blob: 8342c2890b16dd364f9222855e9a6bb7b5b8fc7d [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/init/main.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * GK 2/5/95 - Changed to support mounting root fs via NFS
7 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
8 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
9 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
10 */
11
12#define __KERNEL_SYSCALLS__
13
14#include <linux/config.h>
15#include <linux/types.h>
16#include <linux/module.h>
17#include <linux/proc_fs.h>
18#include <linux/devfs_fs_kernel.h>
19#include <linux/kernel.h>
20#include <linux/syscalls.h>
21#include <linux/string.h>
22#include <linux/ctype.h>
23#include <linux/delay.h>
24#include <linux/utsname.h>
25#include <linux/ioport.h>
26#include <linux/init.h>
27#include <linux/smp_lock.h>
28#include <linux/initrd.h>
29#include <linux/hdreg.h>
30#include <linux/bootmem.h>
31#include <linux/tty.h>
32#include <linux/gfp.h>
33#include <linux/percpu.h>
34#include <linux/kmod.h>
35#include <linux/kernel_stat.h>
36#include <linux/security.h>
37#include <linux/workqueue.h>
38#include <linux/profile.h>
39#include <linux/rcupdate.h>
40#include <linux/moduleparam.h>
41#include <linux/kallsyms.h>
42#include <linux/writeback.h>
43#include <linux/cpu.h>
44#include <linux/cpuset.h>
45#include <linux/efi.h>
46#include <linux/unistd.h>
47#include <linux/rmap.h>
48#include <linux/mempolicy.h>
49#include <linux/key.h>
50
51#include <asm/io.h>
52#include <asm/bugs.h>
53#include <asm/setup.h>
Andi Kleena9401992005-07-28 21:15:30 -070054#include <asm/sections.h>
Arjan van de Ven37b73c82006-01-06 00:12:01 -080055#include <asm/cacheflush.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
57/*
58 * This is one of the first .c files built. Error out early
59 * if we have compiler trouble..
60 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
62#ifdef CONFIG_X86_LOCAL_APIC
63#include <asm/smp.h>
64#endif
65
66/*
67 * Versions of gcc older than that listed below may actually compile
68 * and link okay, but the end product can have subtle run time bugs.
69 * To avoid associated bogus bug reports, we flatly refuse to compile
70 * with a gcc that is known to be too old from the very beginning.
71 */
Andrew Mortonfd285bb2006-01-08 01:04:07 -080072#if (__GNUC__ < 3) || (__GNUC__ == 3 && __GNUC_MINOR__ < 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#error Sorry, your GCC is too old. It builds incorrect kernels.
74#endif
75
76static int init(void *);
77
78extern void init_IRQ(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070079extern void fork_init(unsigned long);
80extern void mca_init(void);
81extern void sbus_init(void);
82extern void sysctl_init(void);
83extern void signals_init(void);
84extern void buffer_init(void);
85extern void pidhash_init(void);
86extern void pidmap_init(void);
87extern void prio_tree_init(void);
88extern void radix_tree_init(void);
89extern void free_initmem(void);
90extern void populate_rootfs(void);
91extern void driver_init(void);
92extern void prepare_namespace(void);
93#ifdef CONFIG_ACPI
94extern void acpi_early_init(void);
95#else
96static inline void acpi_early_init(void) { }
97#endif
Arjan van de Ven37b73c82006-01-06 00:12:01 -080098#ifndef CONFIG_DEBUG_RODATA
99static inline void mark_rodata_ro(void) { }
100#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
102#ifdef CONFIG_TC
103extern void tc_init(void);
104#endif
105
106enum system_states system_state;
107EXPORT_SYMBOL(system_state);
108
109/*
110 * Boot command-line arguments
111 */
112#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
113#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
114
115extern void time_init(void);
116/* Default late time init is NULL. archs can override this later. */
117void (*late_time_init)(void);
118extern void softirq_init(void);
119
120/* Untouched command line (eg. for /proc) saved by arch-specific code. */
121char saved_command_line[COMMAND_LINE_SIZE];
122
123static char *execute_command;
Olof Johanssonffdfc402005-09-06 15:17:19 -0700124static char *ramdisk_execute_command;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126/* Setup configured maximum number of CPUs to activate */
127static unsigned int 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 */
139static int __init nosmp(char *str)
140{
141 max_cpus = 0;
142 return 1;
143}
144
145__setup("nosmp", nosmp);
146
147static int __init maxcpus(char *str)
148{
149 get_option(&str, &max_cpus);
150 return 1;
151}
152
153__setup("maxcpus=", maxcpus);
154
155static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
156char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
157static const char *panic_later, *panic_param;
158
159extern struct obs_kernel_param __setup_start[], __setup_end[];
160
161static int __init obsolete_checksetup(char *line)
162{
163 struct obs_kernel_param *p;
164
165 p = __setup_start;
166 do {
167 int n = strlen(p->str);
168 if (!strncmp(line, p->str, n)) {
169 if (p->early) {
170 /* Already done in parse_early_param? (Needs
171 * exact match on param part) */
172 if (line[n] == '\0' || line[n] == '=')
173 return 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 return 0;
184}
185
186/*
187 * This should be approx 2 Bo*oMips to start (note initial shift), and will
188 * still work even if initially too large, it will just take slightly longer
189 */
190unsigned long loops_per_jiffy = (1<<12);
191
192EXPORT_SYMBOL(loops_per_jiffy);
193
194static int __init debug_kernel(char *str)
195{
196 if (*str)
197 return 0;
198 console_loglevel = 10;
199 return 1;
200}
201
202static int __init quiet_kernel(char *str)
203{
204 if (*str)
205 return 0;
206 console_loglevel = 4;
207 return 1;
208}
209
210__setup("debug", debug_kernel);
211__setup("quiet", quiet_kernel);
212
213static int __init loglevel(char *str)
214{
215 get_option(&str, &console_loglevel);
216 return 1;
217}
218
219__setup("loglevel=", loglevel);
220
221/*
222 * Unknown boot options get handed to init, unless they look like
223 * failed parameters
224 */
225static int __init unknown_bootoption(char *param, char *val)
226{
227 /* Change NUL term back to "=", to make "param" the whole string. */
228 if (val) {
229 /* param=val or param="val"? */
230 if (val == param+strlen(param)+1)
231 val[-1] = '=';
232 else if (val == param+strlen(param)+2) {
233 val[-2] = '=';
234 memmove(val-1, val, strlen(val)+1);
235 val--;
236 } else
237 BUG();
238 }
239
240 /* Handle obsolete-style parameters */
241 if (obsolete_checksetup(param))
242 return 0;
243
244 /*
245 * Preemptive maintenance for "why didn't my mispelled command
246 * line work?"
247 */
248 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
249 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
250 return 0;
251 }
252
253 if (panic_later)
254 return 0;
255
256 if (val) {
257 /* Environment option */
258 unsigned int i;
259 for (i = 0; envp_init[i]; i++) {
260 if (i == MAX_INIT_ENVS) {
261 panic_later = "Too many boot env vars at `%s'";
262 panic_param = param;
263 }
264 if (!strncmp(param, envp_init[i], val - param))
265 break;
266 }
267 envp_init[i] = param;
268 } else {
269 /* Command line option */
270 unsigned int i;
271 for (i = 0; argv_init[i]; i++) {
272 if (i == MAX_INIT_ARGS) {
273 panic_later = "Too many boot init vars at `%s'";
274 panic_param = param;
275 }
276 }
277 argv_init[i] = param;
278 }
279 return 0;
280}
281
282static int __init init_setup(char *str)
283{
284 unsigned int i;
285
286 execute_command = str;
287 /*
288 * In case LILO is going to boot us with default command line,
289 * it prepends "auto" before the whole cmdline which makes
290 * the shell think it should execute a script with such name.
291 * So we ignore all arguments entered _before_ init=... [MJ]
292 */
293 for (i = 1; i < MAX_INIT_ARGS; i++)
294 argv_init[i] = NULL;
295 return 1;
296}
297__setup("init=", init_setup);
298
Olof Johanssonffdfc402005-09-06 15:17:19 -0700299static int __init rdinit_setup(char *str)
300{
301 unsigned int i;
302
303 ramdisk_execute_command = str;
304 /* See "auto" comment in init_setup */
305 for (i = 1; i < MAX_INIT_ARGS; i++)
306 argv_init[i] = NULL;
307 return 1;
308}
309__setup("rdinit=", rdinit_setup);
310
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311extern void setup_arch(char **);
312
313#ifndef CONFIG_SMP
314
315#ifdef CONFIG_X86_LOCAL_APIC
316static void __init smp_init(void)
317{
318 APIC_init_uniprocessor();
319}
320#else
321#define smp_init() do { } while (0)
322#endif
323
324static inline void setup_per_cpu_areas(void) { }
325static inline void smp_prepare_cpus(unsigned int maxcpus) { }
326
327#else
328
329#ifdef __GENERIC_PER_CPU
330unsigned long __per_cpu_offset[NR_CPUS];
331
332EXPORT_SYMBOL(__per_cpu_offset);
333
334static void __init setup_per_cpu_areas(void)
335{
336 unsigned long size, i;
337 char *ptr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
339 /* Copy section for each CPU (we discard the original) */
340 size = ALIGN(__per_cpu_end - __per_cpu_start, SMP_CACHE_BYTES);
341#ifdef CONFIG_MODULES
342 if (size < PERCPU_ENOUGH_ROOM)
343 size = PERCPU_ENOUGH_ROOM;
344#endif
345
346 ptr = alloc_bootmem(size * NR_CPUS);
347
348 for (i = 0; i < NR_CPUS; i++, ptr += size) {
349 __per_cpu_offset[i] = ptr - __per_cpu_start;
350 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
351 }
352}
353#endif /* !__GENERIC_PER_CPU */
354
355/* Called by boot processor to activate the rest. */
356static void __init smp_init(void)
357{
358 unsigned int i;
359
360 /* FIXME: This should be done in userspace --RR */
361 for_each_present_cpu(i) {
362 if (num_online_cpus() >= max_cpus)
363 break;
364 if (!cpu_online(i))
365 cpu_up(i);
366 }
367
368 /* Any cleanup work */
369 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
370 smp_cpus_done(max_cpus);
371#if 0
372 /* Get other processors into their bootup holding patterns. */
373
374 smp_commence();
375#endif
376}
377
378#endif
379
380/*
381 * We need to finalize in a non-__init function or else race conditions
382 * between the root thread and the init thread may cause start_kernel to
383 * be reaped by free_initmem before the root thread has proceeded to
384 * cpu_idle.
385 *
386 * gcc-3.4 accidentally inlines this function, so use noinline.
387 */
388
389static void noinline rest_init(void)
390 __releases(kernel_lock)
391{
392 kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
393 numa_default_policy();
394 unlock_kernel();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200395
396 /*
397 * The boot idle thread must execute schedule()
398 * at least one to get things moving:
399 */
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800400 preempt_enable_no_resched();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200401 schedule();
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800402 preempt_disable();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200403
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800404 /* Call into cpu_idle with preempt disabled */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 cpu_idle();
406}
407
408/* Check for early params. */
409static int __init do_early_param(char *param, char *val)
410{
411 struct obs_kernel_param *p;
412
413 for (p = __setup_start; p < __setup_end; p++) {
414 if (p->early && strcmp(param, p->str) == 0) {
415 if (p->setup_func(val) != 0)
416 printk(KERN_WARNING
417 "Malformed early option '%s'\n", param);
418 }
419 }
420 /* We accept everything at this stage. */
421 return 0;
422}
423
424/* Arch code calls this early on, or if not, just before other parsing. */
425void __init parse_early_param(void)
426{
427 static __initdata int done = 0;
428 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
429
430 if (done)
431 return;
432
433 /* All fall through to do_early_param. */
434 strlcpy(tmp_cmdline, saved_command_line, COMMAND_LINE_SIZE);
435 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
436 done = 1;
437}
438
439/*
440 * Activate the first processor.
441 */
442
443asmlinkage void __init start_kernel(void)
444{
445 char * command_line;
446 extern struct kernel_param __start___param[], __stop___param[];
447/*
448 * Interrupts are still disabled. Do necessary setups, then
449 * enable them
450 */
451 lock_kernel();
452 page_address_init();
453 printk(KERN_NOTICE);
454 printk(linux_banner);
455 setup_arch(&command_line);
456 setup_per_cpu_areas();
457
458 /*
459 * Mark the boot cpu "online" so that it can call console drivers in
460 * printk() and can access its per-cpu storage.
461 */
462 smp_prepare_boot_cpu();
463
464 /*
465 * Set up the scheduler prior starting any interrupts (such as the
466 * timer interrupt). Full topology setup happens at smp_init()
467 * time - but meanwhile we still have a functioning scheduler.
468 */
469 sched_init();
470 /*
471 * Disable preemption - early bootup scheduling is extremely
472 * fragile until we cpu_idle() for the first time.
473 */
474 preempt_disable();
475 build_all_zonelists();
476 page_alloc_init();
477 printk(KERN_NOTICE "Kernel command line: %s\n", saved_command_line);
478 parse_early_param();
479 parse_args("Booting kernel", command_line, __start___param,
480 __stop___param - __start___param,
481 &unknown_bootoption);
482 sort_main_extable();
483 trap_init();
484 rcu_init();
485 init_IRQ();
486 pidhash_init();
487 init_timers();
488 softirq_init();
489 time_init();
490
491 /*
492 * HACK ALERT! This is early. We're enabling the console before
493 * we've done PCI setups etc, and console_init() must be aware of
494 * this. But we do want output early, in case something goes wrong.
495 */
496 console_init();
497 if (panic_later)
498 panic(panic_later, panic_param);
499 profile_init();
500 local_irq_enable();
501#ifdef CONFIG_BLK_DEV_INITRD
502 if (initrd_start && !initrd_below_start_ok &&
503 initrd_start < min_low_pfn << PAGE_SHIFT) {
504 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
505 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
506 initrd_start = 0;
507 }
508#endif
509 vfs_caches_init_early();
Paul Jacksonc417f022006-01-08 01:02:01 -0800510 cpuset_init_early();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 mem_init();
512 kmem_cache_init();
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700513 setup_per_cpu_pageset();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 numa_policy_init();
515 if (late_time_init)
516 late_time_init();
517 calibrate_delay();
518 pidmap_init();
519 pgtable_cache_init();
520 prio_tree_init();
521 anon_vma_init();
522#ifdef CONFIG_X86
523 if (efi_enabled)
524 efi_enter_virtual_mode();
525#endif
526 fork_init(num_physpages);
527 proc_caches_init();
528 buffer_init();
529 unnamed_dev_init();
530 key_init();
531 security_init();
532 vfs_caches_init(num_physpages);
533 radix_tree_init();
534 signals_init();
535 /* rootfs populating might need page-writeback */
536 page_writeback_init();
537#ifdef CONFIG_PROC_FS
538 proc_root_init();
539#endif
540 cpuset_init();
541
542 check_bugs();
543
544 acpi_early_init(); /* before LAPIC and SMP init */
545
546 /* Do the rest non-__init'ed, we're now alive */
547 rest_init();
548}
549
550static int __initdata initcall_debug;
551
552static int __init initcall_debug_setup(char *str)
553{
554 initcall_debug = 1;
555 return 1;
556}
557__setup("initcall_debug", initcall_debug_setup);
558
559struct task_struct *child_reaper = &init_task;
560
561extern initcall_t __initcall_start[], __initcall_end[];
562
563static void __init do_initcalls(void)
564{
565 initcall_t *call;
566 int count = preempt_count();
567
568 for (call = __initcall_start; call < __initcall_end; call++) {
569 char *msg;
570
571 if (initcall_debug) {
572 printk(KERN_DEBUG "Calling initcall 0x%p", *call);
573 print_fn_descriptor_symbol(": %s()", (unsigned long) *call);
574 printk("\n");
575 }
576
577 (*call)();
578
579 msg = NULL;
580 if (preempt_count() != count) {
581 msg = "preemption imbalance";
582 preempt_count() = count;
583 }
584 if (irqs_disabled()) {
585 msg = "disabled interrupts";
586 local_irq_enable();
587 }
588 if (msg) {
589 printk(KERN_WARNING "error in initcall at 0x%p: "
590 "returned with %s\n", *call, msg);
591 }
592 }
593
594 /* Make sure there is no pending stuff from the initcall sequence */
595 flush_scheduled_work();
596}
597
598/*
599 * Ok, the machine is now initialized. None of the devices
600 * have been touched yet, but the CPU subsystem is up and
601 * running, and memory and process management works.
602 *
603 * Now we can finally start doing some real work..
604 */
605static void __init do_basic_setup(void)
606{
607 /* drivers will send hotplug events */
608 init_workqueues();
609 usermodehelper_init();
610 driver_init();
611
612#ifdef CONFIG_SYSCTL
613 sysctl_init();
614#endif
615
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 do_initcalls();
617}
618
619static void do_pre_smp_initcalls(void)
620{
621 extern int spawn_ksoftirqd(void);
622#ifdef CONFIG_SMP
623 extern int migration_init(void);
624
625 migration_init();
626#endif
627 spawn_ksoftirqd();
Ingo Molnar8446f1d2005-09-06 15:16:27 -0700628 spawn_softlockup_task();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629}
630
631static void run_init_process(char *init_filename)
632{
633 argv_init[0] = init_filename;
634 execve(init_filename, argv_init, envp_init);
635}
636
637static inline void fixup_cpu_present_map(void)
638{
639#ifdef CONFIG_SMP
640 int i;
641
642 /*
643 * If arch is not hotplug ready and did not populate
644 * cpu_present_map, just make cpu_present_map same as cpu_possible_map
645 * for other cpu bringup code to function as normal. e.g smp_init() etc.
646 */
647 if (cpus_empty(cpu_present_map)) {
648 for_each_cpu(i) {
649 cpu_set(i, cpu_present_map);
650 }
651 }
652#endif
653}
654
655static int init(void * unused)
656{
657 lock_kernel();
658 /*
659 * init can run on any cpu.
660 */
661 set_cpus_allowed(current, CPU_MASK_ALL);
662 /*
663 * Tell the world that we're going to be the grim
664 * reaper of innocent orphaned children.
665 *
666 * We don't want people to have to make incorrect
667 * assumptions about where in the task array this
668 * can be found.
669 */
670 child_reaper = current;
671
672 /* Sets up cpus_possible() */
673 smp_prepare_cpus(max_cpus);
674
675 do_pre_smp_initcalls();
676
677 fixup_cpu_present_map();
678 smp_init();
679 sched_init_smp();
680
681 cpuset_init_smp();
682
683 /*
684 * Do this before initcalls, because some drivers want to access
685 * firmware files.
686 */
687 populate_rootfs();
688
689 do_basic_setup();
690
691 /*
692 * check if there is an early userspace init. If yes, let it do all
693 * the work
694 */
Olof Johanssonffdfc402005-09-06 15:17:19 -0700695
696 if (!ramdisk_execute_command)
697 ramdisk_execute_command = "/init";
698
699 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
700 ramdisk_execute_command = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 prepare_namespace();
Olof Johanssonffdfc402005-09-06 15:17:19 -0700702 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703
704 /*
705 * Ok, we have completed the initial bootup, and
706 * we're essentially up and running. Get rid of the
707 * initmem segments and start the user-mode stuff..
708 */
709 free_initmem();
710 unlock_kernel();
Arjan van de Ven37b73c82006-01-06 00:12:01 -0800711 mark_rodata_ro();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 system_state = SYSTEM_RUNNING;
713 numa_default_policy();
714
715 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
716 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
717
718 (void) sys_dup(0);
719 (void) sys_dup(0);
Olof Johanssonffdfc402005-09-06 15:17:19 -0700720
721 if (ramdisk_execute_command) {
722 run_init_process(ramdisk_execute_command);
723 printk(KERN_WARNING "Failed to execute %s\n",
724 ramdisk_execute_command);
725 }
726
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 /*
728 * We try each of these until one succeeds.
729 *
730 * The Bourne shell can be used instead of init if we are
731 * trying to recover a really broken machine.
732 */
Avery, Brianc1d7ef72005-09-06 15:16:56 -0700733 if (execute_command) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 run_init_process(execute_command);
Avery, Brianc1d7ef72005-09-06 15:16:56 -0700735 printk(KERN_WARNING "Failed to execute %s. Attempting "
736 "defaults...\n", execute_command);
737 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 run_init_process("/sbin/init");
739 run_init_process("/etc/init");
740 run_init_process("/bin/init");
741 run_init_process("/bin/sh");
742
743 panic("No init found. Try passing init= option to kernel.");
744}