blob: c691f5f7fc27b40ea6b2df39e2fef9ac52b4cd0a [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/types.h>
13#include <linux/module.h>
14#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/kernel.h>
16#include <linux/syscalls.h>
17#include <linux/string.h>
18#include <linux/ctype.h>
19#include <linux/delay.h>
20#include <linux/utsname.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/hdreg.h>
26#include <linux/bootmem.h>
27#include <linux/tty.h>
28#include <linux/gfp.h>
29#include <linux/percpu.h>
30#include <linux/kmod.h>
31#include <linux/kernel_stat.h>
Rusty Russelld7cd5612006-12-07 02:14:08 +010032#include <linux/start_kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <linux/security.h>
34#include <linux/workqueue.h>
35#include <linux/profile.h>
36#include <linux/rcupdate.h>
37#include <linux/moduleparam.h>
38#include <linux/kallsyms.h>
39#include <linux/writeback.h>
40#include <linux/cpu.h>
41#include <linux/cpuset.h>
Paul Menageddbcc7e2007-10-18 23:39:30 -070042#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <linux/efi.h>
Thomas Gleixner906568c2007-02-16 01:28:01 -080044#include <linux/tick.h>
Andrew Morton6168a702007-02-17 21:22:39 -080045#include <linux/interrupt.h>
Shailabh Nagarc7572492006-07-14 00:24:40 -070046#include <linux/taskstats_kern.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070047#include <linux/delayacct.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#include <linux/unistd.h>
49#include <linux/rmap.h>
50#include <linux/mempolicy.h>
51#include <linux/key.h>
Jan Beulich4552d5d2006-06-26 13:57:28 +020052#include <linux/unwind.h>
Adrian Bunkb6cd0b72006-06-27 02:53:54 -070053#include <linux/buffer_head.h>
Ingo Molnar9a11b49a2006-07-03 00:24:33 -070054#include <linux/debug_locks.h>
Ingo Molnarfbb9ce952006-07-03 00:24:50 -070055#include <linux/lockdep.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080056#include <linux/pid_namespace.h>
Adrian Bunk1f217822006-12-19 13:01:28 -080057#include <linux/device.h>
Eric W. Biederman73c27992007-05-09 02:34:32 -070058#include <linux/kthread.h>
Adrian Bunke6fe6642007-11-09 22:39:39 +010059#include <linux/sched.h>
Adrian Bunka1c9eea2008-02-06 01:36:44 -080060#include <linux/signal.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
62#include <asm/io.h>
63#include <asm/bugs.h>
64#include <asm/setup.h>
Andi Kleena9401992005-07-28 21:15:30 -070065#include <asm/sections.h>
Arjan van de Ven37b73c82006-01-06 00:12:01 -080066#include <asm/cacheflush.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#ifdef CONFIG_X86_LOCAL_APIC
69#include <asm/smp.h>
70#endif
71
72/*
Coywolf Qi Hunt69c99ac2006-01-14 13:21:11 -080073 * This is one of the first .c files built. Error out early if we have compiler
74 * trouble.
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070076
Andrew Mortonf1a60db2006-12-06 20:39:59 -080077#if __GNUC__ == 4 && __GNUC_MINOR__ == 1 && __GNUC_PATCHLEVEL__ == 0
78#warning gcc-4.1.0 is known to miscompile the kernel. A different compiler version is recommended.
79#endif
80
Sam Ravnborgaae5f662007-02-26 16:45:41 +010081static int kernel_init(void *);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
83extern void init_IRQ(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070084extern void fork_init(unsigned long);
85extern void mca_init(void);
86extern void sbus_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070087extern void pidhash_init(void);
88extern void pidmap_init(void);
89extern void prio_tree_init(void);
90extern void radix_tree_init(void);
91extern void free_initmem(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092#ifdef CONFIG_ACPI
93extern void acpi_early_init(void);
94#else
95static inline void acpi_early_init(void) { }
96#endif
Arjan van de Ven37b73c82006-01-06 00:12:01 -080097#ifndef CONFIG_DEBUG_RODATA
98static inline void mark_rodata_ro(void) { }
99#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
101#ifdef CONFIG_TC
102extern void tc_init(void);
103#endif
104
105enum system_states system_state;
106EXPORT_SYMBOL(system_state);
107
108/*
109 * Boot command-line arguments
110 */
111#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
112#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
113
114extern void time_init(void);
115/* Default late time init is NULL. archs can override this later. */
116void (*late_time_init)(void);
117extern void softirq_init(void);
118
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800119/* Untouched command line saved by arch-specific code. */
120char __initdata boot_command_line[COMMAND_LINE_SIZE];
121/* Untouched saved command line (eg. for /proc) */
122char *saved_command_line;
123/* Command line for parameter parsing */
124static char *static_command_line;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126static char *execute_command;
Olof Johanssonffdfc402005-09-06 15:17:19 -0700127static char *ramdisk_execute_command;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
Jan Beulich8b3b2952007-07-15 23:41:07 -0700129#ifdef CONFIG_SMP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/* Setup configured maximum number of CPUs to activate */
Andi Kleenca74a6f2008-01-30 13:33:17 +0100131unsigned int __initdata setup_max_cpus = NR_CPUS;
Vivek Goyal7e962872006-09-27 01:50:44 -0700132
133/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 * Setup routine for controlling SMP activation
135 *
136 * Command-line option of "nosmp" or "maxcpus=0" will disable SMP
137 * activation entirely (the MPS table probe still happens, though).
138 *
139 * Command-line option of "maxcpus=<NUM>", where <NUM> is an integer
140 * greater than 0, limits the maximum number of CPUs activated in
141 * SMP mode to <NUM>.
142 */
Len Brown61ec7562007-08-16 03:34:22 -0400143#ifndef CONFIG_X86_IO_APIC
144static inline void disable_ioapic_setup(void) {};
145#endif
146
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147static int __init nosmp(char *str)
148{
Andi Kleenca74a6f2008-01-30 13:33:17 +0100149 setup_max_cpus = 0;
Len Brown61ec7562007-08-16 03:34:22 -0400150 disable_ioapic_setup();
Jan Beulich8b3b2952007-07-15 23:41:07 -0700151 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152}
153
Jan Beulich8b3b2952007-07-15 23:41:07 -0700154early_param("nosmp", nosmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
156static int __init maxcpus(char *str)
157{
Andi Kleenca74a6f2008-01-30 13:33:17 +0100158 get_option(&str, &setup_max_cpus);
159 if (setup_max_cpus == 0)
Len Brown61ec7562007-08-16 03:34:22 -0400160 disable_ioapic_setup();
161
162 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163}
164
Hugh Dickins81340972007-08-27 16:02:12 +0100165early_param("maxcpus", maxcpus);
Jan Beulich8b3b2952007-07-15 23:41:07 -0700166#else
Andi Kleenca74a6f2008-01-30 13:33:17 +0100167#define setup_max_cpus NR_CPUS
Jan Beulich8b3b2952007-07-15 23:41:07 -0700168#endif
169
170/*
171 * If set, this is an indication to the drivers that reset the underlying
172 * device before going ahead with the initialization otherwise driver might
173 * rely on the BIOS and skip the reset operation.
174 *
175 * This is useful if kernel is booting in an unreliable environment.
176 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
177 * skipped and devices will be in unknown state.
178 */
179unsigned int reset_devices;
180EXPORT_SYMBOL(reset_devices);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181
Vivek Goyal7e962872006-09-27 01:50:44 -0700182static int __init set_reset_devices(char *str)
183{
184 reset_devices = 1;
185 return 1;
186}
187
188__setup("reset_devices", set_reset_devices);
189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190static char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
191char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
192static const char *panic_later, *panic_param;
193
194extern struct obs_kernel_param __setup_start[], __setup_end[];
195
196static int __init obsolete_checksetup(char *line)
197{
198 struct obs_kernel_param *p;
Rusty Russell33df0d12006-09-26 10:52:32 +0200199 int had_early_param = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200
201 p = __setup_start;
202 do {
203 int n = strlen(p->str);
204 if (!strncmp(line, p->str, n)) {
205 if (p->early) {
Rusty Russell33df0d12006-09-26 10:52:32 +0200206 /* Already done in parse_early_param?
207 * (Needs exact match on param part).
208 * Keep iterating, as we can have early
209 * params and __setups of same names 8( */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 if (line[n] == '\0' || line[n] == '=')
Rusty Russell33df0d12006-09-26 10:52:32 +0200211 had_early_param = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 } else if (!p->setup_func) {
213 printk(KERN_WARNING "Parameter %s is obsolete,"
214 " ignored\n", p->str);
215 return 1;
216 } else if (p->setup_func(line + n))
217 return 1;
218 }
219 p++;
220 } while (p < __setup_end);
Rusty Russell33df0d12006-09-26 10:52:32 +0200221
222 return had_early_param;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
224
225/*
226 * This should be approx 2 Bo*oMips to start (note initial shift), and will
227 * still work even if initially too large, it will just take slightly longer
228 */
229unsigned long loops_per_jiffy = (1<<12);
230
231EXPORT_SYMBOL(loops_per_jiffy);
232
233static int __init debug_kernel(char *str)
234{
235 if (*str)
236 return 0;
237 console_loglevel = 10;
238 return 1;
239}
240
241static int __init quiet_kernel(char *str)
242{
243 if (*str)
244 return 0;
245 console_loglevel = 4;
246 return 1;
247}
248
249__setup("debug", debug_kernel);
250__setup("quiet", quiet_kernel);
251
252static int __init loglevel(char *str)
253{
254 get_option(&str, &console_loglevel);
255 return 1;
256}
257
258__setup("loglevel=", loglevel);
259
260/*
261 * Unknown boot options get handed to init, unless they look like
262 * failed parameters
263 */
264static int __init unknown_bootoption(char *param, char *val)
265{
266 /* Change NUL term back to "=", to make "param" the whole string. */
267 if (val) {
268 /* param=val or param="val"? */
269 if (val == param+strlen(param)+1)
270 val[-1] = '=';
271 else if (val == param+strlen(param)+2) {
272 val[-2] = '=';
273 memmove(val-1, val, strlen(val)+1);
274 val--;
275 } else
276 BUG();
277 }
278
279 /* Handle obsolete-style parameters */
280 if (obsolete_checksetup(param))
281 return 0;
282
283 /*
Simon Arlott211fee82007-10-20 01:28:29 +0200284 * Preemptive maintenance for "why didn't my misspelled command
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 * line work?"
286 */
287 if (strchr(param, '.') && (!val || strchr(param, '.') < val)) {
288 printk(KERN_ERR "Unknown boot option `%s': ignoring\n", param);
289 return 0;
290 }
291
292 if (panic_later)
293 return 0;
294
295 if (val) {
296 /* Environment option */
297 unsigned int i;
298 for (i = 0; envp_init[i]; i++) {
299 if (i == MAX_INIT_ENVS) {
300 panic_later = "Too many boot env vars at `%s'";
301 panic_param = param;
302 }
303 if (!strncmp(param, envp_init[i], val - param))
304 break;
305 }
306 envp_init[i] = param;
307 } else {
308 /* Command line option */
309 unsigned int i;
310 for (i = 0; argv_init[i]; i++) {
311 if (i == MAX_INIT_ARGS) {
312 panic_later = "Too many boot init vars at `%s'";
313 panic_param = param;
314 }
315 }
316 argv_init[i] = param;
317 }
318 return 0;
319}
320
Ingo Molnar12d6f212008-01-30 13:33:58 +0100321#ifdef CONFIG_DEBUG_PAGEALLOC
322int __read_mostly debug_pagealloc_enabled = 0;
323#endif
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325static int __init init_setup(char *str)
326{
327 unsigned int i;
328
329 execute_command = str;
330 /*
331 * In case LILO is going to boot us with default command line,
332 * it prepends "auto" before the whole cmdline which makes
333 * the shell think it should execute a script with such name.
334 * So we ignore all arguments entered _before_ init=... [MJ]
335 */
336 for (i = 1; i < MAX_INIT_ARGS; i++)
337 argv_init[i] = NULL;
338 return 1;
339}
340__setup("init=", init_setup);
341
Olof Johanssonffdfc402005-09-06 15:17:19 -0700342static int __init rdinit_setup(char *str)
343{
344 unsigned int i;
345
346 ramdisk_execute_command = str;
347 /* See "auto" comment in init_setup */
348 for (i = 1; i < MAX_INIT_ARGS; i++)
349 argv_init[i] = NULL;
350 return 1;
351}
352__setup("rdinit=", rdinit_setup);
353
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354#ifndef CONFIG_SMP
355
356#ifdef CONFIG_X86_LOCAL_APIC
357static void __init smp_init(void)
358{
359 APIC_init_uniprocessor();
360}
361#else
362#define smp_init() do { } while (0)
363#endif
364
365static inline void setup_per_cpu_areas(void) { }
366static inline void smp_prepare_cpus(unsigned int maxcpus) { }
367
368#else
369
Mike Travisdd5af902008-01-30 13:33:32 +0100370#ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
Eric Dumazetb73b4592006-03-23 03:01:07 -0800371unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
373EXPORT_SYMBOL(__per_cpu_offset);
374
375static void __init setup_per_cpu_areas(void)
376{
377 unsigned long size, i;
378 char *ptr;
Eric Dumazet63872f82006-03-23 03:01:04 -0800379 unsigned long nr_possible_cpus = num_possible_cpus();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
381 /* Copy section for each CPU (we discard the original) */
Jeremy Fitzhardingeb6e35902007-05-02 19:27:12 +0200382 size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
383 ptr = alloc_bootmem_pages(size * nr_possible_cpus);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
KAMEZAWA Hiroyuki0a945022006-03-28 01:56:37 -0800385 for_each_possible_cpu(i) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 __per_cpu_offset[i] = ptr - __per_cpu_start;
387 memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
Eric Dumazet63872f82006-03-23 03:01:04 -0800388 ptr += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 }
390}
Mike Travisdd5af902008-01-30 13:33:32 +0100391#endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
393/* Called by boot processor to activate the rest. */
394static void __init smp_init(void)
395{
Christoph Lameter53b8a312007-02-20 13:57:51 -0800396 unsigned int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
398 /* FIXME: This should be done in userspace --RR */
Christoph Lameter53b8a312007-02-20 13:57:51 -0800399 for_each_present_cpu(cpu) {
Andi Kleenca74a6f2008-01-30 13:33:17 +0100400 if (num_online_cpus() >= setup_max_cpus)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 break;
Christoph Lameter53b8a312007-02-20 13:57:51 -0800402 if (!cpu_online(cpu))
403 cpu_up(cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 }
405
406 /* Any cleanup work */
407 printk(KERN_INFO "Brought up %ld CPUs\n", (long)num_online_cpus());
Andi Kleenca74a6f2008-01-30 13:33:17 +0100408 smp_cpus_done(setup_max_cpus);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409}
410
411#endif
412
413/*
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800414 * We need to store the untouched command line for future reference.
415 * We also need to store the touched command line since the parameter
416 * parsing is performed in place, and we should allow a component to
417 * store reference of name/value for future reference.
418 */
419static void __init setup_command_line(char *command_line)
420{
421 saved_command_line = alloc_bootmem(strlen (boot_command_line)+1);
422 static_command_line = alloc_bootmem(strlen (command_line)+1);
423 strcpy (saved_command_line, boot_command_line);
424 strcpy (static_command_line, command_line);
425}
426
427/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 * We need to finalize in a non-__init function or else race conditions
429 * between the root thread and the init thread may cause start_kernel to
430 * be reaped by free_initmem before the root thread has proceeded to
431 * cpu_idle.
432 *
433 * gcc-3.4 accidentally inlines this function, so use noinline.
434 */
435
Sam Ravnborg92080302007-05-17 20:43:54 +0200436static void noinline __init_refok rest_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 __releases(kernel_lock)
438{
Eric W. Biederman73c27992007-05-09 02:34:32 -0700439 int pid;
440
Sam Ravnborgaae5f662007-02-26 16:45:41 +0100441 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 numa_default_policy();
Eric W. Biederman73c27992007-05-09 02:34:32 -0700443 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
444 kthreadd_task = find_task_by_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 unlock_kernel();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200446
447 /*
448 * The boot idle thread must execute schedule()
Ingo Molnar1df21052007-07-09 18:51:58 +0200449 * at least once to get things moving:
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200450 */
Ingo Molnar1df21052007-07-09 18:51:58 +0200451 init_idle_bootup_task(current);
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800452 preempt_enable_no_resched();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200453 schedule();
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800454 preempt_disable();
Ingo Molnarf340c0d2005-06-28 16:40:42 +0200455
Nick Piggin5bfb5d62005-11-08 21:39:01 -0800456 /* Call into cpu_idle with preempt disabled */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 cpu_idle();
Ingo Molnar1df21052007-07-09 18:51:58 +0200458}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459
460/* Check for early params. */
461static int __init do_early_param(char *param, char *val)
462{
463 struct obs_kernel_param *p;
464
465 for (p = __setup_start; p < __setup_end; p++) {
Yinghai Lu18a8bd92007-07-15 23:37:59 -0700466 if ((p->early && strcmp(param, p->str) == 0) ||
467 (strcmp(param, "console") == 0 &&
468 strcmp(p->str, "earlycon") == 0)
469 ) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 if (p->setup_func(val) != 0)
471 printk(KERN_WARNING
472 "Malformed early option '%s'\n", param);
473 }
474 }
475 /* We accept everything at this stage. */
476 return 0;
477}
478
479/* Arch code calls this early on, or if not, just before other parsing. */
480void __init parse_early_param(void)
481{
482 static __initdata int done = 0;
483 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
484
485 if (done)
486 return;
487
488 /* All fall through to do_early_param. */
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800489 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 parse_args("early options", tmp_cmdline, NULL, 0, do_early_param);
491 done = 1;
492}
493
494/*
495 * Activate the first processor.
496 */
497
Stas Sergeev44fd2292006-03-23 02:59:44 -0800498static void __init boot_cpu_init(void)
499{
500 int cpu = smp_processor_id();
501 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
502 cpu_set(cpu, cpu_online_map);
503 cpu_set(cpu, cpu_present_map);
504 cpu_set(cpu, cpu_possible_map);
505}
506
Andrew Morton033ab7f2006-06-30 01:55:50 -0700507void __init __attribute__((weak)) smp_setup_processor_id(void)
508{
509}
510
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511asmlinkage void __init start_kernel(void)
512{
513 char * command_line;
514 extern struct kernel_param __start___param[], __stop___param[];
Andrew Morton033ab7f2006-06-30 01:55:50 -0700515
516 smp_setup_processor_id();
517
Ingo Molnarfbb9ce952006-07-03 00:24:50 -0700518 /*
519 * Need to run as early as possible, to initialize the
520 * lockdep hash:
521 */
Andi Kleenc9538ed2006-09-26 10:52:34 +0200522 unwind_init();
Ingo Molnarfbb9ce952006-07-03 00:24:50 -0700523 lockdep_init();
Paul Menageddbcc7e2007-10-18 23:39:30 -0700524 cgroup_init_early();
Ingo Molnarfbb9ce952006-07-03 00:24:50 -0700525
526 local_irq_disable();
527 early_boot_irqs_off();
Ingo Molnar243c7622006-07-03 00:25:06 -0700528 early_init_irq_lock_class();
Ingo Molnarfbb9ce952006-07-03 00:24:50 -0700529
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530/*
531 * Interrupts are still disabled. Do necessary setups, then
532 * enable them
533 */
534 lock_kernel();
Thomas Gleixner906568c2007-02-16 01:28:01 -0800535 tick_init();
Stas Sergeev44fd2292006-03-23 02:59:44 -0800536 boot_cpu_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 page_address_init();
538 printk(KERN_NOTICE);
Linus Torvalds89937802006-12-11 09:28:46 -0800539 printk(linux_banner);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 setup_arch(&command_line);
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800541 setup_command_line(command_line);
Jan Beulich690a9732006-10-21 18:37:01 +0200542 unwind_setup();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 setup_per_cpu_areas();
Stas Sergeev44fd2292006-03-23 02:59:44 -0800544 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
546 /*
547 * Set up the scheduler prior starting any interrupts (such as the
548 * timer interrupt). Full topology setup happens at smp_init()
549 * time - but meanwhile we still have a functioning scheduler.
550 */
551 sched_init();
552 /*
553 * Disable preemption - early bootup scheduling is extremely
554 * fragile until we cpu_idle() for the first time.
555 */
556 preempt_disable();
557 build_all_zonelists();
558 page_alloc_init();
Ingo Molnar12d6f212008-01-30 13:33:58 +0100559 enable_debug_pagealloc();
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800560 printk(KERN_NOTICE "Kernel command line: %s\n", boot_command_line);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 parse_early_param();
Alon Bar-Lev30d7e0d2007-02-12 00:53:52 -0800562 parse_args("Booting kernel", static_command_line, __start___param,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 __stop___param - __start___param,
564 &unknown_bootoption);
Ard van Breemenc4a68302007-01-05 16:36:19 -0800565 if (!irqs_disabled()) {
566 printk(KERN_WARNING "start_kernel(): bug: interrupts were "
567 "enabled *very* early, fixing it\n");
568 local_irq_disable();
569 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 sort_main_extable();
571 trap_init();
572 rcu_init();
573 init_IRQ();
574 pidhash_init();
575 init_timers();
Thomas Gleixnerc0a31322006-01-09 20:52:32 -0800576 hrtimers_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 softirq_init();
john stultzad596172006-06-26 00:25:06 -0700578 timekeeping_init();
john stultz88fecaa2006-07-03 00:24:04 -0700579 time_init();
Heiko Carstens93e02812006-07-03 00:24:24 -0700580 profile_init();
581 if (!irqs_disabled())
582 printk("start_kernel(): bug: interrupts were enabled early\n");
Ingo Molnarfbb9ce952006-07-03 00:24:50 -0700583 early_boot_irqs_on();
Heiko Carstens93e02812006-07-03 00:24:24 -0700584 local_irq_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
586 /*
587 * HACK ALERT! This is early. We're enabling the console before
588 * we've done PCI setups etc, and console_init() must be aware of
589 * this. But we do want output early, in case something goes wrong.
590 */
591 console_init();
592 if (panic_later)
593 panic(panic_later, panic_param);
Ingo Molnarfbb9ce952006-07-03 00:24:50 -0700594
595 lockdep_info();
596
Ingo Molnar9a11b49a2006-07-03 00:24:33 -0700597 /*
598 * Need to run this when irqs are enabled, because it wants
599 * to self-test [hard/soft]-irqs on/off lock inversion bugs
600 * too:
601 */
602 locking_selftest();
603
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604#ifdef CONFIG_BLK_DEV_INITRD
605 if (initrd_start && !initrd_below_start_ok &&
606 initrd_start < min_low_pfn << PAGE_SHIFT) {
607 printk(KERN_CRIT "initrd overwritten (0x%08lx < 0x%08lx) - "
608 "disabling it.\n",initrd_start,min_low_pfn << PAGE_SHIFT);
609 initrd_start = 0;
610 }
611#endif
612 vfs_caches_init_early();
Paul Jacksonc417f022006-01-08 01:02:01 -0800613 cpuset_init_early();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 mem_init();
Gautham R Shenoyd2219382008-01-25 21:08:01 +0100615 cpu_hotplug_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 kmem_cache_init();
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700617 setup_per_cpu_pageset();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 numa_policy_init();
619 if (late_time_init)
620 late_time_init();
621 calibrate_delay();
622 pidmap_init();
623 pgtable_cache_init();
624 prio_tree_init();
625 anon_vma_init();
626#ifdef CONFIG_X86
627 if (efi_enabled)
628 efi_enter_virtual_mode();
629#endif
630 fork_init(num_physpages);
631 proc_caches_init();
632 buffer_init();
633 unnamed_dev_init();
634 key_init();
635 security_init();
636 vfs_caches_init(num_physpages);
637 radix_tree_init();
638 signals_init();
639 /* rootfs populating might need page-writeback */
640 page_writeback_init();
641#ifdef CONFIG_PROC_FS
642 proc_root_init();
643#endif
Paul Menageddbcc7e2007-10-18 23:39:30 -0700644 cgroup_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 cpuset_init();
Shailabh Nagarc7572492006-07-14 00:24:40 -0700646 taskstats_init_early();
Shailabh Nagarca74e922006-07-14 00:24:36 -0700647 delayacct_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
649 check_bugs();
650
651 acpi_early_init(); /* before LAPIC and SMP init */
652
653 /* Do the rest non-__init'ed, we're now alive */
654 rest_init();
655}
656
657static int __initdata initcall_debug;
658
659static int __init initcall_debug_setup(char *str)
660{
661 initcall_debug = 1;
662 return 1;
663}
664__setup("initcall_debug", initcall_debug_setup);
665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666extern initcall_t __initcall_start[], __initcall_end[];
667
668static void __init do_initcalls(void)
669{
670 initcall_t *call;
671 int count = preempt_count();
672
673 for (call = __initcall_start; call < __initcall_end; call++) {
Ingo Molnar8f0c45c2007-05-08 00:28:26 -0700674 ktime_t t0, t1, delta;
Andrew Mortonc1cda482006-03-25 03:07:15 -0800675 char *msg = NULL;
676 char msgbuf[40];
677 int result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678
679 if (initcall_debug) {
Andrew Mortonc1cda482006-03-25 03:07:15 -0800680 printk("Calling initcall 0x%p", *call);
681 print_fn_descriptor_symbol(": %s()",
682 (unsigned long) *call);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 printk("\n");
Ingo Molnar8f0c45c2007-05-08 00:28:26 -0700684 t0 = ktime_get();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 }
686
Andrew Mortonc1cda482006-03-25 03:07:15 -0800687 result = (*call)();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
Ingo Molnar8f0c45c2007-05-08 00:28:26 -0700689 if (initcall_debug) {
690 t1 = ktime_get();
691 delta = ktime_sub(t1, t0);
692
693 printk("initcall 0x%p", *call);
694 print_fn_descriptor_symbol(": %s()",
695 (unsigned long) *call);
696 printk(" returned %d.\n", result);
697
698 printk("initcall 0x%p ran for %Ld msecs: ",
699 *call, (unsigned long long)delta.tv64 >> 20);
700 print_fn_descriptor_symbol("%s()\n",
701 (unsigned long) *call);
702 }
703
Andrew Mortonf3a19cb2006-05-01 12:15:52 -0700704 if (result && result != -ENODEV && initcall_debug) {
Andrew Mortonc1cda482006-03-25 03:07:15 -0800705 sprintf(msgbuf, "error code %d", result);
706 msg = msgbuf;
707 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 if (preempt_count() != count) {
709 msg = "preemption imbalance";
710 preempt_count() = count;
711 }
712 if (irqs_disabled()) {
713 msg = "disabled interrupts";
714 local_irq_enable();
715 }
716 if (msg) {
Andrew Mortonc1cda482006-03-25 03:07:15 -0800717 printk(KERN_WARNING "initcall at 0x%p", *call);
718 print_fn_descriptor_symbol(": %s()",
719 (unsigned long) *call);
720 printk(": returned with %s\n", msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 }
722 }
723
724 /* Make sure there is no pending stuff from the initcall sequence */
725 flush_scheduled_work();
726}
727
728/*
729 * Ok, the machine is now initialized. None of the devices
730 * have been touched yet, but the CPU subsystem is up and
731 * running, and memory and process management works.
732 *
733 * Now we can finally start doing some real work..
734 */
735static void __init do_basic_setup(void)
736{
737 /* drivers will send hotplug events */
738 init_workqueues();
739 usermodehelper_init();
740 driver_init();
Eric W. Biedermanb04c3af2007-02-14 00:33:57 -0800741 init_irq_proc();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 do_initcalls();
743}
744
Dave Jones97842212007-07-15 23:41:05 -0700745static int __initdata nosoftlockup;
746
747static int __init nosoftlockup_setup(char *str)
748{
749 nosoftlockup = 1;
750 return 1;
751}
752__setup("nosoftlockup", nosoftlockup_setup);
753
Vivek Goyal88d20322007-01-11 01:52:44 +0100754static void __init do_pre_smp_initcalls(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
756 extern int spawn_ksoftirqd(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
758 migration_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 spawn_ksoftirqd();
Dave Jones97842212007-07-15 23:41:05 -0700760 if (!nosoftlockup)
761 spawn_softlockup_task();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762}
763
764static void run_init_process(char *init_filename)
765{
766 argv_init[0] = init_filename;
Arnd Bergmann67608562006-10-02 02:18:26 -0700767 kernel_execve(init_filename, argv_init, envp_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768}
769
Vivek Goyalee5bfa62007-02-13 13:26:22 +0100770/* This is a non __init function. Force it to be noinline otherwise gcc
771 * makes it inline to init() and it becomes part of init.text section
772 */
773static int noinline init_post(void)
774{
775 free_initmem();
776 unlock_kernel();
777 mark_rodata_ro();
778 system_state = SYSTEM_RUNNING;
779 numa_default_policy();
780
781 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
782 printk(KERN_WARNING "Warning: unable to open an initial console.\n");
783
784 (void) sys_dup(0);
785 (void) sys_dup(0);
786
787 if (ramdisk_execute_command) {
788 run_init_process(ramdisk_execute_command);
789 printk(KERN_WARNING "Failed to execute %s\n",
790 ramdisk_execute_command);
791 }
792
793 /*
794 * We try each of these until one succeeds.
795 *
796 * The Bourne shell can be used instead of init if we are
797 * trying to recover a really broken machine.
798 */
799 if (execute_command) {
800 run_init_process(execute_command);
801 printk(KERN_WARNING "Failed to execute %s. Attempting "
802 "defaults...\n", execute_command);
803 }
804 run_init_process("/sbin/init");
805 run_init_process("/etc/init");
806 run_init_process("/bin/init");
807 run_init_process("/bin/sh");
808
809 panic("No init found. Try passing init= option to kernel.");
810}
811
Sam Ravnborgaae5f662007-02-26 16:45:41 +0100812static int __init kernel_init(void * unused)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813{
814 lock_kernel();
815 /*
816 * init can run on any cpu.
817 */
818 set_cpus_allowed(current, CPU_MASK_ALL);
819 /*
820 * Tell the world that we're going to be the grim
821 * reaper of innocent orphaned children.
822 *
823 * We don't want people to have to make incorrect
824 * assumptions about where in the task array this
825 * can be found.
826 */
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -0800827 init_pid_ns.child_reaper = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
Sukadev Bhattiprolu0e29b242007-05-10 22:23:01 -0700829 __set_special_pids(1, 1);
Cedric Le Goater9ec52092006-10-02 02:19:00 -0700830 cad_pid = task_pid(current);
831
Andi Kleenca74a6f2008-01-30 13:33:17 +0100832 smp_prepare_cpus(setup_max_cpus);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
834 do_pre_smp_initcalls();
835
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 smp_init();
837 sched_init_smp();
838
839 cpuset_init_smp();
840
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 do_basic_setup();
842
843 /*
844 * check if there is an early userspace init. If yes, let it do all
845 * the work
846 */
Olof Johanssonffdfc402005-09-06 15:17:19 -0700847
848 if (!ramdisk_execute_command)
849 ramdisk_execute_command = "/init";
850
851 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
852 ramdisk_execute_command = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 prepare_namespace();
Olof Johanssonffdfc402005-09-06 15:17:19 -0700854 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
856 /*
857 * Ok, we have completed the initial bootup, and
858 * we're essentially up and running. Get rid of the
859 * initmem segments and start the user-mode stuff..
860 */
Vivek Goyalee5bfa62007-02-13 13:26:22 +0100861 init_post();
862 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863}