')
209 had_early_param = true;
210 } else if (!p->setup_func) {
211 pr_warn("Parameter %s is obsolete, ignored\n",
212 p->str);
213 return true;
214 } else if (p->setup_func(line + n))
215 return true;
216 }
217 p++;
218 } while (p < __setup_end);
219
220 return had_early_param;
221}
222
223/*
224 * This should be approx 2 Bo*oMips to start (note initial shift), and will
225 * still work even if initially too large, it will just take slightly longer
226 */
227unsigned long loops_per_jiffy = (1<<12);
228EXPORT_SYMBOL(loops_per_jiffy);
229
230static int __init debug_kernel(char *str)
231{
232 console_loglevel = CONSOLE_LOGLEVEL_DEBUG;
233 return 0;
234}
235
236static int __init quiet_kernel(char *str)
237{
238 console_loglevel = CONSOLE_LOGLEVEL_QUIET;
239 return 0;
240}
241
242early_param("debug", debug_kernel);
243early_param("quiet", quiet_kernel);
244
245static int __init loglevel(char *str)
246{
247 int newlevel;
248
249 /*
250 * Only update loglevel value when a correct setting was passed,
251 * to prevent blind crashes (when loglevel being set to 0) that
252 * are quite hard to debug
253 */
254 if (get_option(&str, &newlevel)) {
255 console_loglevel = newlevel;
256 return 0;
257 }
258
259 return -EINVAL;
260}
261
262early_param("loglevel", loglevel);
263
264#ifdef CONFIG_BLK_DEV_INITRD
265static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
266{
267 u32 size, csum;
268 char *data;
269 u32 *hdr;
270
271 if (!initrd_end)
272 return NULL;
273
274 data = (char *)initrd_end - BOOTCONFIG_MAGIC_LEN;
275 if (memcmp(data, BOOTCONFIG_MAGIC, BOOTCONFIG_MAGIC_LEN))
276 return NULL;
277
278 hdr = (u32 *)(data - 8);
279 size = hdr[0];
280 csum = hdr[1];
281
282 data = ((void *)hdr) - size;
283 if ((unsigned long)data < initrd_start) {
284 pr_err("bootconfig size %d is greater than initrd size %ld\n",
285 size, initrd_end - initrd_start);
286 return NULL;
287 }
288
289 /* Remove bootconfig from initramfs/initrd */
290 initrd_end = (unsigned long)data;
291 if (_size)
292 *_size = size;
293 if (_csum)
294 *_csum = csum;
295
296 return data;
297}
298#else
299static void * __init get_boot_config_from_initrd(u32 *_size, u32 *_csum)
300{
301 return NULL;
302}
303#endif
304
305#ifdef CONFIG_BOOT_CONFIG
306
307static char xbc_namebuf[XBC_KEYLEN_MAX] __initdata;
308
309#define rest(dst, end) ((end) > (dst) ? (end) - (dst) : 0)
310
311static int __init xbc_snprint_cmdline(char *buf, size_t size,
312 struct xbc_node *root)
313{
314 struct xbc_node *knode, *vnode;
315 char *end = buf + size;
316 const char *val;
317 int ret;
318
319 xbc_node_for_each_key_value(root, knode, val) {
320 ret = xbc_node_compose_key_after(root, knode,
321 xbc_namebuf, XBC_KEYLEN_MAX);
322 if (ret < 0)
323 return ret;
324
325 vnode = xbc_node_get_child(knode);
326 if (!vnode) {
327 ret = snprintf(buf, rest(buf, end), "%s ", xbc_namebuf);
328 if (ret < 0)
329 return ret;
330 buf += ret;
331 continue;
332 }
333 xbc_array_for_each_value(vnode, val) {
334 ret = snprintf(buf, rest(buf, end), "%s=\"%s\" ",
335 xbc_namebuf, val);
336 if (ret < 0)
337 return ret;
338 buf += ret;
339 }
340 }
341
342 return buf - (end - size);
343}
344#undef rest
345
346/* Make an extra command line under given key word */
347static char * __init xbc_make_cmdline(const char *key)
348{
349 struct xbc_node *root;
350 char *new_cmdline;
351 int ret, len = 0;
352
353 root = xbc_find_node(key);
354 if (!root)
355 return NULL;
356
357 /* Count required buffer size */
358 len = xbc_snprint_cmdline(NULL, 0, root);
359 if (len <= 0)
360 return NULL;
361
362 new_cmdline = memblock_alloc(len + 1, SMP_CACHE_BYTES);
363 if (!new_cmdline) {
364 pr_err("Failed to allocate memory for extra kernel cmdline.\n");
365 return NULL;
366 }
367
368 ret = xbc_snprint_cmdline(new_cmdline, len + 1, root);
369 if (ret < 0 || ret > len) {
370 pr_err("Failed to print extra kernel cmdline.\n");
371 return NULL;
372 }
373
374 return new_cmdline;
375}
376
377static u32 boot_config_checksum(unsigned char *p, u32 size)
378{
379 u32 ret = 0;
380
381 while (size--)
382 ret += *p++;
383
384 return ret;
385}
386
387static int __init bootconfig_params(char *param, char *val,
388 const char *unused, void *arg)
389{
390 if (strcmp(param, "bootconfig") == 0) {
391 bootconfig_found = true;
392 }
393 return 0;
394}
395
396static void __init setup_boot_config(const char *cmdline)
397{
398 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
399 const char *msg;
400 int pos;
401 u32 size, csum;
402 char *data, *copy, *err;
403 int ret;
404
405 /* Cut out the bootconfig data even if we have no bootconfig option */
406 data = get_boot_config_from_initrd(&size, &csum);
407
408 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
409 err = parse_args("bootconfig", tmp_cmdline, NULL, 0, 0, 0, NULL,
410 bootconfig_params);
411
412 if (IS_ERR(err) || !bootconfig_found)
413 return;
414
415 /* parse_args() stops at '--' and returns an address */
416 if (err)
417 initargs_found = true;
418
419 if (!data) {
420 pr_err("'bootconfig' found on command line, but no bootconfig found\n");
421 return;
422 }
423
424 if (size >= XBC_DATA_MAX) {
425 pr_err("bootconfig size %d greater than max size %d\n",
426 size, XBC_DATA_MAX);
427 return;
428 }
429
430 if (boot_config_checksum((unsigned char *)data, size) != csum) {
431 pr_err("bootconfig checksum failed\n");
432 return;
433 }
434
435 copy = memblock_alloc(size + 1, SMP_CACHE_BYTES);
436 if (!copy) {
437 pr_err("Failed to allocate memory for bootconfig\n");
438 return;
439 }
440
441 memcpy(copy, data, size);
442 copy[size] = '\0';
443
444 ret = xbc_init(copy, &msg, &pos);
445 if (ret < 0) {
446 if (pos < 0)
447 pr_err("Failed to init bootconfig: %s.\n", msg);
448 else
449 pr_err("Failed to parse bootconfig: %s at %d.\n",
450 msg, pos);
451 } else {
452 pr_info("Load bootconfig: %d bytes %d nodes\n", size, ret);
453 /* keys starting with "kernel." are passed via cmdline */
454 extra_command_line = xbc_make_cmdline("kernel");
455 /* Also, "init." keys are init arguments */
456 extra_init_args = xbc_make_cmdline("init");
457 }
458 return;
459}
460
461#else
462
463static void __init setup_boot_config(const char *cmdline)
464{
465 /* Remove bootconfig data from initrd */
466 get_boot_config_from_initrd(NULL, NULL);
467}
468
469static int __init warn_bootconfig(char *str)
470{
471 pr_warn("WARNING: 'bootconfig' found on the kernel command line but CONFIG_BOOTCONFIG is not set.\n");
472 return 0;
473}
474early_param("bootconfig", warn_bootconfig);
475
476#endif
477
478/* Change NUL term back to "=", to make "param" the whole string. */
479static void __init repair_env_string(char *param, char *val)
480{
481 if (val) {
482 /* param=val or param="val"? */
483 if (val == param+strlen(param)+1)
484 val[-1] = '=';
485 else if (val == param+strlen(param)+2) {
486 val[-2] = '=';
487 memmove(val-1, val, strlen(val)+1);
488 } else
489 BUG();
490 }
491}
492
493/* Anything after -- gets handed straight to init. */
494static int __init set_init_arg(char *param, char *val,
495 const char *unused, void *arg)
496{
497 unsigned int i;
498
499 if (panic_later)
500 return 0;
501
502 repair_env_string(param, val);
503
504 for (i = 0; argv_init[i]; i++) {
505 if (i == MAX_INIT_ARGS) {
506 panic_later = "init";
507 panic_param = param;
508 return 0;
509 }
510 }
511 argv_init[i] = param;
512 return 0;
513}
514
515/*
516 * Unknown boot options get handed to init, unless they look like
517 * unused parameters (modprobe will find them in /proc/cmdline).
518 */
519static int __init unknown_bootoption(char *param, char *val,
520 const char *unused, void *arg)
521{
522 size_t len = strlen(param);
523
524 repair_env_string(param, val);
525
526 /* Handle obsolete-style parameters */
527 if (obsolete_checksetup(param))
528 return 0;
529
530 /* Unused module parameter. */
531 if (strnchr(param, len, '.'))
532 return 0;
533
534 if (panic_later)
535 return 0;
536
537 if (val) {
538 /* Environment option */
539 unsigned int i;
540 for (i = 0; envp_init[i]; i++) {
541 if (i == MAX_INIT_ENVS) {
542 panic_later = "env";
543 panic_param = param;
544 }
545 if (!strncmp(param, envp_init[i], len+1))
546 break;
547 }
548 envp_init[i] = param;
549 } else {
550 /* Command line option */
551 unsigned int i;
552 for (i = 0; argv_init[i]; i++) {
553 if (i == MAX_INIT_ARGS) {
554 panic_later = "init";
555 panic_param = param;
556 }
557 }
558 argv_init[i] = param;
559 }
560 return 0;
561}
562
563static int __init init_setup(char *str)
564{
565 unsigned int i;
566
567 execute_command = str;
568 /*
569 * In case LILO is going to boot us with default command line,
570 * it prepends "auto" before the whole cmdline which makes
571 * the shell think it should execute a script with such name.
572 * So we ignore all arguments entered _before_ init=... [MJ]
573 */
574 for (i = 1; i < MAX_INIT_ARGS; i++)
575 argv_init[i] = NULL;
576 return 1;
577}
578__setup("init=", init_setup);
579
580static int __init rdinit_setup(char *str)
581{
582 unsigned int i;
583
584 ramdisk_execute_command = str;
585 /* See "auto" comment in init_setup */
586 for (i = 1; i < MAX_INIT_ARGS; i++)
587 argv_init[i] = NULL;
588 return 1;
589}
590__setup("rdinit=", rdinit_setup);
591
592#ifndef CONFIG_SMP
593static const unsigned int setup_max_cpus = NR_CPUS;
594static inline void setup_nr_cpu_ids(void) { }
595static inline void smp_prepare_cpus(unsigned int maxcpus) { }
596#endif
597
598/*
599 * We need to store the untouched command line for future reference.
600 * We also need to store the touched command line since the parameter
601 * parsing is performed in place, and we should allow a component to
602 * store reference of name/value for future reference.
603 */
604static void __init setup_command_line(char *command_line)
605{
606 size_t len, xlen = 0, ilen = 0;
607
608 if (extra_command_line)
609 xlen = strlen(extra_command_line);
610 if (extra_init_args)
611 ilen = strlen(extra_init_args) + 4; /* for " -- " */
612
613 len = xlen + strlen(boot_command_line) + 1;
614
615 saved_command_line = memblock_alloc(len + ilen, SMP_CACHE_BYTES);
616 if (!saved_command_line)
617 panic("%s: Failed to allocate %zu bytes\n", __func__, len + ilen);
618
619 static_command_line = memblock_alloc(len, SMP_CACHE_BYTES);
620 if (!static_command_line)
621 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
622
623 if (xlen) {
624 /*
625 * We have to put extra_command_line before boot command
626 * lines because there could be dashes (separator of init
627 * command line) in the command lines.
628 */
629 strcpy(saved_command_line, extra_command_line);
630 strcpy(static_command_line, extra_command_line);
631 }
632 strcpy(saved_command_line + xlen, boot_command_line);
633 strcpy(static_command_line + xlen, command_line);
634
635 if (ilen) {
636 /*
637 * Append supplemental init boot args to saved_command_line
638 * so that user can check what command line options passed
639 * to init.
640 */
641 len = strlen(saved_command_line);
642 if (initargs_found) {
643 saved_command_line[len++] = ' ';
644 } else {
645 strcpy(saved_command_line + len, " -- ");
646 len += 4;
647 }
648
649 strcpy(saved_command_line + len, extra_init_args);
650 }
651}
652
653/*
654 * We need to finalize in a non-__init function or else race conditions
655 * between the root thread and the init thread may cause start_kernel to
656 * be reaped by free_initmem before the root thread has proceeded to
657 * cpu_idle.
658 *
659 * gcc-3.4 accidentally inlines this function, so use noinline.
660 */
661
662static __initdata DECLARE_COMPLETION(kthreadd_done);
663
664noinline void __ref rest_init(void)
665{
666 struct task_struct *tsk;
667 int pid;
668
669 rcu_scheduler_starting();
670 /*
671 * We need to spawn init first so that it obtains pid 1, however
672 * the init task will end up wanting to create kthreads, which, if
673 * we schedule it before we create kthreadd, will OOPS.
674 */
675 pid = kernel_thread(kernel_init, NULL, CLONE_FS);
676 /*
677 * Pin init on the boot CPU. Task migration is not properly working
678 * until sched_init_smp() has been run. It will set the allowed
679 * CPUs for init to the non isolated CPUs.
680 */
681 rcu_read_lock();
682 tsk = find_task_by_pid_ns(pid, &init_pid_ns);
683 set_cpus_allowed_ptr(tsk, cpumask_of(smp_processor_id()));
684 rcu_read_unlock();
685
686 numa_default_policy();
687 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
688 rcu_read_lock();
689 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
690 rcu_read_unlock();
691
692 /*
693 * Enable might_sleep() and smp_processor_id() checks.
694 * They cannot be enabled earlier because with CONFIG_PREEMPTION=y
695 * kernel_thread() would trigger might_sleep() splats. With
696 * CONFIG_PREEMPT_VOLUNTARY=y the init task might have scheduled
697 * already, but it's stuck on the kthreadd_done completion.
698 */
699 system_state = SYSTEM_SCHEDULING;
700
701 complete(&kthreadd_done);
702
703 /*
704 * The boot idle thread must execute schedule()
705 * at least once to get things moving:
706 */
707 schedule_preempt_disabled();
708 /* Call into cpu_idle with preempt disabled */
709 cpu_startup_entry(CPUHP_ONLINE);
710}
711
712/* Check for early params. */
713static int __init do_early_param(char *param, char *val,
714 const char *unused, void *arg)
715{
716 const struct obs_kernel_param *p;
717
718 for (p = __setup_start; p < __setup_end; p++) {
719 if ((p->early && parameq(param, p->str)) ||
720 (strcmp(param, "console") == 0 &&
721 strcmp(p->str, "earlycon") == 0)
722 ) {
723 if (p->setup_func(val) != 0)
724 pr_warn("Malformed early option '%s'\n", param);
725 }
726 }
727 /* We accept everything at this stage. */
728 return 0;
729}
730
731void __init parse_early_options(char *cmdline)
732{
733 parse_args("early options", cmdline, NULL, 0, 0, 0, NULL,
734 do_early_param);
735}
736
737/* Arch code calls this early on, or if not, just before other parsing. */
738void __init parse_early_param(void)
739{
740 static int done __initdata;
741 static char tmp_cmdline[COMMAND_LINE_SIZE] __initdata;
742
743 if (done)
744 return;
745
746 /* All fall through to do_early_param. */
747 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
748 parse_early_options(tmp_cmdline);
749 done = 1;
750}
751
752void __init __weak arch_post_acpi_subsys_init(void) { }
753
754void __init __weak smp_setup_processor_id(void)
755{
756}
757
758# if THREAD_SIZE >= PAGE_SIZE
759void __init __weak thread_stack_cache_init(void)
760{
761}
762#endif
763
764void __init __weak mem_encrypt_init(void) { }
765
766void __init __weak poking_init(void) { }
767
768void __init __weak pgtable_cache_init(void) { }
769
770bool initcall_debug;
771core_param(initcall_debug, initcall_debug, bool, 0644);
772
773#ifdef TRACEPOINTS_ENABLED
774static void __init initcall_debug_enable(void);
775#else
776static inline void initcall_debug_enable(void)
777{
778}
779#endif
780
781/* Report memory auto-initialization states for this boot. */
782static void __init report_meminit(void)
783{
784 const char *stack;
785
786 if (IS_ENABLED(CONFIG_INIT_STACK_ALL_PATTERN))
787 stack = "all(pattern)";
788 else if (IS_ENABLED(CONFIG_INIT_STACK_ALL_ZERO))
789 stack = "all(zero)";
790 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF_ALL))
791 stack = "byref_all(zero)";
792 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_BYREF))
793 stack = "byref(zero)";
794 else if (IS_ENABLED(CONFIG_GCC_PLUGIN_STRUCTLEAK_USER))
795 stack = "__user(zero)";
796 else
797 stack = "off";
798
799 pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n",
800 stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off",
801 want_init_on_free() ? "on" : "off");
802 if (want_init_on_free())
803 pr_info("mem auto-init: clearing system memory may take some time...\n");
804}
805
806/*
807 * Set up kernel memory allocators
808 */
809static void __init mm_init(void)
810{
811 /*
812 * page_ext requires contiguous pages,
813 * bigger than MAX_ORDER unless SPARSEMEM.
814 */
815 page_ext_init_flatmem();
816 init_debug_pagealloc();
817 report_meminit();
818 mem_init();
819 kmem_cache_init();
820 kmemleak_init();
821 pgtable_init();
822 debug_objects_mem_init();
823 vmalloc_init();
824 ioremap_huge_init();
825 /* Should be run before the first non-init thread is created */
826 init_espfix_bsp();
827 /* Should be run after espfix64 is set up. */
828 pti_init();
829}
830
831void __init __weak arch_call_rest_init(void)
832{
833 rest_init();
834}
835
836asmlinkage __visible void __init __no_sanitize_address start_kernel(void)
837{
838 char *command_line;
839 char *after_dashes;
840
841 set_task_stack_end_magic(&init_task);
842 smp_setup_processor_id();
843 debug_objects_early_init();
844
845 cgroup_init_early();
846
847 local_irq_disable();
848 early_boot_irqs_disabled = true;
849
850 /*
851 * Interrupts are still disabled. Do necessary setups, then
852 * enable them.
853 */
854 boot_cpu_init();
855 page_address_init();
856 pr_notice("%s", linux_banner);
857 early_security_init();
858 setup_arch(&command_line);
859 setup_boot_config(command_line);
860 setup_command_line(command_line);
861 setup_nr_cpu_ids();
862 setup_per_cpu_areas();
863 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
864 boot_cpu_hotplug_init();
865
866 build_all_zonelists(NULL);
867 page_alloc_init();
868
869 pr_notice("Kernel command line: %s\n", saved_command_line);
870 /* parameters may set static keys */
871 jump_label_init();
872 parse_early_param();
873 after_dashes = parse_args("Booting kernel",
874 static_command_line, __start___param,
875 __stop___param - __start___param,
876 -1, -1, NULL, &unknown_bootoption);
877 if (!IS_ERR_OR_NULL(after_dashes))
878 parse_args("Setting init args", after_dashes, NULL, 0, -1, -1,
879 NULL, set_init_arg);
880 if (extra_init_args)
881 parse_args("Setting extra init args", extra_init_args,
882 NULL, 0, -1, -1, NULL, set_init_arg);
883
884 /*
885 * These use large bootmem allocations and must precede
886 * kmem_cache_init()
887 */
888 setup_log_buf(0);
889 vfs_caches_init_early();
890 sort_main_extable();
891 trap_init();
892 mm_init();
893
894 ftrace_init();
895
896 /* trace_printk can be enabled here */
897 early_trace_init();
898
899 /*
900 * Set up the scheduler prior starting any interrupts (such as the
901 * timer interrupt). Full topology setup happens at smp_init()
902 * time - but meanwhile we still have a functioning scheduler.
903 */
904 sched_init();
905 /*
906 * Disable preemption - early bootup scheduling is extremely
907 * fragile until we cpu_idle() for the first time.
908 */
909 preempt_disable();
910 if (WARN(!irqs_disabled(),
911 "Interrupts were enabled *very* early, fixing it\n"))
912 local_irq_disable();
913 radix_tree_init();
914
915 /*
916 * Set up housekeeping before setting up workqueues to allow the unbound
917 * workqueue to take non-housekeeping into account.
918 */
919 housekeeping_init();
920
921 /*
922 * Allow workqueue creation and work item queueing/cancelling
923 * early. Work item execution depends on kthreads and starts after
924 * workqueue_init().
925 */
926 workqueue_init_early();
927
928 rcu_init();
929
930 /* Trace events are available after this */
931 trace_init();
932
933 if (initcall_debug)
934 initcall_debug_enable();
935
936 context_tracking_init();
937 /* init some links before init_ISA_irqs() */
938 early_irq_init();
939 init_IRQ();
940 tick_init();
941 rcu_init_nohz();
942 init_timers();
943 hrtimers_init();
944 softirq_init();
945 timekeeping_init();
946
947 /*
948 * For best initial stack canary entropy, prepare it after:
949 * - setup_arch() for any UEFI RNG entropy and boot cmdline access
950 * - timekeeping_init() for ktime entropy used in rand_initialize()
951 * - rand_initialize() to get any arch-specific entropy like RDRAND
952 * - add_latent_entropy() to get any latent entropy
953 * - adding command line entropy
954 */
955 rand_initialize();
956 add_latent_entropy();
957 add_device_randomness(command_line, strlen(command_line));
958 boot_init_stack_canary();
959
960 time_init();
961 perf_event_init();
962 profile_init();
963 call_function_init();
964 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
965
966 early_boot_irqs_disabled = false;
967 local_irq_enable();
968
969 kmem_cache_init_late();
970
971 /*
972 * HACK ALERT! This is early. We're enabling the console before
973 * we've done PCI setups etc, and console_init() must be aware of
974 * this. But we do want output early, in case something goes wrong.
975 */
976 console_init();
977 if (panic_later)
978 panic("Too many boot %s vars at `%s'", panic_later,
979 panic_param);
980
981 lockdep_init();
982
983 /*
984 * Need to run this when irqs are enabled, because it wants
985 * to self-test [hard/soft]-irqs on/off lock inversion bugs
986 * too:
987 */
988 locking_selftest();
989
990 /*
991 * This needs to be called before any devices perform DMA
992 * operations that might use the SWIOTLB bounce buffers. It will
993 * mark the bounce buffers as decrypted so that their usage will
994 * not cause "plain-text" data to be decrypted when accessed.
995 */
996 mem_encrypt_init();
997
998#ifdef CONFIG_BLK_DEV_INITRD
999 if (initrd_start && !initrd_below_start_ok &&
1000 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
1001 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
1002 page_to_pfn(virt_to_page((void *)initrd_start)),
1003 min_low_pfn);
1004 initrd_start = 0;
1005 }
1006#endif
1007 setup_per_cpu_pageset();
1008 numa_policy_init();
1009 acpi_early_init();
1010 if (late_time_init)
1011 late_time_init();
1012 sched_clock_init();
1013 calibrate_delay();
1014 pid_idr_init();
1015 anon_vma_init();
1016#ifdef CONFIG_X86
1017 if (efi_enabled(EFI_RUNTIME_SERVICES))
1018 efi_enter_virtual_mode();
1019#endif
1020 thread_stack_cache_init();
1021 cred_init();
1022 fork_init();
1023 proc_caches_init();
1024 uts_ns_init();
1025 buffer_init();
1026 key_init();
1027 security_init();
1028 dbg_late_init();
1029 vfs_caches_init();
1030 pagecache_init();
1031 signals_init();
1032 seq_file_init();
1033 proc_root_init();
1034 nsfs_init();
1035 cpuset_init();
1036 cgroup_init();
1037 taskstats_init_early();
1038 delayacct_init();
1039
1040 poking_init();
1041 check_bugs();
1042
1043 acpi_subsystem_init();
1044 arch_post_acpi_subsys_init();
1045 sfi_init_late();
1046 kcsan_init();
1047
1048 /* Do the rest non-__init'ed, we're now alive */
1049 arch_call_rest_init();
1050
1051 prevent_tail_call_optimization();
1052}
1053
1054/* Call all constructor functions linked into the kernel. */
1055static void __init do_ctors(void)
1056{
1057#ifdef CONFIG_CONSTRUCTORS
1058 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
1059
1060 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
1061 (*fn)();
1062#endif
1063}
1064
1065#ifdef CONFIG_KALLSYMS
1066struct blacklist_entry {
1067 struct list_head next;
1068 char *buf;
1069};
1070
1071static __initdata_or_module LIST_HEAD(blacklisted_initcalls);
1072
1073static int __init initcall_blacklist(char *str)
1074{
1075 char *str_entry;
1076 struct blacklist_entry *entry;
1077
1078 /* str argument is a comma-separated list of functions */
1079 do {
1080 str_entry = strsep(&str, ",");
1081 if (str_entry) {
1082 pr_debug("blacklisting initcall %s\n", str_entry);
1083 entry = memblock_alloc(sizeof(*entry),
1084 SMP_CACHE_BYTES);
1085 if (!entry)
1086 panic("%s: Failed to allocate %zu bytes\n",
1087 __func__, sizeof(*entry));
1088 entry->buf = memblock_alloc(strlen(str_entry) + 1,
1089 SMP_CACHE_BYTES);
1090 if (!entry->buf)
1091 panic("%s: Failed to allocate %zu bytes\n",
1092 __func__, strlen(str_entry) + 1);
1093 strcpy(entry->buf, str_entry);
1094 list_add(&entry->next, &blacklisted_initcalls);
1095 }
1096 } while (str_entry);
1097
1098 return 0;
1099}
1100
1101static bool __init_or_module initcall_blacklisted(initcall_t fn)
1102{
1103 struct blacklist_entry *entry;
1104 char fn_name[KSYM_SYMBOL_LEN];
1105 unsigned long addr;
1106
1107 if (list_empty(&blacklisted_initcalls))
1108 return false;
1109
1110 addr = (unsigned long) dereference_function_descriptor(fn);
1111 sprint_symbol_no_offset(fn_name, addr);
1112
1113 /*
1114 * fn will be "function_name [module_name]" where [module_name] is not
1115 * displayed for built-in init functions. Strip off the [module_name].
1116 */
1117 strreplace(fn_name, ' ', '\0');
1118
1119 list_for_each_entry(entry, &blacklisted_initcalls, next) {
1120 if (!strcmp(fn_name, entry->buf)) {
1121 pr_debug("initcall %s blacklisted\n", fn_name);
1122 return true;
1123 }
1124 }
1125
1126 return false;
1127}
1128#else
1129static int __init initcall_blacklist(char *str)
1130{
1131 pr_warn("initcall_blacklist requires CONFIG_KALLSYMS\n");
1132 return 0;
1133}
1134
1135static bool __init_or_module initcall_blacklisted(initcall_t fn)
1136{
1137 return false;
1138}
1139#endif
1140__setup("initcall_blacklist=", initcall_blacklist);
1141
1142static __init_or_module void
1143trace_initcall_start_cb(void *data, initcall_t fn)
1144{
1145 ktime_t *calltime = (ktime_t *)data;
1146
1147 printk(KERN_DEBUG "calling %pS @ %i\n", fn, task_pid_nr(current));
1148 *calltime = ktime_get();
1149}
1150
1151static __init_or_module void
1152trace_initcall_finish_cb(void *data, initcall_t fn, int ret)
1153{
1154 ktime_t *calltime = (ktime_t *)data;
1155 ktime_t delta, rettime;
1156 unsigned long long duration;
1157
1158 rettime = ktime_get();
1159 delta = ktime_sub(rettime, *calltime);
1160 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
1161 printk(KERN_DEBUG "initcall %pS returned %d after %lld usecs\n",
1162 fn, ret, duration);
1163}
1164
1165static ktime_t initcall_calltime;
1166
1167#ifdef TRACEPOINTS_ENABLED
1168static void __init initcall_debug_enable(void)
1169{
1170 int ret;
1171
1172 ret = register_trace_initcall_start(trace_initcall_start_cb,
1173 &initcall_calltime);
1174 ret |= register_trace_initcall_finish(trace_initcall_finish_cb,
1175 &initcall_calltime);
1176 WARN(ret, "Failed to register initcall tracepoints\n");
1177}
1178# define do_trace_initcall_start trace_initcall_start
1179# define do_trace_initcall_finish trace_initcall_finish
1180#else
1181static inline void do_trace_initcall_start(initcall_t fn)
1182{
1183 if (!initcall_debug)
1184 return;
1185 trace_initcall_start_cb(&initcall_calltime, fn);
1186}
1187static inline void do_trace_initcall_finish(initcall_t fn, int ret)
1188{
1189 if (!initcall_debug)
1190 return;
1191 trace_initcall_finish_cb(&initcall_calltime, fn, ret);
1192}
1193#endif /* !TRACEPOINTS_ENABLED */
1194
1195int __init_or_module do_one_initcall(initcall_t fn)
1196{
1197 int count = preempt_count();
1198 char msgbuf[64];
1199 int ret;
1200
1201 if (initcall_blacklisted(fn))
1202 return -EPERM;
1203
1204 do_trace_initcall_start(fn);
1205 ret = fn();
1206 do_trace_initcall_finish(fn, ret);
1207
1208 msgbuf[0] = 0;
1209
1210 if (preempt_count() != count) {
1211 sprintf(msgbuf, "preemption imbalance ");
1212 preempt_count_set(count);
1213 }
1214 if (irqs_disabled()) {
1215 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
1216 local_irq_enable();
1217 }
1218 WARN(msgbuf[0], "initcall %pS returned with %s\n", fn, msgbuf);
1219
1220 add_latent_entropy();
1221 return ret;
1222}
1223
1224
1225extern initcall_entry_t __initcall_start[];
1226extern initcall_entry_t __initcall0_start[];
1227extern initcall_entry_t __initcall1_start[];
1228extern initcall_entry_t __initcall2_start[];
1229extern initcall_entry_t __initcall3_start[];
1230extern initcall_entry_t __initcall4_start[];
1231extern initcall_entry_t __initcall5_start[];
1232extern initcall_entry_t __initcall6_start[];
1233extern initcall_entry_t __initcall7_start[];
1234extern initcall_entry_t __initcall_end[];
1235
1236static initcall_entry_t *initcall_levels[] __initdata = {
1237 __initcall0_start,
1238 __initcall1_start,
1239 __initcall2_start,
1240 __initcall3_start,
1241 __initcall4_start,
1242 __initcall5_start,
1243 __initcall6_start,
1244 __initcall7_start,
1245 __initcall_end,
1246};
1247
1248/* Keep these in sync with initcalls in include/linux/init.h */
1249static const char *initcall_level_names[] __initdata = {
1250 "pure",
1251 "core",
1252 "postcore",
1253 "arch",
1254 "subsys",
1255 "fs",
1256 "device",
1257 "late",
1258};
1259
1260static int __init ignore_unknown_bootoption(char *param, char *val,
1261 const char *unused, void *arg)
1262{
1263 return 0;
1264}
1265
1266static void __init do_initcall_level(int level, char *command_line)
1267{
1268 initcall_entry_t *fn;
1269
1270 parse_args(initcall_level_names[level],
1271 command_line, __start___param,
1272 __stop___param - __start___param,
1273 level, level,
1274 NULL, ignore_unknown_bootoption);
1275
1276 trace_initcall_level(initcall_level_names[level]);
1277 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
1278 do_one_initcall(initcall_from_entry(fn));
1279}
1280
1281static void __init do_initcalls(void)
1282{
1283 int level;
1284 size_t len = strlen(saved_command_line) + 1;
1285 char *command_line;
1286
1287 command_line = kzalloc(len, GFP_KERNEL);
1288 if (!command_line)
1289 panic("%s: Failed to allocate %zu bytes\n", __func__, len);
1290
1291 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++) {
1292 /* Parser modifies command_line, restore it each time */
1293 strcpy(command_line, saved_command_line);
1294 do_initcall_level(level, command_line);
1295 }
1296
1297 kfree(command_line);
1298}
1299
1300/*
1301 * Ok, the machine is now initialized. None of the devices
1302 * have been touched yet, but the CPU subsystem is up and
1303 * running, and memory and process management works.
1304 *
1305 * Now we can finally start doing some real work..
1306 */
1307static void __init do_basic_setup(void)
1308{
1309 cpuset_init_smp();
1310 driver_init();
1311 init_irq_proc();
1312 do_ctors();
1313 usermodehelper_enable();
1314 do_initcalls();
1315}
1316
1317static void __init do_pre_smp_initcalls(void)
1318{
1319 initcall_entry_t *fn;
1320
1321 trace_initcall_level("early");
1322 for (fn = __initcall_start; fn < __initcall0_start; fn++)
1323 do_one_initcall(initcall_from_entry(fn));
1324}
1325
1326static int run_init_process(const char *init_filename)
1327{
1328 const char *const *p;
1329
1330 argv_init[0] = init_filename;
1331 pr_info("Run %s as init process\n", init_filename);
1332 pr_debug(" with arguments:\n");
1333 for (p = argv_init; *p; p++)
1334 pr_debug(" %s\n", *p);
1335 pr_debug(" with environment:\n");
1336 for (p = envp_init; *p; p++)
1337 pr_debug(" %s\n", *p);
1338 return kernel_execve(init_filename, argv_init, envp_init);
1339}
1340
1341static int try_to_run_init_process(const char *init_filename)
1342{
1343 int ret;
1344
1345 ret = run_init_process(init_filename);
1346
1347 if (ret && ret != -ENOENT) {
1348 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
1349 init_filename, ret);
1350 }
1351
1352 return ret;
1353}
1354
1355static noinline void __init kernel_init_freeable(void);
1356
1357#if defined(CONFIG_STRICT_KERNEL_RWX) || defined(CONFIG_STRICT_MODULE_RWX)
1358bool rodata_enabled __ro_after_init = true;
1359static int __init set_debug_rodata(char *str)
1360{
1361 return strtobool(str, &rodata_enabled);
1362}
1363__setup("rodata=", set_debug_rodata);
1364#endif
1365
1366#ifdef CONFIG_STRICT_KERNEL_RWX
1367static void mark_readonly(void)
1368{
1369 if (rodata_enabled) {
1370 /*
1371 * load_module() results in W+X mappings, which are cleaned
1372 * up with call_rcu(). Let's make sure that queued work is
1373 * flushed so that we don't hit false positives looking for
1374 * insecure pages which are W+X.
1375 */
1376 rcu_barrier();
1377 mark_rodata_ro();
1378 rodata_test();
1379 } else
1380 pr_info("Kernel memory protection disabled.\n");
1381}
1382#elif defined(CONFIG_ARCH_HAS_STRICT_KERNEL_RWX)
1383static inline void mark_readonly(void)
1384{
1385 pr_warn("Kernel memory protection not selected by kernel config.\n");
1386}
1387#else
1388static inline void mark_readonly(void)
1389{
1390 pr_warn("This architecture does not have kernel memory protection.\n");
1391}
1392#endif
1393
1394void __weak free_initmem(void)
1395{
1396 free_initmem_default(POISON_FREE_INITMEM);
1397}
1398
1399static int __ref kernel_init(void *unused)
1400{
1401 int ret;
1402
1403 kernel_init_freeable();
1404 /* need to finish all async __init code before freeing the memory */
1405 async_synchronize_full();
1406 kprobe_free_init_mem();
1407 ftrace_free_init_mem();
1408 free_initmem();
1409 mark_readonly();
1410
1411 /*
1412 * Kernel mappings are now finalized - update the userspace page-table
1413 * to finalize PTI.
1414 */
1415 pti_finalize();
1416
1417 system_state = SYSTEM_RUNNING;
1418 numa_default_policy();
1419
1420 rcu_end_inkernel_boot();
1421
1422 do_sysctl_args();
1423
1424 if (ramdisk_execute_command) {
1425 ret = run_init_process(ramdisk_execute_command);
1426 if (!ret)
1427 return 0;
1428 pr_err("Failed to execute %s (error %d)\n",
1429 ramdisk_execute_command, ret);
1430 }
1431
1432 /*
1433 * We try each of these until one succeeds.
1434 *
1435 * The Bourne shell can be used instead of init if we are
1436 * trying to recover a really broken machine.
1437 */
1438 if (execute_command) {
1439 ret = run_init_process(execute_command);
1440 if (!ret)
1441 return 0;
1442 panic("Requested init %s failed (error %d).",
1443 execute_command, ret);
1444 }
1445
1446 if (CONFIG_DEFAULT_INIT[0] != '\0') {
1447 ret = run_init_process(CONFIG_DEFAULT_INIT);
1448 if (ret)
1449 pr_err("Default init %s failed (error %d)\n",
1450 CONFIG_DEFAULT_INIT, ret);
1451 else
1452 return 0;
1453 }
1454
1455 if (!try_to_run_init_process("/sbin/init") ||
1456 !try_to_run_init_process("/etc/init") ||
1457 !try_to_run_init_process("/bin/init") ||
1458 !try_to_run_init_process("/bin/sh"))
1459 return 0;
1460
1461 panic("No working init found. Try passing init= option to kernel. "
1462 "See Linux Documentation/admin-guide/init.rst for guidance.");
1463}
1464
1465/* Open /dev/console, for stdin/stdout/stderr, this should never fail */
1466void __init console_on_rootfs(void)
1467{
1468 struct file *file = filp_open("/dev/console", O_RDWR, 0);
1469
1470 if (IS_ERR(file)) {
1471 pr_err("Warning: unable to open an initial console.\n");
1472 return;
1473 }
1474 init_dup(file);
1475 init_dup(file);
1476 init_dup(file);
1477 fput(file);
1478}
1479
1480static noinline void __init kernel_init_freeable(void)
1481{
1482 /*
1483 * Wait until kthreadd is all set-up.
1484 */
1485 wait_for_completion(&kthreadd_done);
1486
1487 /* Now the scheduler is fully set up and can do blocking allocations */
1488 gfp_allowed_mask = __GFP_BITS_MASK;
1489
1490 /*
1491 * init can allocate pages on any node
1492 */
1493 set_mems_allowed(node_states[N_MEMORY]);
1494
1495 cad_pid = task_pid(current);
1496
1497 smp_prepare_cpus(setup_max_cpus);
1498
1499 workqueue_init();
1500
1501 init_mm_internals();
1502
1503 do_pre_smp_initcalls();
1504 lockup_detector_init();
1505
1506 smp_init();
1507 sched_init_smp();
1508
1509 padata_init();
1510 page_alloc_init_late();
1511 /* Initialize page ext after all struct pages are initialized. */
1512 page_ext_init();
1513
1514 do_basic_setup();
1515
1516 console_on_rootfs();
1517
1518 /*
1519 * check if there is an early userspace init. If yes, let it do all
1520 * the work
1521 */
1522 if (init_eaccess(ramdisk_execute_command) != 0) {
1523 ramdisk_execute_command = NULL;
1524 prepare_namespace();
1525 }
1526
1527 /*
1528 * Ok, we have completed the initial bootup, and
1529 * we're essentially up and running. Get rid of the
1530 * initmem segments and start the user-mode stuff..
1531 *
1532 * rootfs is available now, try loading the public keys
1533 * and default modules
1534 */
1535
1536 integrity_load_keys();
1537}