projects/linux·blob v3.13 · refs

942 lines · 22.7 KiB

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 <[email protected]>
10 */
11
12#define DEBUG /* Enable initcall_debug */
13
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/proc_fs.h>
17#include <linux/kernel.h>
18#include <linux/syscalls.h>
19#include <linux/stackprotector.h>
20#include <linux/string.h>
21#include <linux/ctype.h>
22#include <linux/delay.h>
23#include <linux/ioport.h>
24#include <linux/init.h>
25#include <linux/initrd.h>
26#include <linux/bootmem.h>
27#include <linux/acpi.h>
28#include <linux/tty.h>
29#include <linux/percpu.h>
30#include <linux/kmod.h>
31#include <linux/vmalloc.h>
32#include <linux/kernel_stat.h>
33#include <linux/start_kernel.h>
34#include <linux/security.h>
35#include <linux/smp.h>
36#include <linux/profile.h>
37#include <linux/rcupdate.h>
38#include <linux/moduleparam.h>
39#include <linux/kallsyms.h>
40#include <linux/writeback.h>
41#include <linux/cpu.h>
42#include <linux/cpuset.h>
43#include <linux/cgroup.h>
44#include <linux/efi.h>
45#include <linux/tick.h>
46#include <linux/interrupt.h>
47#include <linux/taskstats_kern.h>
48#include <linux/delayacct.h>
49#include <linux/unistd.h>
50#include <linux/rmap.h>
51#include <linux/mempolicy.h>
52#include <linux/key.h>
53#include <linux/buffer_head.h>
54#include <linux/page_cgroup.h>
55#include <linux/debug_locks.h>
56#include <linux/debugobjects.h>
57#include <linux/lockdep.h>
58#include <linux/kmemleak.h>
59#include <linux/pid_namespace.h>
60#include <linux/device.h>
61#include <linux/kthread.h>
62#include <linux/sched.h>
63#include <linux/signal.h>
64#include <linux/idr.h>
65#include <linux/kgdb.h>
66#include <linux/ftrace.h>
67#include <linux/async.h>
68#include <linux/kmemcheck.h>
69#include <linux/sfi.h>
70#include <linux/shmem_fs.h>
71#include <linux/slab.h>
72#include <linux/perf_event.h>
73#include <linux/file.h>
74#include <linux/ptrace.h>
75#include <linux/blkdev.h>
76#include <linux/elevator.h>
77#include <linux/sched_clock.h>
78#include <linux/context_tracking.h>
79#include <linux/random.h>
80
81#include <asm/io.h>
82#include <asm/bugs.h>
83#include <asm/setup.h>
84#include <asm/sections.h>
85#include <asm/cacheflush.h>
86
87#ifdef CONFIG_X86_LOCAL_APIC
88#include <asm/smp.h>
89#endif
90
91static int kernel_init(void *);
92
93extern void init_IRQ(void);
94extern void fork_init(unsigned long);
95extern void mca_init(void);
96extern void sbus_init(void);
97extern void radix_tree_init(void);
98#ifndef CONFIG_DEBUG_RODATA
99static inline void mark_rodata_ro(void) { }
100#endif
101
102#ifdef CONFIG_TC
103extern void tc_init(void);
104#endif
105
106/*
107 * Debug helper: via this flag we know that we are in 'early bootup code'
108 * where only the boot processor is running with IRQ disabled. This means
109 * two things - IRQ must not be enabled before the flag is cleared and some
110 * operations which are not allowed with IRQ disabled are allowed while the
111 * flag is set.
112 */
113bool early_boot_irqs_disabled __read_mostly;
114
115enum system_states system_state __read_mostly;
116EXPORT_SYMBOL(system_state);
117
118/*
119 * Boot command-line arguments
120 */
121#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
122#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
123
124extern void time_init(void);
125/* Default late time init is NULL. archs can override this later. */
126void (*__initdata late_time_init)(void);
127
128/* Untouched command line saved by arch-specific code. */
129char __initdata boot_command_line[COMMAND_LINE_SIZE];
130/* Untouched saved command line (eg. for /proc) */
131char *saved_command_line;
132/* Command line for parameter parsing */
133static char *static_command_line;
134/* Command line for per-initcall parameter parsing */
135static char *initcall_command_line;
136
137static char *execute_command;
138static char *ramdisk_execute_command;
139
140/*
141 * Used to generate warnings if static_key manipulation functions are used
142 * before jump_label_init is called.
143 */
144bool static_key_initialized __read_mostly = false;
145EXPORT_SYMBOL_GPL(static_key_initialized);
146
147/*
148 * If set, this is an indication to the drivers that reset the underlying
149 * device before going ahead with the initialization otherwise driver might
150 * rely on the BIOS and skip the reset operation.
151 *
152 * This is useful if kernel is booting in an unreliable environment.
153 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
154 * skipped and devices will be in unknown state.
155 */
156unsigned int reset_devices;
157EXPORT_SYMBOL(reset_devices);
158
159static int __init set_reset_devices(char *str)
160{
161 reset_devices = 1;
162 return 1;
163}
164
165__setup("reset_devices", set_reset_devices);
166
167static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
168const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
169static const char *panic_later, *panic_param;
170
171extern const struct obs_kernel_param __setup_start[], __setup_end[];
172
173static int __init obsolete_checksetup(char *line)
174{
175 const struct obs_kernel_param *p;
176 int had_early_param = 0;
177
178 p = __setup_start;
179 do {
180 int n = strlen(p->str);
181 if (parameqn(line, p->str, n)) {
182 if (p->early) {
183 /* Already done in parse_early_param?
184 * (Needs exact match on param part).
185 * Keep iterating, as we can have early
186 * params and __setups of same names 8( */
187 if (line[n] == '\0' || line[n] == '=')
188 had_early_param = 1;
189 } else if (!p->setup_func) {
190 pr_warn("Parameter %s is obsolete, ignored\n",
191 p->str);
192 return 1;
193 } else if (p->setup_func(line + n))
194 return 1;
195 }
196 p++;
197 } while (p < __setup_end);
198
199 return had_early_param;
200}
201
202/*
203 * This should be approx 2 Bo*oMips to start (note initial shift), and will
204 * still work even if initially too large, it will just take slightly longer
205 */
206unsigned long loops_per_jiffy = (1<<12);
207
208EXPORT_SYMBOL(loops_per_jiffy);
209
210static int __init debug_kernel(char *str)
211{
212 console_loglevel = 10;
213 return 0;
214}
215
216static int __init quiet_kernel(char *str)
217{
218 console_loglevel = 4;
219 return 0;
220}
221
222early_param("debug", debug_kernel);
223early_param("quiet", quiet_kernel);
224
225static int __init loglevel(char *str)
226{
227 int newlevel;
228
229 /*
230 * Only update loglevel value when a correct setting was passed,
231 * to prevent blind crashes (when loglevel being set to 0) that
232 * are quite hard to debug
233 */
234 if (get_option(&str, &newlevel)) {
235 console_loglevel = newlevel;
236 return 0;
237 }
238
239 return -EINVAL;
240}
241
242early_param("loglevel", loglevel);
243
244/* Change NUL term back to "=", to make "param" the whole string. */
245static int __init repair_env_string(char *param, char *val, const char *unused)
246{
247 if (val) {
248 /* param=val or param="val"? */
249 if (val == param+strlen(param)+1)
250 val[-1] = '=';
251 else if (val == param+strlen(param)+2) {
252 val[-2] = '=';
253 memmove(val-1, val, strlen(val)+1);
254 val--;
255 } else
256 BUG();
257 }
258 return 0;
259}
260
261/*
262 * Unknown boot options get handed to init, unless they look like
263 * unused parameters (modprobe will find them in /proc/cmdline).
264 */
265static int __init unknown_bootoption(char *param, char *val, const char *unused)
266{
267 repair_env_string(param, val, unused);
268
269 /* Handle obsolete-style parameters */
270 if (obsolete_checksetup(param))
271 return 0;
272
273 /* Unused module parameter. */
274 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
275 return 0;
276
277 if (panic_later)
278 return 0;
279
280 if (val) {
281 /* Environment option */
282 unsigned int i;
283 for (i = 0; envp_init[i]; i++) {
284 if (i == MAX_INIT_ENVS) {
285 panic_later = "Too many boot env vars at `%s'";
286 panic_param = param;
287 }
288 if (!strncmp(param, envp_init[i], val - param))
289 break;
290 }
291 envp_init[i] = param;
292 } else {
293 /* Command line option */
294 unsigned int i;
295 for (i = 0; argv_init[i]; i++) {
296 if (i == MAX_INIT_ARGS) {
297 panic_later = "Too many boot init vars at `%s'";
298 panic_param = param;
299 }
300 }
301 argv_init[i] = param;
302 }
303 return 0;
304}
305
306static int __init init_setup(char *str)
307{
308 unsigned int i;
309
310 execute_command = str;
311 /*
312 * In case LILO is going to boot us with default command line,
313 * it prepends "auto" before the whole cmdline which makes
314 * the shell think it should execute a script with such name.
315 * So we ignore all arguments entered _before_ init=... [MJ]
316 */
317 for (i = 1; i < MAX_INIT_ARGS; i++)
318 argv_init[i] = NULL;
319 return 1;
320}
321__setup("init=", init_setup);
322
323static int __init rdinit_setup(char *str)
324{
325 unsigned int i;
326
327 ramdisk_execute_command = str;
328 /* See "auto" comment in init_setup */
329 for (i = 1; i < MAX_INIT_ARGS; i++)
330 argv_init[i] = NULL;
331 return 1;
332}
333__setup("rdinit=", rdinit_setup);
334
335#ifndef CONFIG_SMP
336static const unsigned int setup_max_cpus = NR_CPUS;
337#ifdef CONFIG_X86_LOCAL_APIC
338static void __init smp_init(void)
339{
340 APIC_init_uniprocessor();
341}
342#else
343#define smp_init() do { } while (0)
344#endif
345
346static inline void setup_nr_cpu_ids(void) { }
347static inline void smp_prepare_cpus(unsigned int maxcpus) { }
348#endif
349
350/*
351 * We need to store the untouched command line for future reference.
352 * We also need to store the touched command line since the parameter
353 * parsing is performed in place, and we should allow a component to
354 * store reference of name/value for future reference.
355 */
356static void __init setup_command_line(char *command_line)
357{
358 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
359 initcall_command_line = alloc_bootmem(strlen (boot_command_line)+1);
360 static_command_line = alloc_bootmem(strlen (command_line)+1);
361 strcpy (saved_command_line, boot_command_line);
362 strcpy (static_command_line, command_line);
363}
364
365/*
366 * We need to finalize in a non-__init function or else race conditions
367 * between the root thread and the init thread may cause start_kernel to
368 * be reaped by free_initmem before the root thread has proceeded to
369 * cpu_idle.
370 *
371 * gcc-3.4 accidentally inlines this function, so use noinline.
372 */
373
374static __initdata DECLARE_COMPLETION(kthreadd_done);
375
376static noinline void __init_refok rest_init(void)
377{
378 int pid;
379
380 rcu_scheduler_starting();
381 /*
382 * We need to spawn init first so that it obtains pid 1, however
383 * the init task will end up wanting to create kthreads, which, if
384 * we schedule it before we create kthreadd, will OOPS.
385 */
386 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
387 numa_default_policy();
388 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
389 rcu_read_lock();
390 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
391 rcu_read_unlock();
392 complete(&kthreadd_done);
393
394 /*
395 * The boot idle thread must execute schedule()
396 * at least once to get things moving:
397 */
398 init_idle_bootup_task(current);
399 schedule_preempt_disabled();
400 /* Call into cpu_idle with preempt disabled */
401 cpu_startup_entry(CPUHP_ONLINE);
402}
403
404/* Check for early params. */
405static int __init do_early_param(char *param, char *val, const char *unused)
406{
407 const struct obs_kernel_param *p;
408
409 for (p = __setup_start; p < __setup_end; p++) {
410 if ((p->early && parameq(param, p->str)) ||
411 (strcmp(param, "console") == 0 &&
412 strcmp(p->str, "earlycon") == 0)
413 ) {
414 if (p->setup_func(val) != 0)
415 pr_warn("Malformed early option '%s'\n", param);
416 }
417 }
418 /* We accept everything at this stage. */
419 return 0;
420}
421
422void __init parse_early_options(char *cmdline)
423{
424 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
425}
426
427/* Arch code calls this early on, or if not, just before other parsing. */
428void __init parse_early_param(void)
429{
430 static __initdata int done = 0;
431 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
432
433 if (done)
434 return;
435
436 /* All fall through to do_early_param. */
437 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
438 parse_early_options(tmp_cmdline);
439 done = 1;
440}
441
442/*
443 * Activate the first processor.
444 */
445
446static void __init boot_cpu_init(void)
447{
448 int cpu = smp_processor_id();
449 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
450 set_cpu_online(cpu, true);
451 set_cpu_active(cpu, true);
452 set_cpu_present(cpu, true);
453 set_cpu_possible(cpu, true);
454}
455
456void __init __weak smp_setup_processor_id(void)
457{
458}
459
460# if THREAD_SIZE >= PAGE_SIZE
461void __init __weak thread_info_cache_init(void)
462{
463}
464#endif
465
466/*
467 * Set up kernel memory allocators
468 */
469static void __init mm_init(void)
470{
471 /*
472 * page_cgroup requires contiguous pages,
473 * bigger than MAX_ORDER unless SPARSEMEM.
474 */
475 page_cgroup_init_flatmem();
476 mem_init();
477 kmem_cache_init();
478 percpu_init_late();
479 pgtable_cache_init();
480 vmalloc_init();
481}
482
483asmlinkage void __init start_kernel(void)
484{
485 char * command_line;
486 extern const struct kernel_param __start___param[], __stop___param[];
487
488 /*
489 * Need to run as early as possible, to initialize the
490 * lockdep hash:
491 */
492 lockdep_init();
493 smp_setup_processor_id();
494 debug_objects_early_init();
495
496 /*
497 * Set up the the initial canary ASAP:
498 */
499 boot_init_stack_canary();
500
501 cgroup_init_early();
502
503 local_irq_disable();
504 early_boot_irqs_disabled = true;
505
506/*
507 * Interrupts are still disabled. Do necessary setups, then
508 * enable them
509 */
510 boot_cpu_init();
511 page_address_init();
512 pr_notice("%s", linux_banner);
513 setup_arch(&command_line);
514 mm_init_owner(&init_mm, &init_task);
515 mm_init_cpumask(&init_mm);
516 setup_command_line(command_line);
517 setup_nr_cpu_ids();
518 setup_per_cpu_areas();
519 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
520
521 build_all_zonelists(NULL, NULL);
522 page_alloc_init();
523
524 pr_notice("Kernel command line: %s\n", boot_command_line);
525 parse_early_param();
526 parse_args("Booting kernel", static_command_line, __start___param,
527 __stop___param - __start___param,
528 -1, -1, &unknown_bootoption);
529
530 jump_label_init();
531
532 /*
533 * These use large bootmem allocations and must precede
534 * kmem_cache_init()
535 */
536 setup_log_buf(0);
537 pidhash_init();
538 vfs_caches_init_early();
539 sort_main_extable();
540 trap_init();
541 mm_init();
542
543 /*
544 * Set up the scheduler prior starting any interrupts (such as the
545 * timer interrupt). Full topology setup happens at smp_init()
546 * time - but meanwhile we still have a functioning scheduler.
547 */
548 sched_init();
549 /*
550 * Disable preemption - early bootup scheduling is extremely
551 * fragile until we cpu_idle() for the first time.
552 */
553 preempt_disable();
554 if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
555 local_irq_disable();
556 idr_init_cache();
557 rcu_init();
558 tick_nohz_init();
559 context_tracking_init();
560 radix_tree_init();
561 /* init some links before init_ISA_irqs() */
562 early_irq_init();
563 init_IRQ();
564 tick_init();
565 init_timers();
566 hrtimers_init();
567 softirq_init();
568 timekeeping_init();
569 time_init();
570 sched_clock_postinit();
571 perf_event_init();
572 profile_init();
573 call_function_init();
574 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
575 early_boot_irqs_disabled = false;
576 local_irq_enable();
577
578 kmem_cache_init_late();
579
580 /*
581 * HACK ALERT! This is early. We're enabling the console before
582 * we've done PCI setups etc, and console_init() must be aware of
583 * this. But we do want output early, in case something goes wrong.
584 */
585 console_init();
586 if (panic_later)
587 panic(panic_later, panic_param);
588
589 lockdep_info();
590
591 /*
592 * Need to run this when irqs are enabled, because it wants
593 * to self-test [hard/soft]-irqs on/off lock inversion bugs
594 * too:
595 */
596 locking_selftest();
597
598#ifdef CONFIG_BLK_DEV_INITRD
599 if (initrd_start && !initrd_below_start_ok &&
600 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
601 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
602 page_to_pfn(virt_to_page((void *)initrd_start)),
603 min_low_pfn);
604 initrd_start = 0;
605 }
606#endif
607 page_cgroup_init();
608 debug_objects_mem_init();
609 kmemleak_init();
610 setup_per_cpu_pageset();
611 numa_policy_init();
612 if (late_time_init)
613 late_time_init();
614 sched_clock_init();
615 calibrate_delay();
616 pidmap_init();
617 anon_vma_init();
618#ifdef CONFIG_X86
619 if (efi_enabled(EFI_RUNTIME_SERVICES))
620 efi_enter_virtual_mode();
621#endif
622 thread_info_cache_init();
623 cred_init();
624 fork_init(totalram_pages);
625 proc_caches_init();
626 buffer_init();
627 key_init();
628 security_init();
629 dbg_late_init();
630 vfs_caches_init(totalram_pages);
631 signals_init();
632 /* rootfs populating might need page-writeback */
633 page_writeback_init();
634#ifdef CONFIG_PROC_FS
635 proc_root_init();
636#endif
637 cgroup_init();
638 cpuset_init();
639 taskstats_init_early();
640 delayacct_init();
641
642 check_bugs();
643
644 acpi_early_init(); /* before LAPIC and SMP init */
645 sfi_init_late();
646
647 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
648 efi_late_init();
649 efi_free_boot_services();
650 }
651
652 ftrace_init();
653
654 /* Do the rest non-__init'ed, we're now alive */
655 rest_init();
656}
657
658/* Call all constructor functions linked into the kernel. */
659static void __init do_ctors(void)
660{
661#ifdef CONFIG_CONSTRUCTORS
662 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
663
664 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
665 (*fn)();
666#endif
667}
668
669bool initcall_debug;
670core_param(initcall_debug, initcall_debug, bool, 0644);
671
672static int __init_or_module do_one_initcall_debug(initcall_t fn)
673{
674 ktime_t calltime, delta, rettime;
675 unsigned long long duration;
676 int ret;
677
678 pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
679 calltime = ktime_get();
680 ret = fn();
681 rettime = ktime_get();
682 delta = ktime_sub(rettime, calltime);
683 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
684 pr_debug("initcall %pF returned %d after %lld usecs\n",
685 fn, ret, duration);
686
687 return ret;
688}
689
690int __init_or_module do_one_initcall(initcall_t fn)
691{
692 int count = preempt_count();
693 int ret;
694 char msgbuf[64];
695
696 if (initcall_debug)
697 ret = do_one_initcall_debug(fn);
698 else
699 ret = fn();
700
701 msgbuf[0] = 0;
702
703 if (preempt_count() != count) {
704 sprintf(msgbuf, "preemption imbalance ");
705 preempt_count_set(count);
706 }
707 if (irqs_disabled()) {
708 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
709 local_irq_enable();
710 }
711 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
712
713 return ret;
714}
715
716
717extern initcall_t __initcall_start[];
718extern initcall_t __initcall0_start[];
719extern initcall_t __initcall1_start[];
720extern initcall_t __initcall2_start[];
721extern initcall_t __initcall3_start[];
722extern initcall_t __initcall4_start[];
723extern initcall_t __initcall5_start[];
724extern initcall_t __initcall6_start[];
725extern initcall_t __initcall7_start[];
726extern initcall_t __initcall_end[];
727
728static initcall_t *initcall_levels[] __initdata = {
729 __initcall0_start,
730 __initcall1_start,
731 __initcall2_start,
732 __initcall3_start,
733 __initcall4_start,
734 __initcall5_start,
735 __initcall6_start,
736 __initcall7_start,
737 __initcall_end,
738};
739
740/* Keep these in sync with initcalls in include/linux/init.h */
741static char *initcall_level_names[] __initdata = {
742 "early",
743 "core",
744 "postcore",
745 "arch",
746 "subsys",
747 "fs",
748 "device",
749 "late",
750};
751
752static void __init do_initcall_level(int level)
753{
754 extern const struct kernel_param __start___param[], __stop___param[];
755 initcall_t *fn;
756
757 strcpy(initcall_command_line, saved_command_line);
758 parse_args(initcall_level_names[level],
759 initcall_command_line, __start___param,
760 __stop___param - __start___param,
761 level, level,
762 &repair_env_string);
763
764 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
765 do_one_initcall(*fn);
766}
767
768static void __init do_initcalls(void)
769{
770 int level;
771
772 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
773 do_initcall_level(level);
774}
775
776/*
777 * Ok, the machine is now initialized. None of the devices
778 * have been touched yet, but the CPU subsystem is up and
779 * running, and memory and process management works.
780 *
781 * Now we can finally start doing some real work..
782 */
783static void __init do_basic_setup(void)
784{
785 cpuset_init_smp();
786 usermodehelper_init();
787 shmem_init();
788 driver_init();
789 init_irq_proc();
790 do_ctors();
791 usermodehelper_enable();
792 do_initcalls();
793 random_int_secret_init();
794}
795
796static void __init do_pre_smp_initcalls(void)
797{
798 initcall_t *fn;
799
800 for (fn = __initcall_start; fn < __initcall0_start; fn++)
801 do_one_initcall(*fn);
802}
803
804/*
805 * This function requests modules which should be loaded by default and is
806 * called twice right after initrd is mounted and right before init is
807 * exec'd. If such modules are on either initrd or rootfs, they will be
808 * loaded before control is passed to userland.
809 */
810void __init load_default_modules(void)
811{
812 load_default_elevator_module();
813}
814
815static int run_init_process(const char *init_filename)
816{
817 argv_init[0] = init_filename;
818 return do_execve(init_filename,
819 (const char __user *const __user *)argv_init,
820 (const char __user *const __user *)envp_init);
821}
822
823static int try_to_run_init_process(const char *init_filename)
824{
825 int ret;
826
827 ret = run_init_process(init_filename);
828
829 if (ret && ret != -ENOENT) {
830 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
831 init_filename, ret);
832 }
833
834 return ret;
835}
836
837static noinline void __init kernel_init_freeable(void);
838
839static int __ref kernel_init(void *unused)
840{
841 int ret;
842
843 kernel_init_freeable();
844 /* need to finish all async __init code before freeing the memory */
845 async_synchronize_full();
846 free_initmem();
847 mark_rodata_ro();
848 system_state = SYSTEM_RUNNING;
849 numa_default_policy();
850
851 flush_delayed_fput();
852
853 if (ramdisk_execute_command) {
854 ret = run_init_process(ramdisk_execute_command);
855 if (!ret)
856 return 0;
857 pr_err("Failed to execute %s (error %d)\n",
858 ramdisk_execute_command, ret);
859 }
860
861 /*
862 * We try each of these until one succeeds.
863 *
864 * The Bourne shell can be used instead of init if we are
865 * trying to recover a really broken machine.
866 */
867 if (execute_command) {
868 ret = run_init_process(execute_command);
869 if (!ret)
870 return 0;
871 pr_err("Failed to execute %s (error %d). Attempting defaults...\n",
872 execute_command, ret);
873 }
874 if (!try_to_run_init_process("/sbin/init") ||
875 !try_to_run_init_process("/etc/init") ||
876 !try_to_run_init_process("/bin/init") ||
877 !try_to_run_init_process("/bin/sh"))
878 return 0;
879
880 panic("No working init found. Try passing init= option to kernel. "
881 "See Linux Documentation/init.txt for guidance.");
882}
883
884static noinline void __init kernel_init_freeable(void)
885{
886 /*
887 * Wait until kthreadd is all set-up.
888 */
889 wait_for_completion(&kthreadd_done);
890
891 /* Now the scheduler is fully set up and can do blocking allocations */
892 gfp_allowed_mask = __GFP_BITS_MASK;
893
894 /*
895 * init can allocate pages on any node
896 */
897 set_mems_allowed(node_states[N_MEMORY]);
898 /*
899 * init can run on any cpu.
900 */
901 set_cpus_allowed_ptr(current, cpu_all_mask);
902
903 cad_pid = task_pid(current);
904
905 smp_prepare_cpus(setup_max_cpus);
906
907 do_pre_smp_initcalls();
908 lockup_detector_init();
909
910 smp_init();
911 sched_init_smp();
912
913 do_basic_setup();
914
915 /* Open the /dev/console on the rootfs, this should never fail */
916 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
917 pr_err("Warning: unable to open an initial console.\n");
918
919 (void) sys_dup(0);
920 (void) sys_dup(0);
921 /*
922 * check if there is an early userspace init. If yes, let it do all
923 * the work
924 */
925
926 if (!ramdisk_execute_command)
927 ramdisk_execute_command = "/init";
928
929 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
930 ramdisk_execute_command = NULL;
931 prepare_namespace();
932 }
933
934 /*
935 * Ok, we have completed the initial bootup, and
936 * we're essentially up and running. Get rid of the
937 * initmem segments and start the user-mode stuff..
938 */
939
940 /* rootfs is available now, try loading default modules */
941 load_default_modules();
942}