]> bbs.cooldavid.org Git - net-next-2.6.git/blame_incremental - init/main.c
Merge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
[net-next-2.6.git] / init / main.c
... / ...
CommitLineData
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/smp_lock.h>
24#include <linux/initrd.h>
25#include <linux/bootmem.h>
26#include <linux/acpi.h>
27#include <linux/tty.h>
28#include <linux/gfp.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/workqueue.h>
37#include <linux/profile.h>
38#include <linux/rcupdate.h>
39#include <linux/moduleparam.h>
40#include <linux/kallsyms.h>
41#include <linux/writeback.h>
42#include <linux/cpu.h>
43#include <linux/cpuset.h>
44#include <linux/cgroup.h>
45#include <linux/efi.h>
46#include <linux/tick.h>
47#include <linux/interrupt.h>
48#include <linux/taskstats_kern.h>
49#include <linux/delayacct.h>
50#include <linux/unistd.h>
51#include <linux/rmap.h>
52#include <linux/mempolicy.h>
53#include <linux/key.h>
54#include <linux/buffer_head.h>
55#include <linux/page_cgroup.h>
56#include <linux/debug_locks.h>
57#include <linux/debugobjects.h>
58#include <linux/lockdep.h>
59#include <linux/kmemleak.h>
60#include <linux/pid_namespace.h>
61#include <linux/device.h>
62#include <linux/kthread.h>
63#include <linux/sched.h>
64#include <linux/signal.h>
65#include <linux/idr.h>
66#include <linux/ftrace.h>
67#include <linux/async.h>
68#include <linux/kmemcheck.h>
69#include <linux/kmemtrace.h>
70#include <linux/sfi.h>
71#include <linux/shmem_fs.h>
72#include <trace/boot.h>
73
74#include <asm/io.h>
75#include <asm/bugs.h>
76#include <asm/setup.h>
77#include <asm/sections.h>
78#include <asm/cacheflush.h>
79
80#ifdef CONFIG_X86_LOCAL_APIC
81#include <asm/smp.h>
82#endif
83
84static int kernel_init(void *);
85
86extern void init_IRQ(void);
87extern void fork_init(unsigned long);
88extern void mca_init(void);
89extern void sbus_init(void);
90extern void prio_tree_init(void);
91extern void radix_tree_init(void);
92extern void free_initmem(void);
93#ifndef CONFIG_DEBUG_RODATA
94static inline void mark_rodata_ro(void) { }
95#endif
96
97#ifdef CONFIG_TC
98extern void tc_init(void);
99#endif
100
101enum system_states system_state __read_mostly;
102EXPORT_SYMBOL(system_state);
103
104/*
105 * Boot command-line arguments
106 */
107#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
108#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
109
110extern void time_init(void);
111/* Default late time init is NULL. archs can override this later. */
112void (*__initdata late_time_init)(void);
113extern void softirq_init(void);
114
115/* Untouched command line saved by arch-specific code. */
116char __initdata boot_command_line[COMMAND_LINE_SIZE];
117/* Untouched saved command line (eg. for /proc) */
118char *saved_command_line;
119/* Command line for parameter parsing */
120static char *static_command_line;
121
122static char *execute_command;
123static char *ramdisk_execute_command;
124
125#ifdef CONFIG_SMP
126/* Setup configured maximum number of CPUs to activate */
127unsigned int __initdata setup_max_cpus = NR_CPUS;
128
129/*
130 * Setup routine for controlling SMP activation
131 *
132 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
133 * activation entirely (the MPS table probe still happens, though).
134 *
135 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
136 * greater than 0, limits the maximum number of CPUs activated in
137 * SMP mode to <NUM>.
138 */
139
140void __weak arch_disable_smp_support(void) { }
141
142static int __init nosmp(char *str)
143{
144 setup_max_cpus = 0;
145 arch_disable_smp_support();
146
147 return 0;
148}
149
150early_param("nosmp", nosmp);
151
152static int __init maxcpus(char *str)
153{
154 get_option(&str, &setup_max_cpus);
155 if (setup_max_cpus == 0)
156 arch_disable_smp_support();
157
158 return 0;
159}
160
161early_param("maxcpus", maxcpus);
162#else
163const unsigned int setup_max_cpus = NR_CPUS;
164#endif
165
166/*
167 * If set, this is an indication to the drivers that reset the underlying
168 * device before going ahead with the initialization otherwise driver might
169 * rely on the BIOS and skip the reset operation.
170 *
171 * This is useful if kernel is booting in an unreliable environment.
172 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
173 * skipped and devices will be in unknown state.
174 */
175unsigned int reset_devices;
176EXPORT_SYMBOL(reset_devices);
177
178static int __init set_reset_devices(char *str)
179{
180 reset_devices = 1;
181 return 1;
182}
183
184__setup("reset_devices", set_reset_devices);
185
186static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
187char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
188static const char *panic_later, *panic_param;
189
190extern struct obs_kernel_param __setup_start[], __setup_end[];
191
192static int __init obsolete_checksetup(char *line)
193{
194 struct obs_kernel_param *p;
195 int had_early_param = 0;
196
197 p = __setup_start;
198 do {
199 int n = strlen(p->str);
200 if (!strncmp(line, p->str, n)) {
201 if (p->early) {
202 /* Already done in parse_early_param?
203 * (Needs exact match on param part).
204 * Keep iterating, as we can have early
205 * params and __setups of same names 8( */
206 if (line[n] == '\0' || line[n] == '=')
207 had_early_param = 1;
208 } else if (!p->setup_func) {
209 printk(KERN_WARNING "Parameter %s is obsolete,"
210 " ignored\n", p->str);
211 return 1;
212 } else if (p->setup_func(line + n))
213 return 1;
214 }
215 p++;
216 } while (p < __setup_end);
217
218 return had_early_param;
219}
220
221/*
222 * This should be approx 2 Bo*oMips to start (note initial shift), and will
223 * still work even if initially too large, it will just take slightly longer
224 */
225unsigned long loops_per_jiffy = (1<<12);
226
227EXPORT_SYMBOL(loops_per_jiffy);
228
229static int __init debug_kernel(char *str)
230{
231 console_loglevel = 10;
232 return 0;
233}
234
235static int __init quiet_kernel(char *str)
236{
237 console_loglevel = 4;
238 return 0;
239}
240
241early_param("debug", debug_kernel);
242early_param("quiet", quiet_kernel);
243
244static int __init loglevel(char *str)
245{
246 get_option(&str, &console_loglevel);
247 return 0;
248}
249
250early_param("loglevel", loglevel);
251
252/*
253 * Unknown boot options get handed to init, unless they look like
254 * failed parameters
255 */
256static int __init unknown_bootoption(char *param, char *val)
257{
258 /* Change NUL term back to "=", to make "param" the whole string. */
259 if (val) {
260 /* param=val or param="val"? */
261 if (val == param+strlen(param)+1)
262 val[-1] = '=';
263 else if (val == param+strlen(param)+2) {
264 val[-2] = '=';
265 memmove(val-1, val, strlen(val)+1);
266 val--;
267 } else
268 BUG();
269 }
270
271 /* Handle obsolete-style parameters */
272 if (obsolete_checksetup(param))
273 return 0;
274
275 /*
276 * Preemptive maintenance for "why didn't my misspelled command
277 * line work?"
278 */
279 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
280 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
281 return 0;
282 }
283
284 if (panic_later)
285 return 0;
286
287 if (val) {
288 /* Environment option */
289 unsigned int i;
290 for (i = 0; envp_init[i]; i++) {
291 if (i == MAX_INIT_ENVS) {
292 panic_later = "Too many boot env vars at `%s'";
293 panic_param = param;
294 }
295 if (!strncmp(param, envp_init[i], val - param))
296 break;
297 }
298 envp_init[i] = param;
299 } else {
300 /* Command line option */
301 unsigned int i;
302 for (i = 0; argv_init[i]; i++) {
303 if (i == MAX_INIT_ARGS) {
304 panic_later = "Too many boot init vars at `%s'";
305 panic_param = param;
306 }
307 }
308 argv_init[i] = param;
309 }
310 return 0;
311}
312
313#ifdef CONFIG_DEBUG_PAGEALLOC
314int __read_mostly debug_pagealloc_enabled = 0;
315#endif
316
317static int __init init_setup(char *str)
318{
319 unsigned int i;
320
321 execute_command = str;
322 /*
323 * In case LILO is going to boot us with default command line,
324 * it prepends "auto" before the whole cmdline which makes
325 * the shell think it should execute a script with such name.
326 * So we ignore all arguments entered _before_ init=... [MJ]
327 */
328 for (i = 1; i < MAX_INIT_ARGS; i++)
329 argv_init[i] = NULL;
330 return 1;
331}
332__setup("init=", init_setup);
333
334static int __init rdinit_setup(char *str)
335{
336 unsigned int i;
337
338 ramdisk_execute_command = str;
339 /* See "auto" comment in init_setup */
340 for (i = 1; i < MAX_INIT_ARGS; i++)
341 argv_init[i] = NULL;
342 return 1;
343}
344__setup("rdinit=", rdinit_setup);
345
346#ifndef CONFIG_SMP
347
348#ifdef CONFIG_X86_LOCAL_APIC
349static void __init smp_init(void)
350{
351 APIC_init_uniprocessor();
352}
353#else
354#define smp_init() do { } while (0)
355#endif
356
357static inline void setup_nr_cpu_ids(void) { }
358static inline void smp_prepare_cpus(unsigned int maxcpus) { }
359
360#else
361
362/* Setup number of possible processor ids */
363int nr_cpu_ids __read_mostly = NR_CPUS;
364EXPORT_SYMBOL(nr_cpu_ids);
365
366/* An arch may set nr_cpu_ids earlier if needed, so this would be redundant */
367static void __init setup_nr_cpu_ids(void)
368{
369 nr_cpu_ids = find_last_bit(cpumask_bits(cpu_possible_mask),NR_CPUS) + 1;
370}
371
372/* Called by boot processor to activate the rest. */
373static void __init smp_init(void)
374{
375 unsigned int cpu;
376
377 /*
378 * Set up the current CPU as possible to migrate to.
379 * The other ones will be done by cpu_up/cpu_down()
380 */
381 set_cpu_active(smp_processor_id(), true);
382
383 /* FIXME: This should be done in userspace --RR */
384 for_each_present_cpu(cpu) {
385 if (num_online_cpus() >= setup_max_cpus)
386 break;
387 if (!cpu_online(cpu))
388 cpu_up(cpu);
389 }
390
391 /* Any cleanup work */
392 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
393 smp_cpus_done(setup_max_cpus);
394}
395
396#endif
397
398/*
399 * We need to store the untouched command line for future reference.
400 * We also need to store the touched command line since the parameter
401 * parsing is performed in place, and we should allow a component to
402 * store reference of name/value for future reference.
403 */
404static void __init setup_command_line(char *command_line)
405{
406 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
407 static_command_line = alloc_bootmem(strlen (command_line)+1);
408 strcpy (saved_command_line, boot_command_line);
409 strcpy (static_command_line, command_line);
410}
411
412/*
413 * We need to finalize in a non-__init function or else race conditions
414 * between the root thread and the init thread may cause start_kernel to
415 * be reaped by free_initmem before the root thread has proceeded to
416 * cpu_idle.
417 *
418 * gcc-3.4 accidentally inlines this function, so use noinline.
419 */
420
421static noinline void __init_refok rest_init(void)
422 __releases(kernel_lock)
423{
424 int pid;
425
426 rcu_scheduler_starting();
427 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
428 numa_default_policy();
429 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
430 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
431 unlock_kernel();
432
433 /*
434 * The boot idle thread must execute schedule()
435 * at least once to get things moving:
436 */
437 init_idle_bootup_task(current);
438 preempt_enable_no_resched();
439 schedule();
440 preempt_disable();
441
442 /* Call into cpu_idle with preempt disabled */
443 cpu_idle();
444}
445
446/* Check for early params. */
447static int __init do_early_param(char *param, char *val)
448{
449 struct obs_kernel_param *p;
450
451 for (p = __setup_start; p < __setup_end; p++) {
452 if ((p->early && strcmp(param, p->str) == 0) ||
453 (strcmp(param, "console") == 0 &&
454 strcmp(p->str, "earlycon") == 0)
455 ) {
456 if (p->setup_func(val) != 0)
457 printk(KERN_WARNING
458 "Malformed early option '%s'\n", param);
459 }
460 }
461 /* We accept everything at this stage. */
462 return 0;
463}
464
465void __init parse_early_options(char *cmdline)
466{
467 parse_args("early options", cmdline, NULL, 0, do_early_param);
468}
469
470/* Arch code calls this early on, or if not, just before other parsing. */
471void __init parse_early_param(void)
472{
473 static __initdata int done = 0;
474 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
475
476 if (done)
477 return;
478
479 /* All fall through to do_early_param. */
480 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
481 parse_early_options(tmp_cmdline);
482 done = 1;
483}
484
485/*
486 * Activate the first processor.
487 */
488
489static void __init boot_cpu_init(void)
490{
491 int cpu = smp_processor_id();
492 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
493 set_cpu_online(cpu, true);
494 set_cpu_present(cpu, true);
495 set_cpu_possible(cpu, true);
496}
497
498void __init __weak smp_setup_processor_id(void)
499{
500}
501
502void __init __weak thread_info_cache_init(void)
503{
504}
505
506/*
507 * Set up kernel memory allocators
508 */
509static void __init mm_init(void)
510{
511 /*
512 * page_cgroup requires countinous pages as memmap
513 * and it's bigger than MAX_ORDER unless SPARSEMEM.
514 */
515 page_cgroup_init_flatmem();
516 mem_init();
517 kmem_cache_init();
518 pgtable_cache_init();
519 vmalloc_init();
520}
521
522asmlinkage void __init start_kernel(void)
523{
524 char * command_line;
525 extern struct kernel_param __start___param[], __stop___param[];
526
527 smp_setup_processor_id();
528
529 /*
530 * Need to run as early as possible, to initialize the
531 * lockdep hash:
532 */
533 lockdep_init();
534 debug_objects_early_init();
535
536 /*
537 * Set up the the initial canary ASAP:
538 */
539 boot_init_stack_canary();
540
541 cgroup_init_early();
542
543 local_irq_disable();
544 early_boot_irqs_off();
545 early_init_irq_lock_class();
546
547/*
548 * Interrupts are still disabled. Do necessary setups, then
549 * enable them
550 */
551 lock_kernel();
552 tick_init();
553 boot_cpu_init();
554 page_address_init();
555 printk(KERN_NOTICE "%s", linux_banner);
556 setup_arch(&command_line);
557 mm_init_owner(&init_mm, &init_task);
558 setup_command_line(command_line);
559 setup_nr_cpu_ids();
560 setup_per_cpu_areas();
561 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
562
563 build_all_zonelists();
564 page_alloc_init();
565
566 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
567 parse_early_param();
568 parse_args("Booting kernel", static_command_line, __start___param,
569 __stop___param - __start___param,
570 &unknown_bootoption);
571 /*
572 * These use large bootmem allocations and must precede
573 * kmem_cache_init()
574 */
575 pidhash_init();
576 vfs_caches_init_early();
577 sort_main_extable();
578 trap_init();
579 mm_init();
580 /*
581 * Set up the scheduler prior starting any interrupts (such as the
582 * timer interrupt). Full topology setup happens at smp_init()
583 * time - but meanwhile we still have a functioning scheduler.
584 */
585 sched_init();
586 /*
587 * Disable preemption - early bootup scheduling is extremely
588 * fragile until we cpu_idle() for the first time.
589 */
590 preempt_disable();
591 if (!irqs_disabled()) {
592 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
593 "enabled *very* early, fixing it\n");
594 local_irq_disable();
595 }
596 rcu_init();
597 /* init some links before init_ISA_irqs() */
598 early_irq_init();
599 init_IRQ();
600 prio_tree_init();
601 init_timers();
602 hrtimers_init();
603 softirq_init();
604 timekeeping_init();
605 time_init();
606 profile_init();
607 if (!irqs_disabled())
608 printk(KERN_CRIT "start_kernel(): bug: interrupts were "
609 "enabled early\n");
610 early_boot_irqs_on();
611 local_irq_enable();
612
613 /* Interrupts are enabled now so all GFP allocations are safe. */
614 set_gfp_allowed_mask(__GFP_BITS_MASK);
615
616 kmem_cache_init_late();
617
618 /*
619 * HACK ALERT! This is early. We're enabling the console before
620 * we've done PCI setups etc, and console_init() must be aware of
621 * this. But we do want output early, in case something goes wrong.
622 */
623 console_init();
624 if (panic_later)
625 panic(panic_later, panic_param);
626
627 lockdep_info();
628
629 /*
630 * Need to run this when irqs are enabled, because it wants
631 * to self-test [hard/soft]-irqs on/off lock inversion bugs
632 * too:
633 */
634 locking_selftest();
635
636#ifdef CONFIG_BLK_DEV_INITRD
637 if (initrd_start && !initrd_below_start_ok &&
638 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
639 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
640 "disabling it.\n",
641 page_to_pfn(virt_to_page((void *)initrd_start)),
642 min_low_pfn);
643 initrd_start = 0;
644 }
645#endif
646 page_cgroup_init();
647 enable_debug_pagealloc();
648 kmemtrace_init();
649 kmemleak_init();
650 debug_objects_mem_init();
651 idr_init_cache();
652 setup_per_cpu_pageset();
653 numa_policy_init();
654 if (late_time_init)
655 late_time_init();
656 sched_clock_init();
657 calibrate_delay();
658 pidmap_init();
659 anon_vma_init();
660#ifdef CONFIG_X86
661 if (efi_enabled)
662 efi_enter_virtual_mode();
663#endif
664 thread_info_cache_init();
665 cred_init();
666 fork_init(totalram_pages);
667 proc_caches_init();
668 buffer_init();
669 key_init();
670 security_init();
671 vfs_caches_init(totalram_pages);
672 radix_tree_init();
673 signals_init();
674 /* rootfs populating might need page-writeback */
675 page_writeback_init();
676#ifdef CONFIG_PROC_FS
677 proc_root_init();
678#endif
679 cgroup_init();
680 cpuset_init();
681 taskstats_init_early();
682 delayacct_init();
683
684 check_bugs();
685
686 acpi_early_init(); /* before LAPIC and SMP init */
687 sfi_init_late();
688
689 ftrace_init();
690
691 /* Do the rest non-__init'ed, we're now alive */
692 rest_init();
693}
694
695/* Call all constructor functions linked into the kernel. */
696static void __init do_ctors(void)
697{
698#ifdef CONFIG_CONSTRUCTORS
699 ctor_fn_t *call = (ctor_fn_t *) __ctors_start;
700
701 for (; call < (ctor_fn_t *) __ctors_end; call++)
702 (*call)();
703#endif
704}
705
706int initcall_debug;
707core_param(initcall_debug, initcall_debug, bool, 0644);
708
709static char msgbuf[64];
710static struct boot_trace_call call;
711static struct boot_trace_ret ret;
712
713int do_one_initcall(initcall_t fn)
714{
715 int count = preempt_count();
716 ktime_t calltime, delta, rettime;
717
718 if (initcall_debug) {
719 call.caller = task_pid_nr(current);
720 printk("calling %pF @ %i\n", fn, call.caller);
721 calltime = ktime_get();
722 trace_boot_call(&call, fn);
723 enable_boot_trace();
724 }
725
726 ret.result = fn();
727
728 if (initcall_debug) {
729 disable_boot_trace();
730 rettime = ktime_get();
731 delta = ktime_sub(rettime, calltime);
732 ret.duration = (unsigned long long) ktime_to_ns(delta) >> 10;
733 trace_boot_ret(&ret, fn);
734 printk("initcall %pF returned %d after %Ld usecs\n", fn,
735 ret.result, ret.duration);
736 }
737
738 msgbuf[0] = 0;
739
740 if (ret.result && ret.result != -ENODEV && initcall_debug)
741 sprintf(msgbuf, "error code %d ", ret.result);
742
743 if (preempt_count() != count) {
744 strlcat(msgbuf, "preemption imbalance ", sizeof(msgbuf));
745 preempt_count() = count;
746 }
747 if (irqs_disabled()) {
748 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
749 local_irq_enable();
750 }
751 if (msgbuf[0]) {
752 printk("initcall %pF returned with %s\n", fn, msgbuf);
753 }
754
755 return ret.result;
756}
757
758
759extern initcall_t __initcall_start[], __initcall_end[], __early_initcall_end[];
760
761static void __init do_initcalls(void)
762{
763 initcall_t *call;
764
765 for (call = __early_initcall_end; call < __initcall_end; call++)
766 do_one_initcall(*call);
767
768 /* Make sure there is no pending stuff from the initcall sequence */
769 flush_scheduled_work();
770}
771
772/*
773 * Ok, the machine is now initialized. None of the devices
774 * have been touched yet, but the CPU subsystem is up and
775 * running, and memory and process management works.
776 *
777 * Now we can finally start doing some real work..
778 */
779static void __init do_basic_setup(void)
780{
781 init_workqueues();
782 cpuset_init_smp();
783 usermodehelper_init();
784 init_tmpfs();
785 driver_init();
786 init_irq_proc();
787 do_ctors();
788 do_initcalls();
789}
790
791static void __init do_pre_smp_initcalls(void)
792{
793 initcall_t *call;
794
795 for (call = __initcall_start; call < __early_initcall_end; call++)
796 do_one_initcall(*call);
797}
798
799static void run_init_process(char *init_filename)
800{
801 argv_init[0] = init_filename;
802 kernel_execve(init_filename, argv_init, envp_init);
803}
804
805/* This is a non __init function. Force it to be noinline otherwise gcc
806 * makes it inline to init() and it becomes part of init.text section
807 */
808static noinline int init_post(void)
809 __releases(kernel_lock)
810{
811 /* need to finish all async __init code before freeing the memory */
812 async_synchronize_full();
813 free_initmem();
814 unlock_kernel();
815 mark_rodata_ro();
816 system_state = SYSTEM_RUNNING;
817 numa_default_policy();
818
819 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
820 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
821
822 (void) sys_dup(0);
823 (void) sys_dup(0);
824
825 current->signal->flags |= SIGNAL_UNKILLABLE;
826
827 if (ramdisk_execute_command) {
828 run_init_process(ramdisk_execute_command);
829 printk(KERN_WARNING "Failed to execute %s\n",
830 ramdisk_execute_command);
831 }
832
833 /*
834 * We try each of these until one succeeds.
835 *
836 * The Bourne shell can be used instead of init if we are
837 * trying to recover a really broken machine.
838 */
839 if (execute_command) {
840 run_init_process(execute_command);
841 printk(KERN_WARNING "Failed to execute %s. Attempting "
842 "defaults...\n", execute_command);
843 }
844 run_init_process("/sbin/init");
845 run_init_process("/etc/init");
846 run_init_process("/bin/init");
847 run_init_process("/bin/sh");
848
849 panic("No init found. Try passing init= option to kernel.");
850}
851
852static int __init kernel_init(void * unused)
853{
854 lock_kernel();
855
856 /*
857 * init can allocate pages on any node
858 */
859 set_mems_allowed(node_possible_map);
860 /*
861 * init can run on any cpu.
862 */
863 set_cpus_allowed_ptr(current, cpu_all_mask);
864 /*
865 * Tell the world that we're going to be the grim
866 * reaper of innocent orphaned children.
867 *
868 * We don't want people to have to make incorrect
869 * assumptions about where in the task array this
870 * can be found.
871 */
872 init_pid_ns.child_reaper = current;
873
874 cad_pid = task_pid(current);
875
876 smp_prepare_cpus(setup_max_cpus);
877
878 do_pre_smp_initcalls();
879 start_boot_trace();
880
881 smp_init();
882 sched_init_smp();
883
884 do_basic_setup();
885
886 /*
887 * check if there is an early userspace init. If yes, let it do all
888 * the work
889 */
890
891 if (!ramdisk_execute_command)
892 ramdisk_execute_command = "/init";
893
894 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
895 ramdisk_execute_command = NULL;
896 prepare_namespace();
897 }
898
899 /*
900 * Ok, we have completed the initial bootup, and
901 * we're essentially up and running. Get rid of the
902 * initmem segments and start the user-mode stuff..
903 */
904
905 init_post();
906 return 0;
907}